# Shubi: full company reference > Shubi is an AI visibility and discovery company. We help brands get found, understood, cited and recommended by AI assistants like ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Based in Malaysia, working across Asia. This file is the long form of https://tryshubi.com/llms.txt. It is generated from the same content as the website. ## Company - Name: Shubi (legal and domain name: Tryshubi) - Website: https://tryshubi.com/ - Category: AI visibility and discovery: generative engine optimization (GEO), large language model optimization (LLMO), answer engine optimization (AEO), brand authority and entity optimization - Base: Malaysia - Markets: Malaysia, Singapore, Indonesia, Philippines, Thailand, Vietnam, India - AI platforms tracked: ChatGPT, Google Gemini, Perplexity, Claude, Google AI Overviews, Microsoft Copilot - Contact: WhatsApp +60 11-3763 4679 (https://wa.me/601137634679) or https://tryshubi.com/contact/. Shubi does not take enquiries by email. - Principle: Shubi does not guarantee AI mentions, citations or rankings. It guarantees the work, a clear method and transparent measurement with sample sizes. Shubi helps businesses become discoverable, accurately understood, trusted and recommended in AI-powered search and discovery: AI assistants such as ChatGPT, Gemini, Claude and Perplexity, and AI search features such as Google AI Overviews and Microsoft Copilot. ## Geographic scope Shubi is based in Malaysia and works with brands across Asia: Malaysia, Singapore, Indonesia, Philippines, Thailand, Vietnam, India. Buyer questions are tracked in the market and language buyers actually use. ## Services ### AI Visibility Intelligence URL: https://tryshubi.com/services/#intelligence We find the questions your buyers ask AI assistants before they buy, test them across the major AI engines in your market and language, and show where you appear, where you do not, how you are described and who is recommended instead. Includes: - Buyer question research by market and language - Visibility and share of voice across 5+ AI engines - Competitor benchmarking - Citation source mapping: which sites AI trusts in your category - Sentiment and accuracy of how AI describes you ### GEO and Content Engineering URL: https://tryshubi.com/services/#content Generative engine optimization (GEO), answer engine optimization (AEO) and LLM optimization (LLMO) applied to your site. We restructure key pages and create new ones so that AI systems can extract a clear, accurate answer and cite you as the source. Includes: - Direct-answer formatting: the answer first, then the detail - Comparison pages, category explainers and buyer guides - Original data, expert commentary and evidence - Content refreshes for pages AI already reads - Multilingual content for Asian markets ### Brand Authority Engineering URL: https://tryshubi.com/services/#authority AI assistants lean heavily on third-party sources: publications, reviews, directories, forums and communities. We decide how your brand should be described, then place consistent, credible signals in the places AI reads. Includes: - Positioning: how AI should describe your brand - Digital PR and publication placements - Review platform and directory presence - Community answers where buyers ask questions - Outreach to sites that mention competitors but not you ### AI Technical Infrastructure URL: https://tryshubi.com/services/#technical The technical foundation that lets AI crawlers access your pages and identify your brand without confusion, plus the tracking to see what AI sends you. Includes: - Structured data: Organization, Service, Product, FAQ, Person - Entity consistency across Wikidata, LinkedIn, Crunchbase and directories - AI crawler access, robots rules and llms.txt - Bing indexing, which several AI assistants rely on - AI referral traffic and lead tracking in GA4 - AI agent readiness for shopping and booking flows ### AI Visibility Growth URL: https://tryshubi.com/services/#growth Our retainer combines all four capabilities above into one monthly program: more buyer questions covered, more markets, more trusted sources, measured the same way every month so you can see the trend. Includes: - Monthly content and authority work - Ongoing technical fixes - Monthly or bi-weekly reporting with sample sizes - Quarterly roadmap review - Expansion into new markets and languages ## Methodology Source: https://tryshubi.com/how-we-measure/ Metrics: - Visibility: The share of tracked AI answers that mention your brand. Formula: answers mentioning you ÷ all tracked answers. - Share of voice: Your visibility compared with every tracked brand in your category. Formula: your visibility ÷ sum of all tracked brands’ visibility. - Citation share: How often AI links to your own domain as a source. Formula: citations to your domain ÷ all citations. - Average position: Where your brand appears when an answer lists several options. Formula: mean rank across answers that mention you. - Sentiment and accuracy: Whether AI describes you positively and correctly, checked against your canonical facts. Formula: positive, accurate mentions ÷ all mentions. - AI referral traffic: Visits and leads that arrive from AI assistants, tracked separately in analytics. Formula: sessions and conversions from AI referrers. Sampling: - Each question is run multiple times per engine, because AI answers are not deterministic. - Questions are run from the target market, in the target language. - The question set stays fixed between reports, so month-on-month comparisons are fair. - Every report states the number of questions, engines, runs and the date range. What Shubi will not do: - Promise guaranteed mentions, rankings or recommendations. - Report a single cherry-picked answer as proof. - Hide the sample size behind a proprietary score. - Count prompt volume or content volume as a business result. ## Frequently asked questions ### What is AI visibility? AI visibility is how often, and how well, AI assistants mention and recommend your brand when people ask questions about your category. It covers whether you appear in the answer, how you are described, which sources the AI cites, and whether you are recommended over competitors. ### How is this different from SEO? SEO aims to rank a page in a list of links. AI visibility aims to get your brand named and cited inside the answer itself. Good SEO helps, because AI systems read the web, but AI answers also depend on how consistently your brand is described across third-party sources, reviews, directories and publications. ### Which AI platforms do you track? We track ChatGPT, Google Gemini, Perplexity, Claude, Google AI Overviews and Microsoft Copilot. Coverage per plan is listed on the pricing page. ### Can you guarantee that AI will recommend us? No. Nobody controls what AI models say, and any agency promising guaranteed mentions is not being honest with you. We guarantee the work, a clear method, and transparent measurement of what changed, including sample sizes. ### How quickly will we see results? Technical and on-site fixes can be picked up within weeks. Changes that depend on third-party sources, such as reviews, publications and directories, usually take two to four months to show consistently in AI answers. That is why our retainers start with a three-month first term. ### Do you work outside Malaysia? Yes. We are based in Malaysia and work with brands across Asia, including Singapore, Indonesia, the Philippines, Thailand, Vietnam and India. We track prompts in the local market and language your buyers actually use. ## Pages - [Home](https://tryshubi.com/): What Shubi does and how an engagement works - [Services](https://tryshubi.com/services/): The five service areas and what each one delivers - [How we measure](https://tryshubi.com/how-we-measure/): Metric definitions (visibility, share of voice, citation share, sentiment, AI referral traffic) and sampling method - [Free AI snapshot](https://tryshubi.com/snapshot/): How to request a free AI visibility snapshot - [About](https://tryshubi.com/about/): Why Shubi exists and what it believes - [Contact](https://tryshubi.com/contact/): How to reach the team - [Privacy policy](https://tryshubi.com/privacy-policy/): How personal data is handled - [Insights](https://tryshubi.com/insights/): Articles on AI visibility, GEO and AEO for brands in Asia ## Insights ## Authorised Dealer ChatGPT Car Recommendations in Singapore > Why Singapore authorised car dealers rank on Google yet go unnamed in ChatGPT car recommendations, and a practical plan to audit prompts and fix the facts AI quotes. URL: https://tryshubi.com/insights/authorised-dealer-chatgpt-car-recommendations-singapore/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Singapore Industry: automotive Topics: AI visibility, AEO, automotive, Singapore, ChatGPT ### Short answer Singapore authorised dealers drop out of ChatGPT car recommendations when their warranty, service and COE facts sit in brochures, sales decks or images, while forums, marketplaces and parallel importers publish plain text AI can quote. The fix is crawlable fact pages, consistent entity names, structured data, credible third-party coverage and a fixed monthly prompt check. ### Key points - ChatGPT search can only cite your site if OpenAI's OAI-SearchBot can crawl it, so check robots.txt and CDN bot rules first. - State authorised status, manufacturer warranty scope, recall handling and service centre locations in plain HTML text, not in PDFs or banner images. - Explain COE, Category A or B and package pricing in neutral terms that match LTA and CCS language, because AI repeats what it can verify. - Service, warranty and after-sales pages often earn citations that model launch pages do not. - Track mention rate, accuracy and cited sources from a fixed prompt set each month, and treat single answers as samples. Your showrooms on Leng Kee Road and in Ubi are busy on weekends, your service centre is booked out, and your brand pages rank for the model names on Google. Yet when a buyer asks ChatGPT where to buy a new car in Singapore with proper warranty and after-sales support, the answer names a marketplace, a forum thread and a parallel importer. That gap in **authorised dealer ChatGPT car recommendations** is now something marketing and brand heads at Singapore dealer groups are being asked to explain. This article sets out why it happens, the prompts Singapore car buyers actually type, how to audit your position this week, and the fixes that make an authorised dealer easier for AI tools to find, describe correctly and cite. It is written for the people who run marketing at franchise and multi-brand groups, not for individual salespeople. ### Why authorised dealers go missing from ChatGPT car recommendations ChatGPT leaves an authorised dealer out when it cannot find clear, crawlable text that links the dealer's name to the brand, the models, the warranty and the service network the buyer asked about. It does not see your ad spend or showroom footfall. It sees pages it can read and quote. #### Training data versus live search ChatGPT forms answers in two ways. The first is what the model learned in training, which is fixed at a point in time and is vague about recent price lists, new models and changes in who holds which franchise. The second is live search, where ChatGPT reads current pages and shows the sources it used. OpenAI's crawler documentation explains the second route. It describes OAI-SearchBot as the crawler used to surface websites in ChatGPT's search features, and says sites that opt out of it will not appear in ChatGPT search answers. GPTBot, which relates to model training, is controlled separately. Dealer sites often sit behind a CDN or bot-protection layer that blocks unfamiliar crawlers, and marketing rarely knows. #### Why marketplaces, forums and parallel importers get quoted A buyer's question needs a short list of facts: model, price including COE, warranty length, who services the car, and whether it is an official unit. A marketplace listing or a long forum thread often puts all of that in one block of text. A parallel importer's page may state its own warranty offer plainly in the first paragraph. Many authorised dealer sites do the opposite. The price is in a downloadable price list, the warranty is in an image, and the service network is behind a "book a test drive" form. Google's guidance on its own AI features recommends making important content available in textual form, and says there are no extra technical requirements beyond normal indexing. ChatGPT is a different product, but the principle carries over: AI quotes what it can parse. #### Why franchise structures confuse the model Singapore dealer groups often run several brands under different trading names, with a listed or private parent, separate service subsidiaries and pre-owned arms. A buyer may know the showroom name, the press may use the group name, and the brand's regional site may list the distributor under its legal entity. If those labels never appear together in plain text, the model has little reason to connect "the official dealer for this brand" with your company. ### What buyers in Singapore actually ask AI Singapore buyers ask AI about cost of ownership, COE timing, warranty, and service convenience, and expats and new citizens ask how the system works at all. These are realistic prompt patterns to test, grouped by stage. **Exploring** - "Is it worth buying a new car in Singapore now with COE prices where they are?" - "Category A or Category B for a family SUV in Singapore, which makes more sense?" - "Just relocated to Singapore. How does buying a car work and should I go to an authorised dealer?" **Comparing** - "Authorised dealer vs parallel importer in Singapore for a Japanese hybrid, what is the difference in warranty?" - "Who is the official distributor for [brand] in Singapore and where are their service centres?" - "Best car dealer in Singapore for a guaranteed COE package" **Deciding and owning** - "Authorised [brand] service centre near Tampines that does weekend appointments" - "新加坡 [brand] 官方代理 保修多久" (asked in Mandarin: how long is the official agent's warranty) Run each prompt in English, and in Mandarin where your buyers use it. Answers differ by language, so treat each as a separate test. ### How to check where you stand You can run a useful audit in a week with a spreadsheet and two or three people. The aim is a baseline you can repeat, not a single screenshot for a board pack. Our [measurement approach](https://tryshubi.com/how-we-measure/) explains why repeated sampling matters. #### Step 1: Confirm AI crawlers can reach you Open your robots.txt and check rules for OAI-SearchBot, GPTBot, Googlebot and Bingbot. Then ask your web team or CDN provider whether bot management blocks or challenges these user agents. A site that returns a captcha to OAI-SearchBot will not be cited in ChatGPT search, however good the content. #### Step 2: Build a fixed prompt set Write 20 to 30 prompts covering your brands, key models, COE and package questions, warranty and service. Include the relocation and expat prompts, because those buyers lean heavily on AI and have no prior dealer loyalty. Keep the wording fixed so next month's results are comparable. #### Step 3: Run each prompt several times Run each prompt three to five times in ChatGPT with search on, and the same set in Gemini, Perplexity and Copilot if budget allows. Log whether you are named, how you are described, which competitors appear, and which URLs are cited. Answers vary between runs and users, so you are measuring a rate, not a rank. #### Step 4: Check the facts AI repeats For every answer that names you, check the details: brand list, showroom addresses, service centre locations, opening hours, warranty length. Flag anything out of date or wrong. Wrong facts are often more damaging than absence, because a buyer acts on them. #### Step 5: Map the cited sources List every domain cited in answers about your brands. You will usually see the same handful: marketplaces, motoring news sites, forums, the manufacturer's regional site, and a few rivals. That list is your target map for the fixes below. ### How to fix it The fixes are mostly editorial and technical work your own team can do. Start with the pages that answer the questions buyers ask most, then widen into third-party sources. #### Publish a plain "Why buy from an authorised dealer" page Write one HTML page per brand that states, in plain sentences: you are the authorised distributor or dealer for that brand in Singapore, the manufacturer warranty length and what it covers, how recalls and software updates are handled, the parts policy, and where the cars are serviced. Keep it factual. Do not attack parallel importers; describe your own offer so AI can quote it accurately. #### Explain COE and pricing in neutral, regulator-aligned language Buyers ask AI about COE constantly, so give them a page that explains it correctly. LTA's OneMotoring site describes the basics: COEs are released through two open bidding exercises a month, a successful bid produces a Temporary COE that must be used to register a vehicle, and a registered vehicle's COE lasts 10 years. It also defines Category A and B by engine capacity and power, with different power limits for fully electric cars. If you sell "guaranteed COE" or package prices, explain what is included and what is not. The Competition and Consumer Commission of Singapore's Guidelines on Price Transparency, in force since 1 November 2020, cover drip pricing, price comparisons, discounts and "free" claims. Pages written to that standard are clearer for buyers and easier for AI to quote without hedging. #### Give after-sales its own content lane Service is where authorised dealers have real, checkable advantages, and where AI answers are often thin. Publish one page per service centre with address, hours, services offered, brands covered, booking options and whether it handles EVs or hybrids. These pages answer "authorised service centre near me" prompts that new-car pages never will. #### Add structured data and consistent entity names Use Organization markup on the corporate site and LocalBusiness markup on each showroom and service centre page. Google's Organization structured data documentation explains properties such as legal name, alternate names, address and `sameAs` links, which help tie trading names to the parent entity. Then use one spelling of each trading name everywhere: site, Google Business Profiles, ACRA-registered name where relevant, LinkedIn and press releases. #### Strengthen the third-party sources AI already reads Use the cited-source map from your audit. Typical targets and what to do: | Source type | Why AI uses it | What to do | | --- | --- | --- | | Motoring news and review sites | Recent, editorial, model-specific | Brief journalists with clear launch facts and warranty terms in text | | Marketplaces and new-car listings | Price, model and dealer in one place | Keep authorised-dealer listings complete and consistent with your site | | Manufacturer regional site | High trust for "official distributor" questions | Ask the brand to list your entity name and showrooms exactly as you do | | Forums and owner communities | Real ownership experience | Answer service questions openly under a named official account | | Google Business Profiles | Hours, addresses, reviews | Keep every showroom and service centre profile accurate | #### Write for the relocation buyer Expats and new arrivals ask AI how the whole system works. A plain guide covering COE, registration, insurance basics and why buying from an authorised dealer matters, written without sales pressure, often becomes the page AI cites for these prompts. Keep it current when rules change. ### What to measure Measure a small set of metrics monthly and report trends, not single answers. Our [AI Visibility Intelligence service](https://tryshubi.com/services/#intelligence) is built around the same idea. - **Mention rate:** share of prompt runs in which your dealer or group is named. - **Citation share:** share of cited URLs that are your own domain or sources you influence. - **Accuracy:** share of mentions with correct brands, addresses, hours and warranty facts. - **Description quality:** whether AI calls you an authorised or official dealer, and whether the tone is neutral, positive or negative. - **Competitor presence:** which rivals and parallel importers appear for the same prompts. Expect month-to-month movement even with no changes, because models, search indexes and answer formats change. Look for direction over three or more months before drawing conclusions, and add an extra run after a launch or a sharp move in COE premiums. ### Common mistakes - Treating Google rankings as proof of AI visibility, without testing ChatGPT, Gemini and Perplexity directly. - Keeping price lists, warranty terms and service details only in PDFs, images or behind forms. - Writing comparative claims about parallel importers instead of stating your own verifiable facts. - Letting showroom names, group names and legal entity names drift across the site, profiles and press. - Blocking OpenAI or other AI crawlers at the CDN without a deliberate decision. - Reporting one screenshot of a ChatGPT answer as a result, good or bad. ### How Shubi helps Shubi works with dealer groups on the research and the fixes: a repeatable prompt panel across ChatGPT, Gemini, Perplexity and Copilot, an accuracy check on what AI says about each brand and showroom, and the [content](https://tryshubi.com/services/#content) and [technical](https://tryshubi.com/services/#technical) work that makes authorised-dealer facts quotable. We cannot guarantee any AI tool will name you, but we can show where you stand and what is most likely to move it. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Why do parallel importers or marketplaces appear in ChatGPT answers instead of us? When ChatGPT searches the web, it cites pages that answer the question in clear text. Marketplace listings, forum threads and parallel importer sites often put the model, price, COE status and warranty claim in one paragraph. If your authorised-dealer facts are spread across a brochure and a contact form, you give the model little to quote. #### Should our pages say parallel imports are worse? No. Stick to checkable facts about your own offer: manufacturer warranty terms, genuine parts, recall handling and service network. Comparative or disparaging claims raise fair trading questions and are exactly the kind of statement AI tools tend to drop or hedge. #### Does ranking first on Google mean we will show up in ChatGPT? Not automatically. Both reward crawlable, well-structured pages, but ChatGPT uses its own crawler and its own source selection. You still need to confirm OAI-SearchBot access and publish the facts buyers actually ask about. #### How often should a dealer group re-test its prompts? Monthly for core brand and model prompts, plus an extra run after a launch, a price change or a move in COE premiums. Run each prompt several times and compare trends, because answers vary between runs and users. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Organization structured data (Google Search Central)](https://developers.google.com/search/docs/appearance/structured-data/organization) - [Certificate of Entitlement (COE), OneMotoring by LTA](https://onemotoring.lta.gov.sg/content/onemotoring/home/buying/upfront-vehicle-costs/certificate-of-entitlement--coe-.html) - [Guidelines on Price Transparency (Competition and Consumer Commission of Singapore)](https://www.ccs.gov.sg/consumer-protection/legislation-and-guidelines/guidelines-on-price-transparency/) --- ## Corporate Clinic ChatGPT Visibility: Losing HR to Rivals > Why Malaysian panel and occupational health clinic groups lose employer shortlists when ChatGPT and Copilot name rival chains, and a practical plan to check and fix it. URL: https://tryshubi.com/insights/malaysia-corporate-clinic-chatgpt-competitor-loss/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: medical Topics: AI visibility, AEO, medical, Malaysia, ChatGPT, Microsoft Copilot, corporate health ### Short answer Malaysian corporate clinic groups lose employer shortlists in ChatGPT when their panel, screening and occupational health facts sit in PDFs, vary between branches or are blocked from AI crawlers, while rivals publish clear employer pages. The fix is crawlable employer-facing pages per city cluster, consistent clinic names and addresses, MAB-compliant wording and a fixed monthly prompt check. ### Key points - HR and benefits teams now ask ChatGPT and Microsoft Copilot which clinic groups can handle panel care and annual screening, before a broker list is opened. - ChatGPT search can only cite your site if OpenAI's search crawler is allowed in, so check robots.txt and your CDN or firewall first. - Employer pages should state locations, services, OHD availability, report turnaround and languages in plain text, not only inside a brochure PDF. - One clinic name, address and phone format across your site, LinkedIn, panel directories and Maps listings reduces wrong or missing answers. - Health service advertising in Malaysia needs Medicine Advertisement Board approval, so keep AI-facing pages factual and check them with your compliance team. - Track mention rate, accuracy and cited sources from a fixed prompt set each month, and treat single AI answers as samples. An HR manager at a manufacturer in Shah Alam needs a clinic group for 2,000 staff: panel GP visits across the Klang Valley and Penang, annual health screening, and pre-employment medicals for new hires. She opens ChatGPT, asks for options, and gets three names. Yours is not one of them, even though you run more branches than two of the three. That is the **corporate clinic ChatGPT visibility** problem, and it means losing employer pipeline before anyone in your partnerships team hears about the tender. This article is for marketing and partnership leads at multi-site clinic groups, occupational health operators and hospital-linked primary care networks in Malaysia. It explains why AI assistants pick rivals, what employer buyers actually type, how to audit your position this week, and what to change, all within Ministry of Health advertising rules. ### Why ChatGPT recommends a rival clinic group ChatGPT names the clinic groups whose employer-relevant facts it can find, read and match to the question. Branch count, Google ratings and ad spend do not carry over on their own; clear, crawlable pages that answer "who serves employers, where, with what" do. #### Training data and live search An AI answer draws on two things. The first is what the model learned during training, which is fixed at a point in time and blurry on details such as which branch now offers audiometry. The second is live retrieval, where the assistant searches the web, reads current pages and shows the sources it used. OpenAI's crawler documentation explains that OAI-SearchBot is used to surface websites in ChatGPT's search features, and that sites which opt out of it will not appear in ChatGPT search answers. GPTBot, used for training, is controlled separately. Healthcare groups often run strict firewall rules, so it is common for these crawlers to be blocked without the marketing team knowing. #### Copilot matters in corporate buying Many HR teams work inside Microsoft 365, so Copilot is often the first assistant they reach for. Microsoft's documentation says that when web search is on, Copilot turns the prompt into a short search query of a few words and sends it to Bing, then uses the results to ground its answer. It also notes that the user's IP address is sent so Bing can localise results. In practice, your pages need to be indexed in Bing, not only Google, and they need to answer short queries such as "corporate health screening Penang". #### Why rivals win the citation Rivals usually win for unglamorous reasons. They have a single "For employers" page that lists cities, services and turnaround in plain sentences. Their clinic names match across their site, LinkedIn and panel directories. Your equivalent facts may be split across a corporate brochure PDF, a careers page and twenty branch pages with three spellings of the group name. Google's guidance on AI features makes the same point from another angle: there is no special markup for AI Overviews, and the basics of being indexable with helpful, text-based content still apply. ### What employers in Malaysia actually ask AI Employer buyers ask practical, operational questions, often in a mix of English and Bahasa Malaysia. Build your test set from prompts like these, grouped by where the buyer is in the process. **Shortlisting** - "Which clinic groups in Malaysia offer corporate panel coverage in Klang Valley, Penang and Johor Bahru?" - "Klinik panel korporat yang ada cawangan di Shah Alam dan Senai untuk syarikat pembuatan" - "Best corporate health screening provider for 2,000 staff in Selangor" **Checking capability** - "Which providers in Malaysia can do medical surveillance for workers exposed to chemicals, with a DOSH registered occupational health doctor?" - "Pre-employment medical check-up for foreign workers in Johor, which clinics handle it at volume?" - "How fast do corporate clinics in KL return annual screening reports to HR?" **Comparing and deciding** - "Compare [your group] and [rival group] for corporate panel clinics in Penang" - "Does [your group] have clinics open on Saturday in Seremban and Melaka?" Note that the last two prompts name you. Even when you appear, the answer may describe you wrongly: old branch lists, closed clinics, services you stopped offering. Accuracy is as important as presence. ### How to check where you stand You can run a useful baseline in a week with a spreadsheet and three assistants. Our [how we measure](https://tryshubi.com/how-we-measure/) page explains the method we use; the steps below are a simplified version. #### Step 1: Fix the prompt set Write 20 to 30 prompts from the list above, adapted to your cities and services. Include 5 that name your group directly. Keep the wording fixed from now on, because changing prompts makes trends meaningless. #### Step 2: Run each prompt several times Run every prompt in ChatGPT, Microsoft Copilot and Google's AI features, three times each, ideally from a Malaysian connection and a logged-out or fresh session. Answers vary between runs and users, so a single result is a sample, not a verdict. #### Step 3: Log names, facts and sources For each answer, record whether you were named, which rivals were named, and what was said about you. Then record the cited sources: your site, a rival's site, a broker or panel directory, a news article, a forum thread. #### Step 4: Check crawler access Ask IT to confirm that robots.txt and the CDN or web application firewall allow OAI-SearchBot and Bingbot. Check Bing Webmaster Tools for indexing of your employer and branch pages, not only the home page. #### Step 5: Grade accuracy Mark each statement about your group as correct, outdated or wrong. Wrong addresses, closed branches and missing services are the most common errors, and each one is a specific fix. ### How to fix corporate clinic ChatGPT visibility The fixes are mostly about making true facts easy to find and quote. None of them needs a new campaign, and all of them help human HR buyers as much as AI assistants. #### Publish employer pages per city cluster Create one HTML page for employers per major cluster, for example Klang Valley, Penang and Johor. Each should state, in plain sentences: which branches serve that area, services offered to employers (panel care, annual screening, pre-employment medicals, medical surveillance, vaccinations), opening hours including Saturdays, languages spoken, and how reports reach HR. Write it the way a procurement team asks, not the way a brochure sells. #### State occupational health credentials precisely DOSH registers occupational health doctors under the Occupational Safety and Health Act 1994, and its Occupational Health Division monitors their activity, particularly medical surveillance programmes. If your group has DOSH-registered OHDs, say which locations they cover and which surveillance work you handle. Vague phrases such as "full occupational health support" give an assistant nothing to match to a prompt about chemical exposure checks. #### Take facts out of PDFs Corporate brochures and screening package PDFs rarely feed AI answers well. Move the core facts into page text: package names, what each includes, minimum headcount if any, and how results are delivered. Keep the PDF for procurement files, but make the HTML page the main source. #### Make names and addresses consistent Use one form of the group name and one address format for every branch, across your site, Google Business Profile, Bing Places, LinkedIn, panel administrator directories and broker listings. Close or update listings for branches that have moved. Contradictions between sources are a common reason for an assistant to skip a brand or describe it wrongly. | Source | Why AI tools may use it | What to do | | --- | --- | --- | | Your employer pages | Direct, current facts in plain text | One page per city cluster, updated quarterly | | Branch pages | Address, hours and services per location | Same template and naming for every branch | | Maps listings (Google, Bing) | Location and hours for local queries | Match names and hours to your site | | Panel and broker directories | Third-party confirmation of coverage | Audit listings, request corrections | | LinkedIn company page | Company description often quoted | Align the "About" text with your site | | Industry news and associations | Independent mention of capability | Share factual updates on new services or sites | #### Stay inside MAB rules The Ministry of Health's Medicine Advertisement Board FAQ says advertisements of health services by private hospitals, clinics, medical laboratories and radiology clinics need prior Board approval. That covers much of what a clinic group puts online. Work with compliance before publishing, and favour factual descriptions of scope, locations and credentials over claims such as "leading" or "best". Factual pages are both safer and easier for AI tools to quote. #### Use Organization and clinic markup Add Organization structured data on the corporate site and consistent LocalBusiness or MedicalClinic markup on branch pages, matching the visible text. Markup does not guarantee an AI mention, but it helps search systems connect branches to the parent group. If you want help with this layer, our [technical work](https://tryshubi.com/services/#technical) covers crawler access and markup. ### What to measure Measure a few things consistently rather than many things once. A monthly cycle with the same prompt set is enough for most clinic groups, with an extra run after opening or closing branches. - **Mention rate:** the share of prompts where your group is named, by assistant and by city cluster. - **Citation share:** how often your own pages are among the cited sources, compared with rivals and directories. - **Accuracy:** the share of statements about you that are correct. Track wrong branches and services separately. - **Rival set:** which competitors appear most often, and for which prompts. - **Sentiment and framing:** whether you are described as a corporate provider or only as a walk-in GP chain. Expect noise. Results move between runs, between users and when assistants change their models, so look at the trend across two or three months before deciding a fix has or has not worked. ### Common mistakes - Treating a Google ranking or Maps rating as proof of AI visibility. They are related but measured differently. - Blocking AI crawlers at the firewall by default, then wondering why ChatGPT cites a directory instead of you. - Publishing "best corporate clinic Malaysia" pages for search bots. They add compliance risk and give AI nothing verifiable. - Leaving closed or moved branches live on directories, so assistants keep sending HR to the wrong address. - Testing once, in one assistant, in English only, and drawing conclusions from it. - Changing the prompt set every month, which makes progress impossible to see. ### How Shubi helps Shubi runs fixed prompt sets across ChatGPT, Copilot, Gemini and Perplexity for clinic groups, logs who is named and which sources are cited, and turns the gaps into a prioritised list of fixes. Our [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) service covers the measurement, and our [content work](https://tryshubi.com/services/#content) helps turn employer facts into pages AI tools can quote. We do not promise mentions; we measure honestly and fix what is in your control. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Why does ChatGPT name a rival clinic chain when our Google ratings are better? Ratings help on Maps, but ChatGPT answers from what it can read and match to the question. If a rival has a clear page saying it runs pre-employment checks and medical surveillance for employers in Shah Alam, and your equivalent facts sit in a PDF, the rival is easier to cite. Answers also vary between runs and users, so judge the pattern, not one reply. #### Which prompts should a corporate clinic group monitor? Use the questions HR, procurement and safety officers actually type: panel clinic coverage by city, annual screening packages for large headcounts, pre-employment medicals, medical surveillance under DOSH rules and turnaround for reports. Include English and Bahasa Malaysia versions, and keep the set fixed so month-on-month results are comparable. #### Can we write 'the best corporate clinic in Malaysia' on our pages to get recommended? No. Health service advertising by private clinics needs prior approval from the Medicine Advertisement Board, and unverifiable superiority claims are a compliance risk. Plain, checkable facts such as locations, services, registered occupational health doctors and accreditation serve HR buyers and AI answers better. #### Does this replace our broker and panel administrator relationships? No. Brokers and third-party administrators still matter, and their directories are among the sources AI tools may read. The aim is to make sure that when an HR manager asks an AI assistant first, your group is described correctly and has a fair chance of being named alongside the broker route. #### How quickly do changes show up in AI answers? Live search features can pick up new or corrected pages once they are crawled, which may take days or weeks. What a model learned in training changes much more slowly. Re-test monthly with the same prompts and look for a trend over two or three cycles. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [Data, privacy, and security for web search in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)](https://learn.microsoft.com/en-us/copilot/microsoft-365/manage-public-web-access) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Occupational Health Division (DOSH Malaysia)](https://dosh.gov.my/en/bahagian-kesihatan-pekerjaan/) - [Medicine Advertisement Board FAQ (Pharmaceutical Services Programme, Ministry of Health Malaysia)](https://pharmacy.moh.gov.my/en/node/1163) --- ## Corporate jewellery gifts: why AI shortlists skip you > Why jewellery groups in Singapore and Malaysia go missing when HR, admin and procurement teams ask ChatGPT, Copilot or Google for corporate gift shortlists, and how to fix it. URL: https://tryshubi.com/insights/corporate-jewelry-ai-gift-shortlists/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Singapore Industry: jewelry Topics: AI visibility, AEO, jewelry, Singapore, Malaysia, Microsoft Copilot, ChatGPT ### Short answer AI assistants build corporate jewellery gift shortlists from pages that state procurement facts plainly: minimum order quantities, engraving lead times, price bands, invoicing and delivery. Jewellery groups in Singapore and Malaysia drop out when those facts sit in PDFs or sales emails. Publish a text corporate gifting hub, keep it current and test buyer prompts before each festive season. ### Key points - Corporate gift prompts ask about quantities, lead times, invoicing and delivery, so bridal and fashion creative alone does not get a jeweller onto the shortlist. - Google says AI Overviews and AI Mode only link to pages that are indexed and eligible for a snippet, so a gifting brochure locked in a PDF or form is easy to miss. - Microsoft Copilot turns a long procurement prompt into a short Bing query, so the gifting hub must answer short phrases like corporate jewellery gifts Singapore. - Singapore buyers check GST on gifts per recipient per occasion, and Malaysian buyers work within gift policies under section 17A of the MACC Act, so publish clear price bands. - Test a fixed set of HR, admin and procurement prompts monthly and before Chinese New Year, Hari Raya, Deepavali and year-end, because answers vary between runs. An HR manager in a Raffles Place office has six weeks until the company's long-service awards. She opens ChatGPT and types: "Shortlist jewellers in Singapore that do corporate gifts, engraved gold pendants, 60 pieces, delivery to two offices." Your jewellery group has a corporate desk, an engraving workshop and outlets across the island. The corporate jewellery shortlist she gets names two rivals and a gifting marketplace. You never knew the order existed. This article is for marketing and B2B leads at multi-outlet jewellery groups in Singapore and Malaysia that sell to companies: festive gifts, long-service awards, client appreciation, retirement gifts and branded pieces. It explains why AI assistants leave corporate jewellery programmes off gift shortlists, what procurement buyers actually ask, how to check where you stand this week, and what to publish so assistants can describe your programme correctly. ### Why AI gift shortlists skip established jewellers AI assistants shortlist the vendors whose pages answer the buyer's operational questions in plain text. Most jewellery websites are built for consumers: bridal campaigns, collection photography and a "contact us for corporate enquiries" form. That gives an assistant nothing to quote when the question is about quantities, engraving, invoices and delivery dates. The engines differ in detail, but they share one dependency: they need readable, indexable text. - **Google AI Overviews and AI Mode.** Google's Search Central documentation says a page must be indexed and eligible to be shown with a snippet to appear as a supporting link, with no extra technical requirements. It also describes "query fan-out": issuing several related searches across subtopics to build one answer. A corporate gift prompt may fan out into searches about engraving, bulk pricing and delivery. If your only gifting content is a PDF or an enquiry form, none of those searches finds you. - **ChatGPT search.** OpenAI's crawler documentation says OAI-SearchBot is used to surface websites in ChatGPT's search features, and that sites which block it will not be shown in ChatGPT search answers. Some retail groups block AI crawlers by default in their CDN or security settings without marketing knowing. - **Microsoft Copilot.** Where a company runs Microsoft 365, HR and admin staff may well ask Copilot Chat at work. Microsoft's documentation says Copilot sends Bing a generated query that "consists of a few words informed by the user's prompt". A detailed procurement request becomes something like "corporate jewellery gifts Singapore engraving". Whatever ranks in Bing for that short phrase is what Copilot reads. There is also the question of trust. When a prompt involves 60 or 200 pieces and a hard deadline, assistants tend to favour sources that state constraints clearly. A rival page that says "minimum 20 pieces, engraving 10 working days, delivery islandwide" is easier to recommend than a beautiful page that says "bespoke corporate solutions". ### What buyers in Singapore and Malaysia actually ask AI Corporate buyers ask in plain English, with local shorthand: "CNY", "Raya", "LSA" for long-service awards, "PO", "tax invoice", "CBD", "KL", "PJ". In Malaysia many will mix in Malay or Chinese. Spelling varies too: "jewelry" and "jewellery" both appear. **Early research (HR, admin, corporate communications)** - "Corporate jewelry gift ideas for long-service awards in Singapore, budget around S$300 per person." - "Which jewellers in KL do corporate gifts with company logo engraving?" - "Hadiah korporat barang kemas untuk Hari Raya, syarikat di Lembah Klang." **Shortlisting (procurement and office managers)** - "Compare three Singapore jewellers for 120 engraved gold coins for Chinese New Year: lead time, minimum order, delivery." - "Jewellers in Penang that issue tax invoices and accept purchase orders for corporate orders." - "公司年终礼品 金饰 吉隆坡 可以刻字 的珠宝店" **Final checks (finance and compliance)** - "Does this jeweller deliver to multiple offices in Singapore and Johor Bahru?" - "Can we pay by bank transfer against invoice for a corporate jewellery order?" Notice what is missing: almost nobody asks about collection stories or celebrity campaigns. They ask about logistics, paperwork and risk. ### How to check where you stand You can run a basic audit in a week with a spreadsheet and the assistants your buyers use. Our [measurement method](https://tryshubi.com/how-we-measure/) explains how we run the same idea at scale with fixed prompt sets and repeated runs. #### Step 1: Write a prompt set from real enquiries Pull the last year of corporate enquiries from your sales team. Turn them into 20 to 30 prompts covering the three stages above, in English, Malay and Chinese where your buyers use them. Include both markets if you sell in both. #### Step 2: Run the prompts across engines Run each prompt in ChatGPT with search on, Copilot Chat with web content on, Google AI Mode and Perplexity. Use a clean browser profile. Run each prompt at least three times on different days, because answers vary between runs and users. #### Step 3: Record who is named and which pages are cited For each answer, note whether you are named, in what position, and which URLs are cited. Copilot Chat shows the exact web search queries it sent to Bing in its citation section, according to Microsoft. Write those down: they tell you the short phrases your gifting hub must answer. #### Step 4: Check accuracy line by line Where you are mentioned, check every fact: minimum order, lead time, engraving options, outlets, delivery areas, payment terms. An answer that names you with a wrong minimum order can cost you the enquiry just as surely as silence. #### Step 5: Check crawler access Ask your web team to confirm that robots.txt and your CDN or firewall rules do not block OAI-SearchBot, Googlebot or Bingbot from your gifting pages. Confirm the pages are indexed in Google Search Console and Bing Webmaster Tools. ### How to fix it The fix is a corporate gifting hub written for procurement, not for brides. Our [content service](https://tryshubi.com/services/#content) does this work, but every step below can be done in-house. #### Build one corporate gifting hub in HTML Create a single page, for example `/corporate-gifts/`, with separate sections for Singapore and Malaysia. State in plain sentences: who the programme is for, minimum order quantities, typical lead times with and without engraving, price bands per piece, what can be engraved or branded, packaging options, delivery areas and how to place an order. Keep the brochure PDF as a download, but never as the only source. #### Publish the facts procurement asks for Use a short table that an assistant can lift directly. | Fact | Example of a quotable line | Why it matters | | --- | --- | --- | | Minimum order | "Corporate orders start from 20 pieces per design." | Filters you in or out of volume prompts | | Lead time | "Allow 10 working days for engraving after artwork approval." | Deadline prompts are common before festive seasons | | Price bands | "Pendants from S$150 to S$600 per piece before GST." | Buyers match bands to gift policies | | Payment | "Purchase orders and bank transfer against tax invoice accepted." | Finance checks this before approving | | Delivery | "Delivery to offices across Singapore; Klang Valley and Penang via our Malaysian outlets." | Multi-office prompts need this | Use your real figures. If a fact changes by season, say so and date it. #### Explain gold pricing honestly Gold jewellery prices move with the daily gold rate, which makes fixed corporate quotes risky. Publish your method instead: for example, how the gold rate and workmanship combine, and how many days a corporate quote holds. AI tools repeat whatever number they last saw, so an explained method ages far better than a stale price. #### Write for the buyer's tax and compliance checks You should not give tax or legal advice, but you can make your buyers' checks easy. In Singapore, IRAS guidance on gifts says a GST-registered business accounts for output tax when the total cost of gifts to the same recipient for the same occasion exceeds S$200, excluding GST, where input tax was claimed. Finance teams therefore ask for per-unit prices before GST. Show them. Separately, the Ministry of Law's AML/CFT Division requires regulated precious stones and metals dealers to file a cash transaction report for cash or cash-equivalent payments above the threshold, currently S$20,000. Stating that you invoice corporate orders and accept bank transfers removes a question before it is asked. In Malaysia, corporate liability under section 17A of the MACC Act means many companies run gifts and hospitality policies under the government's Guidelines on Adequate Procedures. Clear price tiers help a compliance officer match your pieces to their policy limits. Do not suggest that a gift helps anyone win business. #### Make outlets and corporate contacts consistent Many groups list corporate enquiries under the consumer hotline, so AI answers send procurement to a retail counter. Publish one corporate contact route per market on the hub and keep outlet names, addresses and hours identical across your site, Google Business Profile, Bing Places and mall directories. Add Organization structured data that matches the visible text, as Google recommends. Our [technical service](https://tryshubi.com/services/#technical) covers crawler access, structured data and indexing. #### Earn mentions where buyers already look Assistants often cite third-party pages for gifting prompts: corporate gifting guides, chamber of commerce member directories, HR and office-management publications, and marketplace listings. List your programme where it is genuinely relevant, with the same facts as your hub. A short case note about a real, consented corporate order, with no invented figures, gives an assistant something specific to cite. #### Publish ahead of the festive calendar Corporate buying for Chinese New Year, Hari Raya, Deepavali and year-end starts weeks ahead. Update the hub with seasonal lead times and cut-off dates at least six weeks before each season, and say on the page when it was last updated. ### What to measure Measure four things on a fixed prompt set, monthly and again before each festive season. Treat any single answer as a sample, not a verdict. - **Mention rate:** the share of prompts where you are named at all. - **Citation share:** how often your own pages, rather than marketplaces or rivals, are cited as sources. - **Accuracy:** the share of mentions where minimum order, lead time, outlets and payment terms are correct. - **Position and framing:** whether you appear as a primary option for corporate orders or as a footnote under consumer jewellery. Add two business signals: corporate enquiries that mention an AI tool, and visits to the gifting hub from AI referrers in your analytics. Expect movement over months rather than weeks, and expect noise between runs. ### Common mistakes - Treating corporate gifting as a contact form rather than a page with facts. - Publishing the programme only as a designed PDF or inside a slide deck. - Using the bridal brand voice ("timeless", "bespoke") where buyers need numbers and dates. - Letting the corporate hub show last year's festive lead times or old price bands. - Blocking AI crawlers at the CDN without anyone in marketing knowing. - Writing one hub for both markets with Singapore dollars only, which leaves Malaysian buyers guessing. ### How Shubi helps Shubi is an AI visibility company based in Malaysia, working across Asia. Through [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) we track how ChatGPT, Copilot, Gemini, Perplexity and Google describe your corporate gifting programme against your rivals, and help you publish the facts procurement teams need. We do not promise mentions or rankings; we measure them honestly and improve what AI can read. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### How is corporate jewellery AI discovery different from bridal? Bridal prompts ask about design, ring size and budget for one couple. Corporate prompts ask whether a jeweller can deliver 80 engraved pieces to three offices by a date, issue a tax invoice and quote per unit. AI can only shortlist you for the second kind of question if those operational facts are on a page it can read. #### Should a PDF brochure replace web copy for our gifting programme? Keep the PDF for sales meetings, but put the same facts in HTML. Google's documentation says supporting links in AI features must be indexed and eligible for a snippet, and important content should be in text form. Many AI crawlers read HTML far more reliably than designed PDFs. #### Do Malaysian and Singapore procurement teams ask the same prompts? The structure is similar, but the details differ: currency, tax wording, delivery areas and festive calendars. Singapore buyers may ask about GST and CBD delivery, while Klang Valley and Penang buyers ask in English, Malay or Chinese and mention Hari Raya or long-service awards. Write separate sections for each market. #### Can we publish gold prices for corporate orders when gold moves daily? Yes, with care. Publish how you price, for example a daily rate plus workmanship, and state how long a corporate quote is valid. AI tools repeat old numbers, so a dated, explained pricing method is safer than a fixed figure that goes stale. #### Will Shubi guarantee that AI recommends our corporate gifting programme? No. Nobody can guarantee mentions or citations in AI answers, and answers vary between users and runs. Shubi helps make your programme facts clear and retrievable, then measures how often and how accurately you appear over time. ### Sources - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Overview of OpenAI crawlers (OpenAI)](https://developers.openai.com/api/docs/bots) - [Data, privacy, and security for web search in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)](https://learn.microsoft.com/en-us/copilot/plan/manage-public-web-access) - [Gifts and samples: do I charge GST? (IRAS)](https://www.iras.gov.sg/taxes/goods-services-tax-(gst)/charging-gst-(output-tax)/common-scenarios---do-i-charge-gst/gifts-and-samples) - [Transaction-based requirements for regulated dealers (Ministry of Law, AML/CFT Division)](https://acd.mlaw.gov.sg/compliance/transaction-based-requirements/) - [Introduction of Guidelines on Adequate Procedures, Section 17A MACC Act 2009 (Skrine)](https://www.skrine.com/insights/alerts/february-2019/introduction-of-guidelines-on-adequate-procedures) --- ## Corporate Wellness Supplements in AI Search: Malaysia > Why pharma-owned and licensed supplement brands selling to Malaysian employers lose AI shortlists to rivals, and how to fix it without risky health claims. URL: https://tryshubi.com/insights/malaysia-corporate-wellness-supplement-ai-search/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: pharmaceuticals Topics: AI visibility, AEO, pharmaceuticals, Malaysia, corporate wellness, ChatGPT ### Short answer Corporate wellness supplement brands in Malaysia lose AI search answers when their employer programmes exist only in sales decks and their public pages are consumer shop listings. AI tools cite brands whose programme structure, registered products, ingredients and logistics are described in plain HTML, kept separate from consumer copy and within Medicine Advertisements Board rules. ### Key points - HR and benefits teams ask AI to shortlist workforce wellness vendors before they ask for proposals, so programme facts must be public and quotable. - Consumer influencer reach does not translate into employer AI answers, which need programme structure, logistics and reporting details. - Ingredient and allergen tables belong in HTML pages, not brochure scans or product images. - Health supplement advertising in Malaysia is regulated by the Medicine Advertisements Board, so pre-approve neutral programme facts with your regulatory team. - Give the employer programme its own canonical pages, separate from the D2C storefront, and track AI mentions on fixed prompts monthly. An HR director at a Penang manufacturer is building next year's wellness budget. Before calling vendors, she asks ChatGPT which supplement brands run corporate wellness programmes for Malaysian employers, and gets back three names. Your company, a licensed brand with a pharmaceutical parent and a range already sold in Guardian and Watsons, is not one of them. This is the **corporate wellness supplement AI search** problem: strong consumer presence, absent from the employer shortlist. This article is for brand, employer solutions and digital leaders at pharmaceutical and nutraceutical companies selling workforce supplement programmes in Malaysia. It covers why AI tools skip you, the prompts HR and benefits buyers actually use, a one-week audit, and fixes that stay inside Malaysia's health advertising rules. ### Why do AI tools leave corporate wellness supplement brands out? AI tools leave you out when the employer programme is invisible on the open web. The assistant can find your consumer products, but it cannot find a page that says you run workforce programmes, what they include and how they work, so it names the brands that wrote that down. #### Training data versus live search Assistants answer from two sources. One is what the model learned in training, which reflects how often and how clearly the web discussed a brand up to a cut-off date. The other is live retrieval: the assistant searches, reads current pages and cites them. For live retrieval, access comes first. OpenAI's crawler documentation says OAI-SearchBot is used to surface websites in ChatGPT's search features and that sites which opt out of it will not appear in ChatGPT search answers. Corporate pharma sites often sit behind strict security settings inherited from the global group, and those settings can block AI crawlers without the local team knowing. #### Why consumer pages do not answer employer questions An HR buyer's prompt asks about programme structure: who it suits, what is in the pack, how delivery to several sites works, what reporting HR receives, how employees are told about it. Your consumer shop page answers none of that. It lists a single product, a price and a promotion. Google's guidance on AI features recommends making sure important content is available in textual form. Employer programme facts usually live in a proposal PDF sent after a sales call. A rival that publishes even a plain programme overview gives the assistant something to cite. #### Why compliance caution leads to silence Regulatory teams are right to be careful. In Malaysia, the Ministry of Health controls health product advertising through the Medicine Advertisements Board, and its guidance says approved advertisements carry an approval number beginning with KKLIU. Many brands respond by publishing as little as possible. But factual programme information is not the same as a health claim, and silence hands the AI answer to whoever published more. ### What buyers in Malaysia actually ask AI Employer buyers ask about programmes, logistics, compliance and fit with existing wellness plans. HR teams in multinationals prompt in English. Local corporates and GLCs often mix English and Bahasa Malaysia, and Chinese-owned firms sometimes ask in Mandarin. **Exploring** - "Which supplement brands offer corporate wellness programmes for employers in Malaysia?" - "Ideas for a workplace wellness package for 800 factory staff in Penang" - "Program kesihatan pekerja dengan suplemen untuk syarikat di Malaysia" **Comparing** - "Compare corporate vitamin programmes in Malaysia for shift workers" - "Which wellness supplement suppliers can deliver to offices in KL, Johor and Kuching?" - "Are the supplements in [brand]'s employee programme registered with NPRA?" **Deciding** - "What do HR managers say about [brand]'s corporate wellness programme?" - "马来西亚有哪些品牌为公司员工提供保健品福利计划？" Run each prompt several times and in more than one assistant. Answers vary between runs and users, so you are looking for patterns. ### How to check where you stand A useful audit takes a week and needs a marketer and someone from regulatory affairs. Our [measurement approach](https://tryshubi.com/how-we-measure/) explains why fixed prompts and repeated runs give a fairer picture than one-off checks. #### Step 1: Write a fixed prompt set Draft 20 to 30 prompts across the stages above. Include generic category prompts, prompts naming your brand, and prompts that mention the workforce types you serve: factory, office, healthcare, logistics. #### Step 2: Run them in the main assistants Test ChatGPT with search on, Google Gemini, Perplexity, Microsoft Copilot and Google Search with AI Overviews. Record whether you are named, which rivals appear and which pages are cited. #### Step 3: Check what is said about you When you are named, check every fact. Look for discontinued products, old pack sizes, wrong distributors and, most importantly, health claims you never made. If an assistant attributes an unapproved claim to your brand, trace the source page. #### Step 4: Check crawl access and format Confirm with IT that OAI-SearchBot, Googlebot, Bingbot and PerplexityBot are not blocked. Then list the pages that describe your employer programme and note which are HTML, which are PDFs and which are behind a form. #### Step 5: Map the cited sources List the domains the assistants cite. Benefits platforms, HR media, marketplace listings and rivals' programme pages each point to a different fix. ### How to fix it The work is to publish the facts you already share in proposals, in a form AI tools can read, with regulatory sign-off. None of it requires new health claims. #### Give the employer programme its own pages Create a dedicated section, for example /employers/, separate from the consumer shop. Open with a short paragraph stating what the programme is, who it is for and where in Malaysia you deliver. That paragraph is what an assistant is most likely to quote. #### Publish programme facts, not benefit claims Describe what an employer gets: which registered products are included, pack formats, minimum headcount, delivery to multiple sites, how employees receive information, what reporting HR gets and how renewals work. These are operational facts. Have regulatory affairs review the page, and keep any product claims to wording that has already been approved. #### Put ingredients, allergens and registration details in HTML Buyers and their occupational health advisers ask about ingredients and allergens. Publish them as HTML tables per product, with the product's registration reference written the same way every time. MOH's guidance notes that a product's registration number and its KKLIU advertising approval number are different identifiers, so do not mix them up on the page. #### Connect to how Malaysian employers already think about wellness Many HR teams know the Ministry of Health's KOSPEN WOW workplace programme, which MOH describes as a workplace health intervention aimed at reducing non-communicable disease risk among workers. If your programme is designed to sit alongside an employer's wider wellness activities, say so factually. Do not imply MOH endorsement you do not have. #### Earn third-party sources AI tools trust | Source type | Why AI tools use it | What to do | | --- | --- | --- | | Employee benefits and flexi-benefit platforms | Structured vendor listings employers browse | Make sure your listing describes the programme, not just products | | HR media and events | Named, dated coverage aimed at the buyer | Share programme facts and approved data, not slogans | | Pharmacy chain and marketplace listings | Frequently crawled product detail | Keep names, pack sizes and registration details consistent | | Industry associations | Seen as neutral and credible | Keep member listings current and linked | #### Answer the questions procurement asks Corporate buyers do not stop at HR. Procurement and finance teams join once a pilot is on the table, and they ask assistants different things: how billing works, whether the supplier can invoice a Malaysian entity, how stock is replaced when a product changes, what happens when an employee reports a reaction. Publish a short employer FAQ page that answers these in plain sentences. Keep the answers factual and dated. If your adverse event process routes reports to your pharmacovigilance team, say that in one line and link to the relevant contact route. A buyer reading an AI summary that mentions a clear safety process sees a serious supplier, and the assistant has a reason to describe you that way. #### Keep distributors and partners consistent Many brands reach employers through distributors, benefits platforms or pharmacy chains with corporate desks. Each of them publishes its own description of your products. Send partners one approved programme summary, product names in their registered form and current pack images, and check their pages each quarter. When ten partner pages agree with your own site, AI tools have a consistent picture to repeat. #### Retire old campaign pages Expired microsites and old promotions that mention discontinued SKUs or past claims keep getting crawled. Redirect them to current pages or remove them, so assistants stop repeating outdated information. ### What to measure Measure mention rate, citation share, accuracy and claim safety on your fixed prompt set every month, and judge trends over at least three months. - **Mention rate:** the share of employer prompts where your brand is named. - **Citation share:** how often your own programme pages are cited, against rivals and platforms. - **Accuracy:** the share of mentions with correct products, packs and programme details. - **Claim safety:** any answer attributing unapproved health claims to your brand, logged and traced. Ask new employer enquiries which tools they used during their search. Expect variance between runs; one answer is a sample, not a verdict. ### Common mistakes - Treating influencer reach and consumer reviews as proof of employer visibility. - Keeping the employer programme only in proposal PDFs sent after a sales call. - Writing new benefit claims for SEO or AI pages without MAB-aware review. - Scanning ingredient sheets as images instead of publishing them as text. - Letting old campaign microsites with discontinued SKUs stay live. - Implying government or MOH endorsement because a programme fits a workplace health theme. ### How Shubi helps Shubi tracks how ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews answer employer wellness prompts in Malaysia, and shows which rivals and sources take the citations. We then help brand and regulatory teams turn approved programme facts into pages AI can quote, through [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) and [content work](https://tryshubi.com/services/#content). [Message Shubi on WhatsApp](https://wa.me/601137634679). ### Frequently asked questions #### Why does ChatGPT name a rival supplement brand when HR asks for workplace wellness options? When ChatGPT searches the web, it cites pages that describe what the buyer asked about in clear text. If a rival publishes how its employer programme works and you only publish consumer product pages, the rival gives the assistant more to quote. Answers also vary between runs and users, so test several times before drawing conclusions. #### Can we publish health benefit claims to get mentioned more often in AI answers? No. In Malaysia, health product advertising is controlled by the Medicine Advertisements Board under the Ministry of Health, and approved advertisements carry a KKLIU approval number. Publish factual programme information and claims your regulatory team has cleared, not new benefit claims written for AI. #### Should our employer programme sit on the same site as our online shop? It can sit on the same domain, but it should have its own section and URLs. Employer pages need different facts: pack structure, minimum orders, delivery to multiple sites, employee communication and reporting. Mixing them into shop listings makes it harder for AI tools to tell what you offer to companies. #### Do AI Overviews ignore regulated health brands? Google says there are no special requirements for appearing in AI features beyond normal indexing and helpful content. Regulated brands often lose out because their facts are locked in PDFs or behind forms, not because they are regulated. Clear, accurate HTML pages help. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [How to recognise an approved advertisement (Pharmaceutical Services Programme, MOH Malaysia)](https://pharmacy.moh.gov.my/en/content/how-recognize-approved-advertisement.html) - [KOSPEN WOW overview (Ministry of Health Malaysia)](https://qhub.moh.gov.my/web/overview-kospen) --- ## East Malaysia Gold: Missing From Copilot Answers > Why Sabah and Sarawak gold retail groups with busy showrooms go unnamed in Microsoft Copilot, and a practical plan to check Bing visibility and fix it. URL: https://tryshubi.com/insights/east-malaysia-gold-copilot-visibility/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: gold Topics: AI visibility, AEO, gold, Malaysia, Microsoft Copilot, Sabah, Sarawak ### Short answer East Malaysia gold retailers go missing from Microsoft Copilot because Copilot grounds web answers in Bing, and many Sabah and Sarawak groups have thin Bing coverage, facts locked in social posts, and Peninsula-centred listicles naming other brands. The fix is Bing-indexed branch and product pages, consistent facts in English and Bahasa Malaysia, and monthly Kuching and Kota Kinabalu prompt tests. ### Key points - Microsoft says Copilot turns a prompt into a short Bing search query and uses the results to ground its answer, so Bing indexing comes first. - Gold buying guides online are often written from Kuala Lumpur, which leaves Sabah and Sarawak brands out of the sources assistants read. - Branch, product and buyback facts posted only on Facebook, Instagram or TikTok are hard for Copilot to find and cite. - Questions about buying in Kuching and collecting in Kota Kinabalu, or delivery across states, need explicit written answers on your site. - Publish the same facts in English and Bahasa Malaysia, and make sure Peninsula and Borneo branch details never contradict each other. - Gold investment wording must stay factual: Bank Negara has warned the public about unsustainable gold investment schemes. Your showrooms in Kuching, Miri and Kota Kinabalu are full on weekends and before Hari Raya, Gawai and Chinese New Year. Customers know your name, your 916 designs and your buyback counter. But when an office worker opens Microsoft Copilot and asks where to buy gold in Kuching, the answer lists Peninsula chains and a couple of generic tips. That gap in **East Malaysia gold** visibility is quiet, but it shapes who gets shortlisted before anyone walks into a mall. This article is for marketing and digital leads at Sabah and Sarawak gold retail groups with several branches and central accountability for the brand. It explains why Copilot leaves you out, what buyers in Borneo actually ask, how to audit your position in a week, and the specific fixes that give your group a fair chance of being named and described correctly. ### Why East Malaysia gold brands miss Copilot answers Copilot names the gold retailers it can find through Bing and match to the buyer's question. A busy shop floor, strong Facebook following or long history in Kuching does not count unless it shows up in pages Bing has indexed. #### Copilot grounds answers in Bing Microsoft's documentation on Copilot web search explains the mechanics. When web search is on, Copilot reads the prompt, generates a short search query of a few words, and sends it to the Bing search service. It then uses the results to compose its answer. Microsoft also notes that the user's IP address is sent so Bing can localise results. Two things follow. First, if your branch and product pages are not in Bing's index, they cannot be used. Second, the query Copilot sends is short, something like "gold shop Kuching" or "kedai emas Kota Kinabalu", so your pages need to answer exactly that kind of query in plain text. #### Peninsula content dominates the sources Much of the online writing about buying gold in Malaysia is produced from Kuala Lumpur: listicles, comparison posts, finance blogs and news. When those pages are the best match for a broad query, the brands they name become the brands Copilot repeats. A Sabah or Sarawak group that is never mentioned in that layer is easy to leave out, even for a prompt that names Kuching. #### Social-first proof is hard to cite Many East Malaysia gold groups put their energy where customers are: Facebook, Instagram, TikTok and WhatsApp broadcasts. That works for existing customers, but daily price posts, new designs and promotions in social video are hard for a search index to read and quote. If your website is a thin catalogue with a phone number, the assistant has little to work with. ### What buyers in Sabah and Sarawak actually ask AI Buyers in East Malaysia ask in English, Bahasa Malaysia and, in many Kuching and Sibu households, Mandarin. Use prompts like these, adapted to your branches, grouped by buying stage. **Early research** - "Where to buy 916 gold in Kuching with good buyback?" - "Kedai emas yang dipercayai di Kota Kinabalu untuk beli barang kemas kahwin" - "古晋哪里买金条比较好?" **Comparing options** - "Gold shops in Miri open on Sunday" - "Difference between 916 and 999 gold, where can I buy both in Sandakan?" - "Can I buy gold online from a Sarawak shop and collect in Sibu?" **Brand checks** - "Is [your brand] a reliable gold shop in Kuching?" - "Does [your brand] have a branch in Tawau, and what are its hours?" The brand-check prompts matter as much as the discovery ones. Assistants sometimes list branches that have moved, quote old hours, or mix up two brands with similar names. ### How to check where you stand A useful baseline takes about a week and a spreadsheet. Our [how we measure](https://tryshubi.com/how-we-measure/) page sets out the full approach; this is a version you can run in-house. #### Step 1: Confirm Bing can see you Verify your domain in Bing Webmaster Tools if you have not already, submit your sitemap, and check that branch, product and buyback pages are indexed, not just the home page. Ask whoever runs your CDN or firewall to confirm Bingbot is not blocked. #### Step 2: Build a fixed prompt set Write 20 to 30 prompts across your cities and languages, including at least five that name your brand. Keep the wording unchanged from month to month so you can compare results. #### Step 3: Run prompts in Copilot and others Run each prompt three times in Microsoft Copilot, then in ChatGPT and Gemini for comparison. Use a Malaysian connection, ideally from Sabah or Sarawak, because Copilot localises using the user's location. Answers vary between runs and users, so record each run. #### Step 4: Log names and sources For each answer, note whether you were named, which competitors were named, and which sources were cited: your site, a rival, a mall directory, a news article, a listicle. #### Step 5: Grade accuracy Mark every statement about your brand as correct, outdated or wrong. Branch addresses, opening hours and the products each branch carries are the usual errors. ### How to fix East Malaysia gold visibility in Copilot The work is mostly about turning facts you already have into pages Bing can index and Copilot can quote. It also makes your site more useful to buyers who never touch an AI tool. #### Publish a page for every branch and city Give each branch its own HTML page with the full address, mall or shoplot name, opening hours including Sundays and public holidays, phone or WhatsApp number, and what is available there (jewellery, bars and coins, buyback, trade-in). Add a city page for Kuching, Kota Kinabalu, Miri and any other town with more than one branch, summarising what buyers can do in that city. #### Answer the cross-state questions in writing Buyers in Borneo regularly ask about ordering online, collecting at another branch, or delivery between Sabah, Sarawak and the Peninsula. Whatever your policy is, write it down plainly: what you offer, which branches, typical timing, and any conditions. If you do not offer it, say so. An assistant that finds no answer will guess or point to a competitor that has one. #### Keep English and Bahasa Malaysia in step If you publish in both languages, make sure the facts match: the same branches, hours and products. Where you serve many Chinese-speaking customers, consider key pages in Mandarin too. Mismatched versions create contradictions that make assistants less confident about your brand. #### Make the website the source of truth, not social Keep posting on social, but mirror the facts that matter on the website: new collections, purity and weight details, buyback process and any change in hours. If you change prices, stock or branch details, notify Bing. Microsoft's IndexNow guidance says sites should notify search engines when they add pages, remove content, change prices or stock, or substantially update existing pages. #### Earn mentions outside your own site Copilot is more likely to name brands that other sources also name. Look for factual coverage in Sabah and Sarawak news outlets, mall directories, wedding and festive shopping guides, and chamber of commerce listings. Correct any directory that lists old branches. | Source | Why AI tools may use it | What to do | | --- | --- | --- | | Your branch and city pages | Direct facts in plain text | One page per branch, reviewed monthly | | Bing and Google Maps listings | Location and hours for local queries | Match names and hours exactly to the site | | Mall directories | Independent confirmation of branch locations | Check every mall you trade in | | Local news and guides | Third-party mention outside KL-focused content | Share factual updates on openings and events | | Social profiles | Brand name and links | Link back to the matching site pages | #### Add structured data that matches the page Use Organization markup on the corporate site and consistent store markup on branch pages, with the same name, address and phone shown in the visible text. Google's Organization documentation describes how this markup helps search systems understand your business details; it supports, but does not guarantee, how assistants describe you. Our [technical work](https://tryshubi.com/services/#technical) covers indexing, crawler access and markup. #### Keep investment wording careful Gold bars and coins invite investment questions. Describe purity, weight, pricing basis and the buyback process, and avoid promising returns. Bank Negara Malaysia has publicly described gold investment schemes that paid promised returns from investors' own money rather than from gold trading. Calm, factual wording protects buyers and is easier for assistants to repeat accurately. ### What to measure Track a small set of measures every month with the same prompts, and add a run after opening or closing a branch. - **Mention rate:** how often you are named, split by Copilot, ChatGPT and Gemini, and by city. - **Citation share:** how often your own pages appear among cited sources, compared with rivals and listicles. - **Accuracy:** the share of correct statements about branches, hours and products. - **Bing coverage:** indexed branch and product pages in Bing Webmaster Tools. - **Framing:** whether you are described as a local trusted name, a generic shop, or not at all. Expect movement between runs and as Microsoft updates Copilot. Judge progress across two or three months, not one result. ### Common mistakes - Checking only Google and assuming Bing sees the same thing. - Running the business mainly through social posts while the website stays a thin catalogue. - Letting Peninsula and Borneo branch lists drift apart on the site, directories and Maps listings. - Writing English pages only, then testing prompts in Bahasa Malaysia or Mandarin and wondering why you vanish. - Using investment language such as guaranteed growth on bar and coin pages. - Testing once, from KL, and drawing conclusions about Kuching buyers. ### How Shubi helps Shubi runs fixed prompt sets in Copilot, ChatGPT, Gemini and Perplexity for gold retailers, logs which brands are named and which sources are cited, and turns the gaps into a prioritised list of fixes. Our [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) service covers the measurement, and our [content work](https://tryshubi.com/services/#content) helps turn branch and product facts into pages assistants can quote. We cannot promise mentions, but we can show you clearly where you stand and what to change. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Why focus on Copilot rather than ChatGPT for East Malaysia gold? Both matter, but Copilot is built into Windows, Edge and Microsoft 365, which many office workers in Kuching, Miri and Kota Kinabalu use every day. Microsoft's documentation says Copilot grounds web answers using Bing, so Bing coverage is a direct lever. Fixing facts and pages helps ChatGPT and Gemini too. #### Do our Peninsula pages help our Borneo branches appear? Only if the facts line up. If the corporate site lists one set of East Malaysia branches and a regional page or directory lists another, assistants may skip you or repeat the wrong version. One source of truth for every branch helps both markets. #### Should we publish our inter-state delivery and collection policies? Yes, if you offer them. Buyers ask AI whether they can order online and collect in another state, and an assistant will guess or name a rival if your site does not say. State what you offer, where, and any conditions, in plain text. #### Can we describe gold as a safe investment on these pages? Keep it factual. Describe products, purity, weights and how buyback works, and avoid promised returns. Bank Negara Malaysia has publicly warned about gold investment schemes that paid promised returns from investors' own money, so cautious wording protects buyers and your brand. #### Does Shubi guarantee that Copilot will name us? No. Nobody can guarantee what an AI assistant says, and answers vary between runs and users. Shubi measures prompt outcomes over time and improves the facts and sources that assistants can find. ### Sources - [Data, privacy, and security for web search in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)](https://learn.microsoft.com/en-us/copilot/microsoft-365/manage-public-web-access) - [IndexNow: When and How Websites Should Notify Search Engines (Microsoft Bing Blogs)](https://blogs.bing.com/webmaster/September-2024/IndexNow-When-and-How-Websites-Should-Notify-Search-Engines) - [Organization structured data (Google Search Central)](https://developers.google.com/search/docs/appearance/structured-data/organization) - [Update on joint raids conducted by enforcement agencies on companies suspected of conducting illegal investment schemes using gold (Bank Negara Malaysia)](https://www.bnm.gov.my/-/update-on-joint-raids-conducted-by-enforcement-agencies-on-companies-suspected-of-conducting-illegal-investment-schemes-using-gold) --- ## Johor Industrial Park ChatGPT Visibility for Developers > Why Johor industrial parks with strong occupancy still lose ChatGPT site-selection answers to rivals, and how developers and REIT marketers can check and fix it. URL: https://tryshubi.com/insights/johor-industrial-park-chatgpt-visibility/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: property Topics: AI visibility, AEO, property, Malaysia, ChatGPT ### Short answer Johor industrial parks go missing from ChatGPT site-selection answers when power, land, access and zone facts sit in gated PDF decks, while trade press and agency reports publish them in plain text. Developers fix it with public HTML spec pages per park, consistent park and developer names, structured data, trade coverage and a monthly technical prompt check. ### Key points - Operations and real estate leads now paste technical site requirements into ChatGPT before any consultant is hired, so the AI longlist matters. - Each park needs a public HTML spec page with power, water, plot sizes, zoning, access and distance to ports and the Causeway or Second Link. - Name the developer on every park URL and use one spelling of the park name in English, Bahasa Malaysia and Chinese. - Explain how the park sits within the Johor-Singapore Special Economic Zone in plain, checkable terms, without overstating incentives. - Track mention rate, accuracy and cited sources from a fixed set of technical prompts monthly, prioritising parks with the most vacant land. A regional operations director in Shenzhen, Osaka or Singapore is asked to find 10 acres for a new electronics line within reach of Singapore. Before calling a consultant, she types the requirement into ChatGPT: power, plot size, access to a port, a sensible commute for Singapore-based managers. The answer names three parks. Yours, with committed multinationals and real power capacity, is not one of them. That is the problem of **Johor industrial park ChatGPT** visibility, and it now shapes longlists before your leasing team hears about the project. This article is for industrial park developers, logistics REIT marketers and investment promotion teams in southern Malaysia. It explains why AI tools leave strong parks out, the prompts site-selection teams actually use, how to audit your position this week, and the fixes that make a park easier to find, describe correctly and cite. ### Why Johor industrial parks drop out of ChatGPT answers ChatGPT leaves a park out when it cannot find plain, crawlable text linking the park name and developer to the technical facts the buyer asked about. Occupancy rates and tenant logos in a pitch deck do not count if the model cannot read them. #### Training data versus live search ChatGPT answers from two places. One is what the model learned in training, which is fixed at a point in time and often out of date on new phases, infrastructure and policy. The other is live search, where it reads current pages and shows the sources it used. OpenAI's crawler documentation describes OAI-SearchBot as the crawler used to surface websites in ChatGPT's search features, and says sites that opt out of it will not be shown in ChatGPT search answers. GPTBot, used for model training, is controlled separately. Corporate sites of listed developers and REITs are often behind security tools that block unfamiliar crawlers by default. #### Why trade press and agency reports win A site-selection prompt needs specific facts: available land, power supply, water, zoning, ceiling heights for built-to-suit units, distance to Port of Tanjung Pelepas or Johor Port, and travel time to Singapore. Property news sites, agency research notes and investment promotion pages often put several of those facts in one readable paragraph. Many park websites put them in a gated brochure, a map image or a video. Google's guidance on its AI features recommends making important content available in textual form and notes there are no extra technical requirements beyond normal indexing. ChatGPT is a different product, but the lesson holds: facts hidden in PDFs and images are hard to cite. #### Why naming and geography confuse the model A park may have a marketing name, a Malay name, a Chinese name, an older phase name and a developer that is a subsidiary of a listed group. Press may call it by the town, such as Senai, Kulai, Pasir Gudang or Tanjung Langsat. If those labels never appear together in plain text, the model has little reason to connect "the park near Senai airport with high-voltage supply" with your project. ### What site-selection teams actually ask AI Buyers of industrial space ask AI technical, comparative questions, often in English but also in Mandarin and Bahasa Malaysia. These are realistic prompt patterns to test, grouped by stage. **Scoping** - "Best industrial parks in Johor for semiconductor or electronics manufacturing near Singapore" - "Johor vs Penang vs Kedah for a new EMS plant, compare land cost, power and talent" - "Which Johor locations are inside the Johor-Singapore Special Economic Zone?" **Shortlisting** - "Industrial land in Johor with 10 acres freehold, high power supply, within 30 minutes of the Second Link" - "Built-to-suit warehouse Johor with 12m clear height near Tanjung Pelepas port" - "柔佛 工业园 靠近新加坡 适合设厂" (Mandarin: Johor industrial park near Singapore suitable for a factory) **Validating** - "Who is the developer of [park name] and which companies are already there?" - "Taman perindustrian di Johor dengan status zon bebas atau LMW" (Bahasa Malaysia: Johor industrial parks with free zone or LMW status) Treat each language as a separate test. The answers, and the sources cited, often differ. ### How to check where you stand You can build a baseline in a week with a spreadsheet and two people. The goal is a repeatable measurement, not a screenshot. Our [measurement approach](https://tryshubi.com/how-we-measure/) explains why sampling over time matters more than any single answer. #### Step 1: Check crawler access Look at robots.txt for OAI-SearchBot, GPTBot, Googlebot and Bingbot, then ask IT or your CDN provider whether bot protection blocks or challenges them. Do this for every domain you use: group corporate site, park microsites and REIT investor pages. #### Step 2: Write a technical prompt panel Draft 20 to 30 prompts based on real enquiries your leasing team has received. Include power, land size, tenure, logistics, labour, zone status and Singapore access. Keep the wording fixed so you can compare months. #### Step 3: Run and log Run each prompt three to five times in ChatGPT with search on, then in Gemini, Perplexity and Copilot if you can. Record whether each park is named, how it is described, which rival parks appear, and which URLs are cited. #### Step 4: Check accuracy For every mention, check the facts AI states: developer name, location, land available, tenure, utilities, and any claim about incentives or zone status. Out-of-date phase information and wrong developer attribution are common and can cost you a shortlist. #### Step 5: Map cited sources by park List the domains cited for each park and for your rivals. Separate parks by vacancy: a fully let park missing from AI answers matters less than one with 40 acres still to sell. ### How to fix it Most fixes are content and technical work your own team can do. Start with the parks that have the most vacant land or upcoming phases. #### Publish one public spec page per park Give each park a stable HTML page that states, in plain sentences and a simple table: developer and parent company, location and nearest town, total and available land, plot sizes, tenure, zoning and permitted uses, power supply and substation details, water, gas and telecoms, road access, and distances to ports, airports and the Causeway and Second Link. Keep the PDF as a download, but make the HTML page the source of truth. #### Explain the JS-SEZ context accurately Buyers ask AI about the Johor-Singapore Special Economic Zone, so say plainly how your park relates to it. Singapore's Economic Development Board reports that the agreement was signed in January 2025, sets a target of 100 projects within the first decade and 20,000 skilled jobs, covers 11 sectors including manufacturing and logistics, and names nine flagship zones such as Senai-Skudai, Sedenak, Pasir Gudang, Tanjung Pelepas and Tanjung Bin, and Pengerang. State which flagship zone you are in, if any, and link to the official source. Do not promise tax rates. Describe incentives as available subject to eligibility and approval, and let the investment agencies speak for the detail. #### Keep cross-border access facts current Cross-border commuting is central to many Johor site decisions, and AI answers often get the timing of new infrastructure wrong. For example, The Star reported in October 2026 that the Johor Bahru-Singapore RTS Link is targeted to start passenger operations in February 2027, subject to safety certification and regulatory approvals. If your park pitch mentions the RTS Link, give the source and update the page when the date changes. #### Tie every park to the developer entity Use Organization structured data on the group site and link each park page back to it. Google's Organization structured data documentation covers legal name, alternate names, address and `sameAs` links, which help connect a park's marketing name to the developer. Use one spelling of each park name everywhere, and list the Malay and Chinese names on the page so the variants are explicitly linked. #### Earn coverage in the sources AI already reads Use your cited-source map. Typical targets: | Source type | Why AI uses it | What to do | | --- | --- | --- | | Business and property press | Recent, named, factual | Announce lettings with the park name and developer in the headline and first paragraph | | Agency research notes | Comparative, data-led | Supply accurate park facts to the agencies that cover Johor industrial | | Investment promotion agencies | High trust for "where to invest" prompts | Make sure your park is listed with correct details where listings exist | | Trade associations and chambers | Sector credibility | Speak at or publish with associations your tenants belong to | | LinkedIn and company pages | Entity signals | Keep developer and park pages consistent with the website | #### Publish an honest Johor-Singapore FAQ Site-selection teams ask practical questions: commute times, checkpoint clearance for staff and goods, which activities suit Johor rather than Singapore. A plain FAQ that answers them without overselling is the kind of page AI can lift directly. Date it and review it when policy or infrastructure changes. ### What to measure Track a short set of metrics monthly, by park, and report trends over at least a quarter. This is the core of our [AI Visibility Intelligence work](https://tryshubi.com/services/#intelligence). - **Mention rate:** share of prompt runs in which each park is named. - **Citation share:** share of cited URLs that are your own pages or coverage you influenced. - **Accuracy:** share of mentions with correct developer, location, land and utility facts. - **Competitive set:** which rival parks appear for the same prompts, and how often. - **Coverage against inventory:** mention rate for parks with the most available land, since those carry the most revenue. Expect variation even when nothing changes, because models, indexes and answer formats shift. Look for direction across several months, and run extra checks after big letting news or policy announcements. ### Common mistakes - Testing consumer property prompts and assuming the results apply to industrial site selection. - Keeping power, land and access specs only in gated PDFs or map images. - Naming parks inconsistently across the group site, microsites, press and Chinese-language material. - Overstating JS-SEZ incentives or infrastructure dates that later change, which leaves AI repeating stale or wrong claims. - Blocking AI crawlers at the firewall without a deliberate decision. - Reporting a single ChatGPT answer to leadership as proof of success or failure. ### How Shubi helps Shubi helps industrial developers and REIT teams build a technical prompt panel, measure how each park appears across ChatGPT, Gemini, Perplexity and Copilot, and fix the [content](https://tryshubi.com/services/#content) and [third-party sources](https://tryshubi.com/services/#authority) AI relies on. We cannot guarantee that any AI tool will name a park, but we can show leadership where each asset stands and what is most likely to change it. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Are PDF spec sheets and investor decks enough for AI visibility? They help human readers but are weak for AI citation. Key facts in a gated or image-heavy PDF are hard for search crawlers and AI tools to extract and quote. Publish the same facts as plain HTML text on a stable park URL and keep the PDF as a download. #### Should we test consumer property prompts too? Not for this purpose. Condo and landed-home prompts tell you nothing about industrial site selection. Build a panel of technical prompts written the way plant managers, supply chain heads and relocation consultants actually ask. #### Can we say our park offers JS-SEZ tax incentives? Only state what is official and current, and point to the government source. Incentives depend on the activity and approval by the relevant agencies, so phrase it as eligibility subject to approval rather than a promise. AI tools tend to drop or hedge claims they cannot verify. #### How often should we re-check ChatGPT answers about our parks? Monthly, using the same prompts, with extra runs after a major letting announcement, new infrastructure news or a policy update. Run each prompt several times, because answers vary between runs and users. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Organization structured data (Google Search Central)](https://developers.google.com/search/docs/appearance/structured-data/organization) - [Landmark Johor-Singapore SEZ pact signed: 100 projects, 20,000 jobs, investments across 11 sectors (Singapore EDB)](https://www.edb.gov.sg/en/business-insights/insights/landmark-johor-singapore-sez-pact-signed-100-projects-20000-jobs-investments-across-11-sectors.html) - [JB-Singapore RTS Link to commence operations in February 2027 (The Star)](https://www.thestar.com.my/news/nation/2026/10/02/jb-singapore-rts-link-to-commence-operations-in-february-2027) --- ## Malaysia Cloud Platform AI Rivals: Why Hyperscalers Win > Why Malaysian cloud and data centre providers lose sovereign hosting questions in ChatGPT, Gemini and Perplexity to global hyperscalers, and how to fix it. URL: https://tryshubi.com/insights/malaysia-cloud-platform-ai-rivals/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: saas Topics: AI visibility, AEO, saas, Malaysia, ChatGPT, cloud ### Short answer Malaysian cloud providers lose AI answers to global hyperscalers because hyperscalers publish vast amounts of crawlable documentation, while local providers keep residency, certification and compliance facts in slides and RFP packs. Publishing plain HTML pages on data centre locations, certifications, PDPA and BNM RMiT support, and workload fit gives AI tools facts they can cite. ### Key points - AI assistants answer sovereign cloud questions from crawlable text, and hyperscalers publish far more of it than most Malaysian providers. - Regulated buyers ask about PDPA cross-border rules and BNM RMiT cloud requirements, so publish how your service supports each one, reviewed by counsel. - Put facility locations, tier ratings, certifications and service scope in HTML pages, not in PDFs, data sheets or sales decks. - Use one spelling of your company, platform and facility names across your site, Bursa filings, partner directories and press. - Track mention rate, accuracy and cited sources on a fixed prompt set each month, and treat single answers as samples. A CIO at a Malaysian insurer opens ChatGPT and asks which cloud providers can host core workloads in Malaysia with data kept onshore. The answer lists three global hyperscalers and a consultancy blog. Your company runs Tier III facilities in the Klang Valley, has banks on its customer list and a team that knows the RMiT paperwork by heart, yet it is not named. That is the **Malaysia cloud platform AI rivals** problem: local providers with the right answer losing the AI shortlist to global brands with more pages. This article is for marketing, product marketing and alliance leaders at Malaysian cloud, colocation and managed hosting companies. It explains why AI tools favour hyperscalers on sovereignty questions, what Malaysian buyers actually type, how to audit your position this week, and which fixes give you a fair chance of being cited. ### Why do AI tools pick hyperscalers over Malaysian cloud providers? AI assistants pick the providers whose facts they can find and quote. On sovereign hosting questions, hyperscalers usually win because they publish thousands of crawlable pages on regions, compliance and services, while local providers keep the same facts in slides, data sheets and RFP responses. #### Training data and live search Assistants answer in two ways. The first is what the model learned in training, which is fixed at a point in time and weighted towards brands the web writes about most. Global cloud vendors dominate that corpus. The second is live retrieval: the assistant searches the web, reads current pages and cites the ones that answer the question. Live retrieval is where a local provider can compete, but only if it can be crawled. OpenAI's crawler documentation says OAI-SearchBot is used to surface websites in ChatGPT's search features, and that sites which opt out of it will not appear in ChatGPT search answers. Data centre operators often run strict bot rules at the firewall or CDN. Security teams block unknown agents by default, and nobody in marketing finds out. #### Why documentation volume matters Google's guidance on AI features in Search recommends making sure important content is available in textual form and says there are no extra technical requirements beyond normal indexing. Hyperscalers already do this at scale: every region, service and compliance programme has its own page. A Malaysian provider's equivalent facts often sit in a 40-page PDF behind a form. AI tools cannot cite what they cannot read. #### Why the sovereignty story gets lost The policy climate favours local capability. When the Prime Minister announced the National Cloud Computing Policy in October 2024, MIDA reported that one of its four focus areas was strengthening user trust and data security. Buyers now ask AI about sovereignty, residency and local support. But if your site says "trusted sovereign cloud" without stating where the data sits, who operates it and which controls apply, the assistant has a slogan and no facts. It will fill the gap with a vendor that published the facts. ### What buyers in Malaysia actually ask AI Malaysian cloud buyers ask about residency, regulation, workload fit and local support. IT leaders mostly prompt in English, with Bahasa Malaysia common in public sector and GLC procurement. These are realistic patterns to test, grouped by stage. **Exploring** - "Which cloud providers have data centres in Malaysia for regulated workloads?" - "Sovereign cloud options in Malaysia for a bank, local versus hyperscaler" - "Penyedia awan tempatan di Malaysia yang sesuai untuk agensi kerajaan" **Comparing** - "Compare Malaysian managed cloud providers for RMiT compliance support" - "Which colocation providers in Cyberjaya offer Tier III and ISO 27001?" - "Does keeping data in Malaysia help with PDPA cross-border transfer rules?" **Deciding** - "Is [your company] a good choice for hosting core insurance systems?" - "What do customers say about [your company] support response times in KL?" Run each prompt several times and in more than one assistant. Answers vary between runs, users and locations, so look for patterns rather than single results. ### How to check where you stand You can audit your AI position in a week with a spreadsheet and two people. The aim is a baseline you can repeat, not a perfect score. Our [measurement approach](https://tryshubi.com/how-we-measure/) explains why fixed prompts and repeated runs matter. #### Step 1: Build a fixed prompt set Write 20 to 30 prompts across the three stages above, covering your main workloads: core banking, insurance, government, healthcare records, AI training. Include both "best provider" questions and questions about your own name. #### Step 2: Run them across the main assistants Test ChatGPT with search on, Google Gemini, Perplexity, Microsoft Copilot and Google Search with AI Overviews. Record whether you are named, where in the answer, which rivals appear and which URLs are cited. #### Step 3: Score accuracy For every answer that names you, check the facts. Common errors for local providers are wrong facility locations, outdated tier ratings, a merged or renamed entity, or services you stopped offering years ago. Wrong facts are often more damaging than absence. #### Step 4: Check crawl access Ask your network team to confirm that OAI-SearchBot, Googlebot, Bingbot and PerplexityBot are not blocked in robots.txt, at the WAF or by the CDN. Then check which of your key pages are HTML and which are PDFs. #### Step 5: Map the cited sources List every domain the assistants cite on your prompts. If they cite hyperscaler compliance pages, analyst lists, partner directories and news you are absent from, that is your work list. ### How to fix it The fixes are about turning facts you already have into pages AI can find and quote. None needs inflated claims, and in a regulated buying process, inflated claims backfire. #### Publish a residency and facilities page in plain HTML One page should state each facility's city, the operator, tier certification and the certifying body, power and connectivity basics, and which services run there. Name places the way buyers do: Cyberjaya, Klang Valley, Johor, Kulai. Keep it current when you open or retire a site. #### Explain PDPA and RMiT support without overclaiming Regulated buyers ask AI how a provider helps with Malaysian rules, so answer on your own site. Mayer Brown notes that the PDPA amendments removed the old whitelist approach to cross-border transfers in favour of a risk-based framework, and made data processors directly subject to the security principle. Skrine reports that BNM's 2023 RMiT policy document requires financial institutions to run a comprehensive risk assessment before cloud adoption, covering infrastructure location and legal risks. Write a page for each that explains what your service provides: onshore storage, audit rights, contract clauses, incident reporting, support for customer risk assessments. Do not call yourself "RMiT certified"; the policy applies to the institution. Have counsel review the wording. #### Turn workload fit into reference pages Hyperscalers win "best cloud for X" prompts because they publish reference architectures. Publish your own for the workloads you actually serve, each with a one-paragraph summary at the top that an assistant can lift: what it is for, where data sits, which controls apply, which partners are involved. #### Make the entity consistent Many Malaysian providers have a listed parent, an operating subsidiary, a platform brand and facility brands. Use the same names everywhere: your site, Bursa announcements, LinkedIn, partner directories and press releases. Add Organization structured data with your legal name, logo and official profiles, and state the relationships in plain text on your About page. #### Earn the third-party sources AI already trusts Assistants weigh what others say about you. This table shows where to focus. | Source type | Why AI tools use it | What to do | | --- | --- | --- | | Hyperscaler and software partner directories | Structured, regularly crawled listings | Make sure your listing names Malaysian facilities and certifications | | Industry associations and government programmes | Seen as neutral and official | Keep member and programme listings accurate and linked | | Certification registries | Verifiable proof of tier and ISO claims | Use the registered entity name exactly | | Trade and business press | Recent, named, dated coverage | Brief journalists with facts, not adjectives | | Customer case pages | Concrete workload detail | Publish only with written client approval | #### Cover Bahasa Malaysia where buyers use it If you sell to agencies or GLCs, publish Bahasa Malaysia versions of your residency, security and service pages. Properly translated pages, not machine output, give you a chance at prompts written in Malay. ### What to measure Measure four things on your fixed prompt set every month: mention rate, citation share, accuracy and sentiment. Track trends over three months or more before drawing conclusions. - **Mention rate:** the share of prompts where you are named. - **Citation share:** how often your own pages are cited, against hyperscaler and third-party pages. - **Accuracy:** the share of mentions with correct facility, certification and service facts. - **Sentiment and framing:** whether you appear as a credible option or an afterthought. Also watch AI referral traffic in analytics and ask new enquiries which tools they used. Expect variance between runs. A single bad answer is a sample, not a verdict. ### Common mistakes - Blocking AI crawlers at the firewall while asking marketing to "get us into ChatGPT". - Writing "sovereign cloud" in every headline without stating where data sits and who operates it. - Keeping certifications and tier ratings only in PDFs and RFP packs. - Claiming certifications that belong to the customer's regulator, such as "RMiT certified". - Using different names for the same entity across the site, filings and partner listings. - Publishing a comparison page that attacks hyperscalers with claims you cannot prove. ### How Shubi helps Shubi tracks how ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews describe Malaysian cloud providers on a fixed set of buyer prompts, and shows where hyperscalers and third-party pages take the citations. From there we help teams publish the residency and compliance pages AI can quote. See [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) and [technical work](https://tryshubi.com/services/#technical), or [Message Shubi on WhatsApp](https://wa.me/601137634679). ### Frequently asked questions #### Why does ChatGPT recommend global hyperscalers when we have local data centres? When ChatGPT searches the web, it cites pages that answer the question in clear text. Hyperscalers publish detailed region, compliance and service pages, while many local providers keep the same facts in gated decks. If your residency and compliance facts are not on crawlable pages, the assistant has little to quote. #### Can we claim to be 'PDPA compliant' or 'RMiT certified' on our site? Be careful. The PDPA places duties on data controllers and processors, and BNM's RMiT policy applies to financial institutions, not to the cloud provider as a certificate. Describe the controls and contract terms you offer that help customers meet their obligations, and have counsel review the wording. #### Does ranking on Google for 'cloud hosting Malaysia' mean we appear in AI answers? Not automatically. Good SEO helps because both depend on crawlable, well-linked pages, but AI tools assemble answers from the passages that best match a specific question. A page that ranks for a broad keyword may still lack the residency, certification or workload facts a CIO's prompt asks for. #### How often should we check how AI describes our cloud platform? Monthly is enough for most providers, with an extra check after a new facility, certification or major partnership. Run the same prompts several times each round, because answers vary between runs and users. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [PM unveils National Cloud Policy, focus on four key areas (MIDA)](https://www.mida.gov.my/mida-news/pm-unveils-national-cloud-policy-focus-on-four-key-areas/) - [Bank Negara issues new Policy Document on Risk Management in Technology (Skrine)](https://www.skrine.com/insights/alerts/june-2023/bank-negara-issues-new-policy-document-on-risk-man) - [Key amendments to Malaysia's PDPA and the launch of cross-border transfer guidelines (Mayer Brown)](https://www.mayerbrown.com/en/insights/publications/2025/07/from-legislative-reform-to-practical-guidance-key-amendments-to-malaysias-pdpa-and-the-launch-of-cross-border-transfer-guidelines) --- ## Malaysia Energy Logistics AI Accuracy: Fix Wrong Facts > Why ChatGPT, Gemini and Perplexity get ports, fleets and joint venture names wrong for Malaysian energy logistics firms, and a practical plan to correct them. URL: https://tryshubi.com/insights/malaysia-energy-logistics-ai-accuracy/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: oil and gas Topics: AI visibility, AEO, oil and gas, Malaysia, ChatGPT, Perplexity ### Short answer Malaysia energy logistics AI accuracy breaks down when ChatGPT, Gemini or Perplexity piece together your ports, fleet, licences and joint venture names from old directories and press. The fix is one dated, crawlable facts page, consistent naming across registers and profiles, Organization markup, prompt checks run every month and corrections pushed to the third-party sources AI already reads. ### Key points - AI assistants describe energy logistics firms from whatever text they can find, so stale directory entries and old press releases often become the answer. - A single dated facts page listing ports, bases, vessel types, licences and joint venture names gives AI tools a current source to quote. - Use one spelling of every legal entity, joint venture and asset name across your site, PETRONAS registration, LinkedIn and press. - Check that OpenAI's search crawler and Googlebot can reach the facts page, because blocked pages cannot be cited. - Score accuracy monthly on a fixed prompt set and treat each answer as a sample, since outputs vary between runs and users. A project engineer at an operator in KL asks ChatGPT for a quick background on your company before a pre-qualification call. The answer names a supply base you closed, lists a vessel you sold, and calls your joint venture by the partner's old name. Nobody in your team saw it, and nobody can correct it in the room. That is the real cost of poor **Malaysia energy logistics AI accuracy**: wrong facts reach buyers before your sales team does. This article explains why AI assistants get bunkering, terminal, marine and supply base operators wrong, what buyers in Malaysia actually ask, how to audit your own answers this week, and the specific fixes that give ChatGPT, Gemini, Perplexity and Google's AI features a current, quotable source. The steps work whether or not you hire anyone. ### Why AI gets energy logistics facts wrong AI assistants describe your company from the text they can find, and in energy logistics most of that text is old. Directories, tender notices, conference brochures and news reports outlive the assets they describe, and AI tools have no way to know which version is current unless you publish one. #### Training data versus live search There are two routes into an answer. The first is what a model learned in training, which is frozen at a point in time and blurry on companies that are not widely covered. The second is live retrieval, where the assistant searches the web and reads current pages. OpenAI's crawler documentation describes the second route for ChatGPT. It says OAI-SearchBot "is for search" and that "sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers". GPTBot, which collects training data, is controlled separately, and OpenAI notes that "each setting is independent of the others". Many corporate sites in this sector sit behind strict firewalls or CDN bot rules, so the search crawler may never reach the page that would have corrected the answer. #### Why the sector is especially prone to errors Energy logistics firms change shape often. Vessels are chartered in and out, supply base contracts move between Kemaman, Labuan, Miri and Bintulu, terminals change hands and joint ventures are renamed when a partner exits. Each change leaves a trail of older pages that still describe the previous arrangement. Ports and fuels are also shifting. MIDA's June 2025 article on bunkering notes Malaysia's first LNG bunkering operation at the Port of Tanjung Pelepas in February 2024 and biofuel bunkering at Port Klang in July 2024. When services move this fast, an AI answer built from 2022 sources will be out of date even if every source was accurate when written. #### Why rivals sound more accurate Google's guidance on its AI features says there are no special optimisations needed beyond normal indexing, and recommends "making sure that important content is available in textual form". It also says AI Overviews and AI Mode may use "query fan-out", running several related searches to build one answer. A competitor whose site states its ports, vessel classes and licences in plain text gives every one of those searches something to match. A rival with a glossy PDF capability statement gives them almost nothing. ### What buyers in Malaysia actually ask AI Buyers ask AI about energy logistics vendors by asset, location, licence and track record, mostly in English, with Bahasa Malaysia common among local operators and contractors. These are realistic prompts to test, grouped by stage. **Scoping the market** - "Which companies provide offshore supply vessels out of Kemaman Supply Base?" - "Who are the main bunkering suppliers at Port Klang and Tanjung Pelepas?" - "Syarikat logistik minyak dan gas yang berlesen PETRONAS di Sarawak" **Checking a specific vendor** - "What vessels does [your company] operate and where are they based?" - "Is [your company] a joint venture? Who are the partners?" - "Does [your company] supply LNG or biofuel bunkers in Malaysia?" **Comparing before a tender** - "Compare [your company] and [competitor] for tank storage in Johor" - "Which Malaysian marine logistics firms have handled decommissioning projects?" Run each with and without your company name. The unbranded prompts show whether you are named at all; the branded ones show whether what AI says about you is true. ### How to check where you stand You can audit AI accuracy in two or three working days with a spreadsheet and a few accounts. The aim is a baseline you can repeat, not a one-off screenshot. We describe the method we use in more detail on [how we measure](https://tryshubi.com/how-we-measure/). #### Step 1: Write down the truth first Before you test anything, list the facts an AI answer should contain: legal entity names, joint venture names and partners, operating ports and bases, vessel types or terminal capacity categories, fuels handled, key certifications and your PETRONAS licence or registration status. Have operations and legal sign it off. Without this list, nobody can score an answer. #### Step 2: Build a fixed prompt set Pick 20 to 30 prompts from the patterns above, covering your main service lines and locations. Include a few in Bahasa Malaysia if your buyers use it. Save the exact wording, because small changes in phrasing produce different answers. #### Step 3: Run each prompt several times Test in ChatGPT with search on, Gemini, Perplexity, Microsoft Copilot and a Google search that triggers an AI Overview. Run each prompt at least three times on different days. Answers vary between runs and users, so one result is a sample, not a verdict. #### Step 4: Score accuracy field by field For every answer that mentions you, mark each fact as correct, wrong, outdated or missing. Use five fields: entity name, locations, assets, services and partners. A simple score out of five per answer is enough to see trends. #### Step 5: Log the cited sources Where the tool shows sources, record every URL. You will usually find the same few culprits: an old directory profile, a conference speaker bio, a dated news item, a partner's page that still lists the previous JV name. Those URLs become your correction list. #### Step 6: Check crawl access Ask IT to confirm that robots.txt and any firewall or CDN rules allow OAI-SearchBot and Googlebot to reach your main pages. If your services are only described in a gated portal or a downloadable PDF, note that too. ### How to fix it The fixes are mostly about publishing one current version of the truth and making the rest of the web agree with it. Start with your own site, then work outwards. #### Publish a dated facts page Create one HTML page, linked from the main navigation, that states your current facts in short sentences: who you are, which legal entities trade under the brand, where you operate, what you operate, what you handle and which licences and certifications you hold. Put a visible "last updated" date at the top and change it whenever a fact changes. This page is what you want ChatGPT and Perplexity to cite when someone asks "what does [company] do?". #### Write one page per service line and location Bunkering at Port Klang, tank storage in Johor, OSV charter from Kemaman and marine services in Labuan are different questions, so they deserve different pages. Each should name the port or base, the service, the vessel or asset types and the customers you serve, without claims you cannot back. These pages match the specific prompts buyers type far better than a single "Services" page. #### Make your licence status clear and accurate PETRONAS distinguishes between a licence, which allows a company to supply goods and services to both upstream and downstream, and a registration, which covers downstream only. Its licensing page also explains that SWEC is "the list of work categories in upstream and downstream". State which you hold and for which work categories, in the same terms PETRONAS uses, and keep it current at renewal. Do not imply a licence or category you do not hold; AI tools will repeat the claim and buyers will check. #### Fix joint venture and entity naming Write one plain sentence for each joint venture: its legal name, both partners and what it does. Use the same sentence on your site, in press releases, on LinkedIn and in conference bios, and ask partners to match it. When a JV is restructured or renamed, publish a short dated note saying so, with both the old and new names, so AI tools can connect them. #### Add Organization structured data Google's Organization structured data documentation explains how to describe a company's legal name, logo, address, contact points and profiles. Add it to your home page and facts page, list your subsidiaries where appropriate and make sure every value matches the visible text. Google's AI features guidance specifically recommends structured data that "matches the visible text on the page". #### Correct the third-party sources AI already reads Use the URLs from your audit to drive this work. | Source type | Why AI tools use it | What to do | | --- | --- | --- | | Industry and business directories | Structured, easy to parse, often old | Claim, update or ask for removal of stale profiles | | Port and terminal partner pages | Seen as independent confirmation | Ask partners to update your name, services and contact | | Trade press and news | Authoritative and frequently cited | Issue dated releases for asset, port or JV changes when disclosable | | Conference and association pages | Long-lived speaker and exhibitor bios | Send current boilerplate after every event | | LinkedIn and Google Business Profile | Frequently crawled, often the first profile found | Align descriptions word for word with the facts page | #### Keep asset changes flowing When you take delivery of a vessel, start a new bunkering service or exit a base, publish it. A short, factual release and an update to the facts page on the same day gives search-based AI a fresh source. Silence leaves the old version standing. ### What to measure Measure accuracy first, then visibility. An AI answer that names you with the wrong ports is worse than no mention at all, because it travels with your brand attached. We set this up as part of [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence). Track these monthly on the same prompt set: - **Accuracy score:** correct fields out of five per answer, averaged per engine. - **Mention rate:** share of unbranded prompts where your company is named. - **Citation share:** share of cited sources that are your own pages, and which third-party pages appear most. - **Stale source count:** how many outdated URLs still appear in citations. - **Description quality:** whether the answer describes your current services, not just your history. Expect noise. Results vary between runs, engines and users, so look at trends across three or more months rather than reacting to a single bad answer. Re-run the audit after any major asset or JV change. ### Common mistakes - Treating the capability statement PDF as the source of truth. AI tools struggle to quote it, and it is often years old. - Correcting your own site but leaving old directory profiles and partner pages untouched. - Using three versions of a joint venture name across the site, press and LinkedIn. - Blocking all AI crawlers by default, then wondering why ChatGPT's search answers cite a directory instead of you. - Claiming licences, work categories or capacities you cannot document, which AI tools then repeat to buyers who will check. - Measuring once, finding a good answer and assuming the problem is solved. ### How Shubi helps Shubi is an AI visibility company based in Malaysia. For energy logistics firms we build the prompt set, score accuracy across the main AI engines, trace wrong answers back to their sources and help your team publish facts AI can quote, through [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) and [content and entity work](https://tryshubi.com/services/#content). We do not promise mentions or rankings, only a clear method and honest reporting. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Why does ChatGPT say we operate at a port we left years ago? The model may be drawing on older training data or on a directory, tender notice or news report that still lists that port. When ChatGPT searches the web it reads current pages, so a dated facts page and corrected directory entries give it a better source. Old mentions fade slowly, so expect improvement over months, not days. #### Do procurement teams really use AI to research energy logistics vendors? We cannot point to a reliable Malaysian survey on this, so treat it as a risk rather than a statistic. Engineers and buyers do use ChatGPT, Gemini and Perplexity for quick background checks, and a wrong answer about your assets or licences can shape a first impression before any formal tender. #### Should we publish fleet and asset details if some are commercially sensitive? Publish what is already public or disclosable, such as vessel types, operating bases, certifications and service lines, and leave out what is not. A short, accurate, dated page beats a detailed one that legal will not keep current. Where you cannot share a figure, state the category instead. #### How do we handle joint venture names that AI keeps mixing up? Write one sentence per joint venture that names both partners, the legal entity and what the venture does, and repeat that wording on your site, in press releases and on partner pages. Ask partners to use the same form. AI tools confuse entities when the labels never appear together in plain text. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Organization structured data (Google Search Central)](https://developers.google.com/search/docs/appearance/structured-data/organization) - [Licensing and Procurement in Malaysia (PETRONAS)](https://www.petronas.com/suppliers/licensing-procurement-malaysia) - [Bunkering Advancements and Their Economic Spill-Overs: Unlocking Strategic Value for Malaysia (MIDA, 24 June 2025)](https://www.mida.gov.my/bunkering-advancements-and-their-economic-spill-overs-unlocking-strategic-value-for-malaysia/) --- ## Malaysia general insurers missing from AI shortlists > Why Malaysian general insurers that rank on Google for fire, engineering and liability cover go missing when risk managers and brokers ask AI for a shortlist, and what to publish. URL: https://tryshubi.com/insights/malaysia-general-insurer-ai-shortlist/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: insurance Topics: AI visibility, AEO, insurance, Malaysia, ChatGPT, Microsoft Copilot, Google AI Mode ### Short answer Malaysian general insurers drop off AI shortlists when their commercial line pages target keywords but not the scenario questions risk managers ask. ChatGPT, Copilot and Google AI Mode quote pages that state who a product covers, typical exclusions, claims steps and appetite in plain text. Publish line-by-line facts, keep licence details consistent and measure answers monthly. ### Key points - Risk managers ask AI scenario questions, such as cover for a factory in Shah Alam after a flood, which keyword-led product pages rarely answer directly. - PIAM reported gross written premium of RM24.2 billion for Malaysia's general insurance industry in 2025, with fire the second-largest class at 20.9 per cent. - Google says AI Overviews and AI Mode only link to pages that are indexed and eligible for a snippet, and may run several related searches for one prompt. - Microsoft Copilot turns a long prompt into a short Bing query, so commercial line pages must answer short phrases like contractors all risk insurance Malaysia. - Measure mention rate, citation share and accuracy on a fixed prompt set, because AI answers vary between runs and no insurer can be guaranteed a place. A risk manager at a manufacturer in Shah Alam is reviewing the group's property and engineering programme before renewal. Before calling the broker, she asks ChatGPT: "Which general insurers in Malaysia are strong in industrial all risks and business interruption for manufacturing plants?" Your company writes that business, has a long claims record and ranks well on Google for "fire insurance Malaysia". The AI answer names three other insurers and a broker's blog. For a Malaysia general insurer, that is a seat on the shortlist lost before the formal process starts. This article is for marketing, digital and commercial lines leaders at general insurers and general takaful operators in Malaysia. It explains why AI assistants leave established commercial insurers off their shortlists, what corporate buyers and intermediaries actually ask, how to audit your position, and what to publish so assistants describe your commercial lines accurately. ### Why the general insurer AI shortlist gap happens AI assistants build shortlists from pages that answer the buyer's scenario in plain words. Most insurer websites are organised by product name and keyword, with the detail locked in policy wordings, PDFs or "contact your agent" forms. That works for search rankings but gives an assistant little it can quote. The commercial market is large enough to attract plenty of competing content. PIAM, the general insurance association, reported gross written premium of RM24.2 billion in 2025, according to The Star. Motor made up 45.2 per cent of the business, fire 20.9 per cent and marine, aviation and transit 7.4 per cent. Fire and specialty lines are where risk managers do the most research, and where AI answers draw on whoever has explained the risk most clearly. How the main engines pick sources: - **Google AI Overviews and AI Mode.** Google's Search Central documentation says supporting links must come from pages that are indexed and eligible for a snippet. It describes "query fan-out", where one prompt becomes several related searches across subtopics. A question about a plant's property programme may fan out into business interruption, flood, machinery breakdown and claims. A single product page per class covers only one of those. - **ChatGPT search.** OpenAI's documentation says OAI-SearchBot is used to surface websites in ChatGPT's search features, and that sites which block it will not be shown in ChatGPT search answers. Security teams at regulated firms sometimes block unfamiliar bots by default. - **Microsoft Copilot.** Where a company runs Microsoft 365, staff may ask Copilot at work. Microsoft's documentation says Copilot sends the Bing search service a generated query that "consists of a few words informed by the user's prompt", and that a user's referenced document can shape that query. A risk manager who asks Copilot to compare insurers named in last year's renewal pack will get answers built from what Bing returns for each name. The result is predictable. Brokers' explainers, comparison articles and rivals with detailed coverage guides become the retrieved proof, even when your underwriting capability is stronger. ### What buyers in Malaysia actually ask AI Commercial buyers ask in English, often with Malay terms and local places: "Klang Valley", "Pasir Gudang", "kilang", "banjir", "CAR", "EAR", "PL". General takaful is part of the same conversation for many buyers. **Understanding the risk (risk managers, CFOs, company secretaries)** - "What does industrial all risks insurance cover in Malaysia and how is it different from a fire policy?" - "Is flood covered under a standard fire policy for a factory in Selangor?" - "Insurans kebakaran untuk kilang di Johor, apa yang perlu ada dalam polisi?" **Shortlisting insurers (procurement and risk teams)** - "Which general insurers in Malaysia write contractors' all risks for large infrastructure projects?" - "Compare insurers for directors' and officers' liability for a Bursa-listed company." - "General takaful operators offering cyber cover for Malaysian SMEs and mid-size firms." **Checking claims and service (finance and operations)** - "How long does a business interruption claim take with Malaysian insurers?" - "Is this insurer licensed by Bank Negara Malaysia?" These prompts mix education, comparison and reassurance. The insurers that appear are the ones whose pages answer all three. ### How to check where you stand You can run a first audit within a week. Our [measurement method](https://tryshubi.com/how-we-measure/) describes how we do the same thing with larger, repeated prompt sets. #### Step 1: Build a prompt set by class of business Ask your commercial underwriters and broker relationship managers for the questions they hear most. Turn them into 25 to 40 prompts covering fire and industrial all risks, engineering, liability, marine, cyber and any specialty lines you write. Include Malay prompts and takaful variants if relevant. #### Step 2: Run them across engines and days Run each prompt in ChatGPT with search, Copilot Chat with web content on, Google AI Mode and Perplexity. Repeat on at least three different days. Answers vary between runs and users, so one screenshot proves little. #### Step 3: Record mentions, sources and the queries behind them Log whether you are named, where, and which URLs are cited. Microsoft says Copilot Chat shows the exact web search queries it sent to Bing in the citation section. Those short queries are the phrases your line pages must answer. #### Step 4: Check accuracy Where you appear, check every fact against your current position: classes written, risk appetite, claims contacts, branch locations and licence status. Old product names and closed branches are common errors. #### Step 5: Check access and indexing Confirm that robots.txt and your firewall do not block OAI-SearchBot, Googlebot or Bingbot from public product and claims pages. Confirm those pages are indexed in Google Search Console and Bing Webmaster Tools. ### How to fix it The goal is to make your commercial line knowledge readable, factual and consistent. Our [content service](https://tryshubi.com/services/#content) does this with insurers, but every step below can start in-house with your underwriting and compliance teams. #### Build a hub for each class of business Give each major class its own HTML page: what it covers, who typically buys it, common extensions, common exclusions in plain words, what an underwriter needs for a quotation and how long quotation usually takes. Link the policy wording PDF, but do not make it the only source. #### Answer scenarios, not just keywords Add short scenario sections a risk manager would recognise: a plant after a flood, a contractor with a damaged crane, a data breach at a listed company. Explain which classes respond and what affects the outcome. Keep it educational and factual, and have compliance review every page. #### Publish claims facts plainly Buyers ask AI about claims service. State the claims notification route, documents needed, how loss adjusters are appointed and who the corporate claims contact is. Avoid unsupported promises about speed or outcomes. | Page type | Why AI uses it | What to publish | | --- | --- | --- | | Class of business hub | Answers "what covers X" prompts | Scope, extensions, exclusions, underwriting information needed | | Claims guide by class | Answers service and reassurance prompts | Notification steps, documents, contacts | | Broker submission page | Answers intermediary prompts | Submission requirements, appetite, contacts | | About and licence page | Answers "is this insurer licensed" | Legal name, licence type, link to BNM's list | #### Keep entity facts consistent Use your exact legal name as it appears on Bank Negara Malaysia's list of licensed general insurers, and keep it identical on your site, LinkedIn, PIAM or MTA listings and press releases. Mergers and rebrands leave old names scattered across the web, and AI tools mix them up. Add Organization structured data that matches the visible text, as Google advises. #### Earn expert citations AI answers about commercial risk often cite trade press, broker explainers and industry association material. Put your underwriters and claims leaders forward for bylined articles and conference talks on construction, flood, cyber and marine risk. Our [authority service](https://tryshubi.com/services/#authority) helps plan this. Accurate third-party coverage gives assistants a second source that confirms your own pages. #### Give intermediaries their own page Commercial risks are often placed through brokers and agents, and some of them now use AI tools to scan the market before approaching insurers. A dedicated intermediary page helps both groups. State the classes you write, the information a complete submission needs, typical turnaround for a quotation and the right contact for each region. Keep it factual and current, and date it so an assistant can tell old appetite statements from new ones. #### Publish in Malay where buyers do Many owner-managed firms and their finance staff ask in Malay. Translate core class hubs and claims guides properly rather than machine-translating policy text, and keep terms consistent with your Malay policy documents. ### What to measure Track four measures on a fixed prompt set every month, and after any product launch, rebrand or major claims event. - **Mention rate:** the share of prompts in which you are named. - **Citation share:** how often your own pages are the cited source rather than brokers or rivals. - **Accuracy:** the share of mentions with correct classes, appetite, contacts and licence details. - **Framing:** whether you are presented as a specialist in a class or only as a general name. Report results by class of business, because you may be visible for marine and invisible for engineering. Expect gradual change and month-to-month noise, and treat trends across many prompts as the signal. ### Common mistakes - Leaving commercial lines detail only in policy wordings and PDFs. - Writing product pages for keywords while ignoring the scenarios risk managers describe. - Making superlative claims that compliance would never sign off and AI rarely repeats. - Letting old company names from mergers or rebrands persist in directories. - Blocking AI crawlers by default without a decision from marketing and compliance. - Measuring only Google rankings and assuming AI visibility follows. ### How Shubi helps Shubi is an AI visibility company based in Malaysia, working across Asia. Through [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) we measure how ChatGPT, Copilot, Gemini, Perplexity and Google describe your commercial lines against rivals, then help you publish facts AI can quote accurately. We do not promise mentions or rankings, and every result we report comes with its prompt set. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Do Google rankings for commercial lines carry over to ChatGPT and Copilot? Only partly. ChatGPT search, Copilot and Google AI Mode all retrieve web pages, so good SEO helps. But they answer the buyer's full scenario, not a keyword, and they favour pages that state cover, exclusions and claims steps in plain language. Track AI answers separately from rankings. #### Can we tell AI tools we are the best fire or engineering insurer in Malaysia? Avoid it. Unsupported superlatives are a compliance risk for a licensed insurer and AI tools rarely repeat them anyway. Publish verifiable facts instead: lines written, typical risks accepted, claims process and your entry on Bank Negara Malaysia's list of licensed insurers. #### Should we publish rates or premiums for commercial lines? Usually not, because commercial premiums depend on underwriting. You can explain what drives the price, what information an underwriter needs and how long a quotation takes. That gives AI tools something accurate to say without quoting numbers that would mislead. #### Do brokers use AI tools too? Some do, for market research and drafting. That makes clear, factual line pages useful for intermediaries as well as direct corporate buyers. Keep a separate section for brokers covering submission requirements and contacts. #### Will Shubi guarantee that AI recommends our company? No. Nobody can guarantee mentions or citations in AI answers, and they vary between users and runs. Shubi improves how retrievable and accurate your facts are and measures citations with transparent prompt sets. ### Sources - [General insurance industry grows 4.8% in 2025, posts RM1.2bil underwriting profit: PIAM (The Star)](https://www.thestar.com.my/business/business-news/2026/05/06/general-insurance-industry-grows-48-in-2025-posts-rm12bil-underwriting-profit---piam) - [List of licensed financial institutions: general business (Bank Negara Malaysia)](https://www.bnm.gov.my/general-business) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Overview of OpenAI crawlers (OpenAI)](https://developers.openai.com/api/docs/bots) - [Data, privacy, and security for web search in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)](https://learn.microsoft.com/en-us/copilot/plan/manage-public-web-access) --- ## Malaysia Jewellery ChatGPT Visibility for Fine Jewellers > Why established fine jewellers in Malaysia go unnamed when buyers ask ChatGPT about bridal and gifting jewellery, and how to fix it with facts AI can quote. URL: https://tryshubi.com/insights/malaysia-jewellery-chatgpt-visibility/ Published: 2026-11-13 Updated: 2026-11-13 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: jewelry Topics: AI visibility, AEO, jewellery, Malaysia, ChatGPT ### Short answer Malaysia jewellery ChatGPT visibility depends on whether ChatGPT can find clear text that pairs your brand with boutiques, collections, services and purity facts. Jewellers who rely on Instagram and mall listings lose answers to listicles and multi-brand retailers. Publish one page per boutique and service, keep names consistent, allow OpenAI's search crawler and test prompts monthly. ### Key points - ChatGPT can only cite your site in search answers if OpenAI's OAI-SearchBot is allowed to crawl it, so check robots.txt and your CDN first. - Instagram posts and mall directory listings rarely give ChatGPT a quotable paragraph about your boutiques, services or purity standards. - Each boutique needs its own page with address, mall, hours, appointment options and the services offered there, in plain text. - Buyers ask in English, Bahasa Malaysia and Mandarin, and the festive calendar changes what they ask, so test across both. - Track mention rate, accuracy and cited sources monthly on fixed prompts, and treat single answers as samples, not verdicts. Your flagship in Pavilion or Suria KLCC is busy every festive season, your bridal collection has a loyal following on Instagram and your name is known to every jewellery buyer in the Klang Valley. Yet when a couple asks ChatGPT where to buy an engagement ring in KL, the answer lists a mall, a lifestyle listicle and two multi-brand retailers. That gap in **Malaysia jewellery ChatGPT** visibility is now a question country managers and marketing heads are being asked by their boards. This article explains why established jewellers go missing, what Malaysian buyers actually type into ChatGPT, how to audit your position this week and the specific fixes that make your brand easier to find, describe and cite. It is written for groups and maisons with several boutiques, not single-outlet shops, and it works without hiring anyone. ### Why ChatGPT leaves established jewellers out ChatGPT names jewellers it can read about in clear text that pairs the brand with the place, the occasion and the service the buyer asked about. Footfall, ad spend and follower counts do not enter into it directly. Pages it can read and quote do. #### Two routes into an answer ChatGPT forms answers in two ways. The first is what the model learned in training, which is fixed at a point in time and often vague on specific boutiques, recent openings and current collections. The second is live search, where ChatGPT reads current web pages and shows the sources it used. OpenAI's crawler documentation is direct about the second route. It says OAI-SearchBot "is for search" and that "sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers". Training data is collected by a separate crawler, GPTBot, and each robots.txt setting "is independent of the others". Luxury brand sites often sit behind aggressive bot protection set by a regional or global team, and nobody in Malaysia knows the search crawler is turned away. #### Why listicles and malls win the citation Think about what a bridal question needs: a brand, a location, a price feel, the services offered (sizing, engraving, certification, aftercare) and a reason to trust. A "best jewellers in KL" article or a mall's store directory often puts that in one paragraph. A brand site built around campaign imagery, a lookbook carousel and a store locator powered by a map widget may not say any of it in readable text. Google's guidance on its own AI features makes the same point for AI Overviews and AI Mode. It recommends "making sure that important content is available in textual form" and says there are no special optimisations required beyond being indexed. ChatGPT is a different product, but the lesson transfers: if your facts live in images and scripts, AI tools quote someone else's version. #### Why global templates struggle locally Many maisons in Malaysia run on a regional or global site template. The Malaysia pages often carry translated campaign copy but not the local facts buyers ask about: which boutique offers private appointments, whether bridal consultations are available on weekends, what purity the local gold collection uses. The local retailer or the mall fills the gap, and AI tools follow. ### What buyers in Malaysia actually ask ChatGPT Malaysian jewellery buyers ask ChatGPT by occasion, location, purity and service, in English, Bahasa Malaysia and Mandarin, and the questions shift with the festive calendar. These are realistic prompts to test, grouped by stage. **Exploring** - "Best places to buy an engagement ring in KL for under RM10,000" - "Kedai barang kemas yang dipercayai di Kuala Lumpur untuk cincin kahwin" - "吉隆坡哪里买916金首饰比较可靠" (where in KL is reliable for 916 gold jewellery) **Comparing** - "Which jewellers in Malaysia offer lab-grown diamonds with certificates?" - "Compare [your brand] and [rival] for bridal sets" - "Jewellers in Penang that do custom engraving in a week" **Ready to buy or gift** - "Corporate gifting jewellery suppliers in Malaysia for Chinese New Year" - "Does [your brand] at Pavilion resize rings and how long does it take?" Repeat the unbranded prompts before Chinese New Year, Hari Raya Aidilfitri, Deepavali, Valentine's Day and the year-end wedding season, because buyers' wording changes with each. ### How to check where you stand You can run a useful baseline in a week. The aim is a repeatable record, not a screenshot for one meeting. Our method is set out on [how we measure](https://tryshubi.com/how-we-measure/). #### Step 1: Agree the facts List what a correct answer about your brand should contain: brand and legal entity names, every boutique with its mall and address, collections, purity standards used, certification bodies for diamonds, and services such as sizing, engraving, cleaning and trade-in. Have retail operations and legal sign it off. #### Step 2: Build a fixed prompt set Choose 25 to 40 prompts covering bridal, festive gold, gifting and corporate orders, across the languages your buyers use. Save the exact wording and do not edit it between rounds. #### Step 3: Run each prompt several times Test in ChatGPT with search on, and repeat each prompt at least three times on different days. Answers vary between runs and users, so one result tells you little on its own. If you can, check Gemini, Perplexity and Google's AI Overviews with the same set for comparison. #### Step 4: Log mentions, accuracy and sources For each answer, record whether you were named, where in the list, whether the boutique details and services were right, and which sources were cited. Note every rival and retailer that appears. #### Step 5: Read what gets cited Open the cited pages and ask why they won. Usually it is a clear sentence such as "the brand offers free engraving and resizing within 30 days at its KL boutiques" that your own site never states. #### Step 6: Check crawl access Ask the team that controls your site, regional or global, to confirm that OAI-SearchBot and Googlebot can reach your boutique and collection pages, and that key facts are in the HTML rather than loaded by a script after the page opens. ### How to fix it The fixes give ChatGPT one clear, consistent and current version of your brand, then make sure the sources it already trusts agree. Start on your own site and work outwards. #### Publish one page per boutique Each boutique deserves its own page with the mall, full address, hours, phone, appointment options, parking or MRT notes, and the services and collections available there. Add a short paragraph on what makes that boutique different, such as a bridal salon, a high jewellery room or a workshop on site. This is the page you want quoted when someone asks "which jeweller at Mid Valley does custom rings?". #### Write service pages in plain text Sizing, engraving, repairs, cleaning, trade-in, certification and warranty are what buyers ask about once they have a shortlist. Give each a short page or section with turnaround times and conditions where you can commit to them. These pages answer the follow-up questions that decide which brand ChatGPT recommends. #### Be precise and compliant on purity Malaysia's Trade Descriptions (Articles Made of Precious Metals) Regulations 1994 govern how gold, silver and platinum articles may be described and require fineness marks. State purity in permitted terms, such as 916 or 750, and explain what they mean. Precise, lawful facts are easier for AI to repeat accurately, and they avoid the vague "pure gold" language that attracts enforcement and buyer distrust. #### Add LocalBusiness structured data Google's LocalBusiness structured data documentation explains how to mark up a store's name, address, opening hours and contact details. Add it to each boutique page, keep it in line with your Google Business Profiles and make sure every value matches what visitors see on the page. #### Earn mentions where AI already looks | Source type | Why ChatGPT uses it | What to do | | --- | --- | --- | | Lifestyle and bridal listicles | Answer "best jeweller" questions directly | Offer editors accurate boutique and service facts, not just images | | Mall store directories | Structured and trusted for location | Correct names, hours and unit numbers with each mall | | Bridal and wedding planning sites | Match engagement and wedding prompts | Supply consistent descriptions and current collections | | Trade and business press | Seen as independent and authoritative | Announce openings, collections and craftsmanship stories with clear facts | | Industry context sources | Frame the category | Reference recognised sources on Malaysian jewellery, such as MIDA's industry profile | MIDA's 2023 profile of the sector describes Penang as Malaysia's jewellery manufacturing centre and names several local brands. If your brand has a real manufacturing or craft story, tell it on your site with facts that a journalist or an AI tool can repeat. #### Match the languages your buyers use Publish key boutique, bridal and gold pages in Bahasa Malaysia and Chinese where your buyers use them, not just English. Keep brand, collection and boutique names identical in every version so AI tools recognise one entity, not three. #### Keep names consistent everywhere Use one spelling of the brand, each collection and each boutique across your site, Google Business Profiles, mall listings, social profiles and press releases. When a boutique moves or closes, update every one of them on the same day. ### What to measure Measure visibility and accuracy together, monthly, on the same prompts. Our [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) service tracks these across engines. - **Mention rate:** share of unbranded prompts where your brand is named. - **Position:** whether you appear first, in the middle or last when ChatGPT lists jewellers. - **Accuracy:** whether boutiques, services and purity details are correct. - **Citation share:** how often your own pages are cited versus listicles, malls and retailers. - **Sentiment:** whether descriptions are neutral, positive or misleading. Compare festive periods year on year rather than month to month, because prompts and competition change with the season. Results vary between runs and users, so watch trends over a quarter. ### Common mistakes - Relying on Instagram and campaign imagery to carry facts that only text can carry. - Leaving bot protection set by a global team untested for OpenAI's search crawler. - One generic store locator instead of a page per boutique. - Vague or non-compliant purity language that AI tools repeat back to buyers. - Different brand or collection spellings across Malaysia, regional and global sites. - Checking ChatGPT once before a campaign and drawing conclusions from a single answer. ### How Shubi helps Shubi is an AI visibility company based in Malaysia. For jewellery groups we build the prompt set across languages and festive seasons, measure how ChatGPT and other AI tools describe your brand and help your team publish boutique and service facts they can quote, through [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) and [content work](https://tryshubi.com/services/#content). We never promise mentions or rankings, only a clear method and honest reporting. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Why does ChatGPT recommend a mall or a listicle instead of our boutique? When ChatGPT searches the web, it tends to use pages that answer the question in one readable passage. A 'best jewellers in KL' article or a mall directory often does that better than a brand site built around imagery. Give ChatGPT a page that states your boutiques, services and specialities in plain text and you give it a fair chance of citing you. #### Do we need Bahasa Malaysia and Chinese pages? If your buyers use those languages, yes. Bridal and festive gold buyers in Malaysia often search in Bahasa Malaysia or Mandarin, and an AI answer in that language will draw on sources in that language. Keep brand, collection and boutique names identical across every version. #### Can we describe our pieces as 'pure gold' to match what buyers type? Only describe purity in the terms Malaysian law allows. The Trade Descriptions (Articles Made of Precious Metals) Regulations 1994 set how gold, silver and platinum articles may be described and marked. Accurate fineness terms such as 916 or 750 protect you legally and give AI tools a precise fact to repeat. #### How quickly will ChatGPT pick up changes we make? It varies. Answers built from live search can change within weeks once new pages are crawled, while what the model learned in training changes far more slowly. Measure monthly and look for a trend over a quarter. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Local business (LocalBusiness) structured data (Google Search Central)](https://developers.google.com/search/docs/appearance/structured-data/local-business) - [Trade Descriptions (Articles Made of Precious Metals) Regulations 1994 (WIPO Lex)](https://www.wipo.int/wipolex/en/legislation/details/3134) - [Shining Bright Malaysia: The Glittering World of High-End Jewellery Industry (MIDA, 30 June 2023)](https://www.mida.gov.my/shining-bright-malaysia-the-glittering-world-of-high-end-jewellery-industry/) --- ## ChatGPT Property Developer Visibility in Malaysia > Why Malaysian property developers with strong launches and portal listings still go unnamed in ChatGPT answers, and a practical plan to check and fix it. URL: https://tryshubi.com/insights/chatgpt-property-developer-malaysia/ Published: 2026-11-12 Updated: 2026-11-12 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: property Topics: AI visibility, AEO, property, Malaysia, ChatGPT ### Short answer Malaysian property developers go missing from ChatGPT answers when their project facts are locked in brochures, named inconsistently, or found mostly on portals and news under other labels. The fix is plain-text project pages that pair the developer and project names, consistent naming everywhere, Organization markup, third-party coverage and a fixed monthly prompt check. ### Key points - ChatGPT search can only cite your site if OAI-SearchBot is allowed to crawl it, so check robots.txt and your CDN before anything else. - Every launch needs one HTML page that states the developer's legal name, project name, location, tenure, unit types and KPKT permit details in plain text. - Use one spelling of the developer and project names across the corporate site, Bursa filings, portals, LinkedIn and press releases. - Buyers ask AI in English, Bahasa Malaysia and Mandarin, so test and publish in the languages your buyers actually use. - Track mention rate, accuracy and cited sources from a fixed prompt set every month, and treat single answers as samples, not verdicts. Your launches sell. The showroom in Cyberjaya or Bukit Jalil is busy on weekends, your projects sit near the top of PropertyGuru and iProperty, and the corporate site ranks for your own township names. Yet when a buyer asks ChatGPT which developers to trust for a landed home in the Klang Valley, your name is missing, and that is the gap in **ChatGPT property developer** visibility that marketing heads in Malaysia are now being asked to explain to the board. This article explains why it happens, what Malaysian buyers actually type into AI tools, how to audit your position in an afternoon, and the specific fixes that make a developer easier for ChatGPT to find, describe correctly and cite. None of it depends on hiring anyone. ### Why ChatGPT property developer answers leave you out ChatGPT leaves a developer out when it cannot find clear, crawlable text that pairs the developer's name with the projects, places and facts the buyer asked about. It does not reward ad spend or showroom traffic. It rewards pages it can read and quote. #### Two ways ChatGPT forms an answer There are two routes. The first is what the model learned during training, which is fixed at a point in time and fuzzy on recent launches. The second is live search: when ChatGPT searches the web, it reads current pages and shows the sources it used. OpenAI's crawler documentation is explicit about the second route. It says OAI-SearchBot "is used to surface websites in search results in ChatGPT's search features", and that sites which opt out of OAI-SearchBot will not be shown in ChatGPT search answers. A separate crawler, GPTBot, covers model training, and each robots.txt setting is independent. Many developer sites block unfamiliar bots at the CDN or firewall level without anyone in marketing knowing. #### Why portals and news win the citation Think about what a buyer's question needs: a project name, a developer, a location, a price range, tenure, completion status. A PropertyGuru listing, an EdgeProp review or a Star Property launch report often puts all of that in one readable paragraph. Your own launch microsite may put it in a hero video, a downloadable e-brochure and an image of the price table. AI tools quote what they can parse. Google's guidance on its own AI features makes the same point for Search: it recommends "making sure that important content is available in textual form", and says there are no special technical requirements beyond normal indexing. ChatGPT is a different product, but the lesson carries over. Facts trapped in PDFs and images are hard to cite. #### Why naming breaks the link Malaysian developers are often groups with listed parents, subsidiaries, joint ventures and township brands. A buyer may know the township, the press may use the parent group, the SPA carries the subsidiary's legal name and the portal shows yet another variation. If those labels never appear together in plain text, an AI system has little reason to connect "the township near the new highway interchange" with your company. ### What buyers in Malaysia actually ask AI Buyers ask AI about developers by reputation, location, price and risk, and they ask in English, Bahasa Malaysia and Mandarin, often mixed. The prompts below are realistic patterns to test, grouped by buying stage. **Exploring** - "Which property developers in Malaysia have a good track record for landed homes in Selangor?" - "Best townships in the Klang Valley for young families under RM800k" - "Pemaju hartanah yang dipercayai di Johor Bahru untuk rumah teres" **Comparing** - "Compare [Township A] and [Township B] in Semenyih for build quality and amenities" - "Is [project name] freehold or leasehold, and who is the developer?" - "吉隆坡哪家发展商的公寓质量比较好？" (Which KL developers build better-quality condos?) **Checking risk before booking** - "Has [developer name] had any abandoned or delayed projects?" - "Does [project name] have a valid KPKT advertising and sales permit?" That last group matters more than many marketing teams expect. In September 2026, EdgeProp reported that the Ministry of Housing and Local Government (KPKT) urged buyers to check developers on the TEDUH portal before buying, including licence status, advertising and sales permits (APDL) and whether projects are classed as delayed, sick or abandoned. Buyers who have absorbed "check before you buy" will ask AI the same questions, and the answer they get is shaped by whatever the AI can find. ### How to check where you stand You can run a useful baseline audit this week with a spreadsheet, two or three colleagues and a list of prompts. The aim is a record you can repeat, not a one-off screenshot. Our [measurement method](https://tryshubi.com/how-we-measure/) describes the same approach at a larger scale. #### Step 1: Confirm OpenAI can crawl you Open `yourdomain.com/robots.txt` and look for rules covering OAI-SearchBot and GPTBot. Then ask your IT or web vendor whether the CDN, WAF or bot-protection service challenges or blocks them. Do this for the corporate site and every launch microsite, since microsites are often built by different agencies on different hosting. #### Step 2: Build a fixed prompt set Write 20 to 40 prompts across the three stages above. Cover each region where you sell (Klang Valley, Johor, Penang, Perak), each language your buyers use and each flagship project. Include at least five prompts that name your developer directly, so you can check accuracy as well as presence. #### Step 3: Run each prompt several times Use a logged-out browser or a fresh account, and run each prompt three times. Answers vary between runs and users, so one result proves little. Note whether ChatGPT searched the web for that answer, because that tells you whether live pages are in play. #### Step 4: Log four things per answer For each run, record: whether your developer is named, whether the facts are correct (tenure, location, developer of record, completion), which competitors appear, and which sources are cited. The cited-sources column is the most useful one. It shows you exactly which pages ChatGPT trusts for your category. #### Step 5: Compare against your own pages Take the five prompts where you were missing or described wrongly. Open your own page for that project and ask: could someone answer this question from the visible text alone, without opening a PDF or watching a video? If not, you have found your first fix. ### How to fix it Make your own pages the clearest, most consistent source of developer and project facts, then make sure trusted third parties repeat those facts. The steps below are specific to how Malaysian developers publish and sell. #### Publish one fact page per project Each launch needs a permanent HTML page, not only a campaign microsite that disappears after sales close. In the first paragraph, state: project name, developer of record (legal name), parent group if relevant, location by mukim or area and state, tenure, unit types, built-up range and indicative price range. Put the KPKT advertising and sales permit details and developer licence number on the page in text, as they already appear on your ads and brochures under the housing development regulations. Add a short fact table. AI systems and buyers both find tables easy to read: | Fact | Example of what to state | | --- | --- | | Developer of record | Full legal name of the subsidiary, plus parent group | | Location | Township, area, district and state | | Tenure | Freehold or leasehold, with lease term | | Status | Launched, under construction, vacant possession date | | Permits | APDL and developer licence numbers, validity | | Official channels | Sales gallery address, official website URL | #### Fix your naming everywhere Pick one form for the group name, each subsidiary and each township, and use it on the corporate site, Bursa Malaysia announcements, annual reports, LinkedIn, Google Business Profiles for sales galleries, portal developer pages and press releases. Where a joint venture is involved, say plainly on the project page who the partners are and which entity is the developer of record. This one sentence often resolves AI confusion about "who built this". #### Add Organization structured data Google says Organization markup helps it "disambiguate your organization", with properties such as `legalName`, `url`, `logo` and `sameAs` (links to other pages about you, such as your LinkedIn or Bursa listing). Google is clear that structured data must match the visible text. It is not a ChatGPT ranking switch, but it reinforces the same entity facts in a machine-readable form and costs little to add. Our [technical AI visibility work](https://tryshubi.com/services/#technical) usually starts here. #### Answer the questions buyers ask, in their words Add a short FAQ to each project page and to the corporate "About" page, written the way buyers ask: "Who is the developer of [project]?", "Is [project] freehold?", "When is vacant possession?", "Is [project] registered on TEDUH?". Keep answers factual and short. Avoid unverifiable superlatives like "best investment in Johor"; they read as advertising, may sit badly with housing advertising rules, and AI tools tend not to repeat them. #### Publish in Bahasa Malaysia and Chinese where buyers do If your Johor or Penang buyers search in Mandarin, or your Selangor affordable launches attract Malay-speaking first-time buyers, publish real translated fact pages, not auto-translated banners. Keep the developer and project names identical across languages so the entity stays connected. Our [content service](https://tryshubi.com/services/#content) covers this kind of multilingual fact writing. #### Earn coverage that pairs your name with the project The citations you saw in Step 4 tell you which outside sources matter. For most developers that means property media (EdgeProp, Star Property, The Edge), portal developer profiles, REHDA membership listings and award pages, and business press for listed groups. When you issue a launch release, put the developer of record and project name together in the first sentence, so any article that reuses it carries the link AI needs. #### Keep facts current after launch Price ranges, phases, completion dates and handover status change. Update the project page when they do, and add a visible "last updated" date. If ChatGPT is still describing a phase as "upcoming" a year after handover, the outdated fact usually lives on a page you control or one you can ask a portal to correct. ### What to measure Measure four things monthly from the same prompt set: mention rate, accuracy, citation share and competitor presence. Treat them as trends over several months, because individual answers shift between runs and users. - **Mention rate:** the share of prompt runs in which your developer is named, split by region, language and stage. - **Accuracy:** the share of mentions with correct developer of record, tenure, location and status. A wrong fact about a permit or tenure is worth fixing before anything else. - **Citation share:** how often your own pages are among the cited sources, compared with portals, media and competitors. - **Competitor presence:** which developers appear in the prompts where you do not, and which sources back them. Report these to leadership with the method stated: number of prompts, runs per prompt, dates and engines tested. A small, honest panel tracked for six months says more than a large one-off test. Expect movement to be uneven; a new project page may be cited within weeks for one prompt and not at all for another. ### Common mistakes - **Blocking OpenAI's crawlers by accident** through a CDN rule or bot shield, then wondering why the site is never cited. - **Relying on e-brochures and image price lists** for the facts buyers ask about most. - **Using different names** for the same township or subsidiary on the website, portals and press releases. - **Taking campaign microsites offline** after sales close, which removes the only page that described the project. - **Writing for ads, not answers:** superlatives and lifestyle copy with no checkable facts give AI nothing to quote. - **Judging visibility from one screenshot** shared in a management meeting instead of a repeatable prompt panel. ### How Shubi helps Shubi is an AI visibility company based in Malaysia. We run the prompt panels, crawler checks and source analysis described above for developers across their regions and languages, then help fix the project facts and naming that AI tools rely on. Our [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) service is the usual starting point, and we do not promise rankings or guaranteed mentions. If you want to talk it through, [Message Shubi on WhatsApp](https://wa.me/601137634679). ### Frequently asked questions #### Why does ChatGPT name the portal or a news site instead of us? When ChatGPT searches the web, it cites pages that answer the question in clear text. Portal listings and news reports often pair your project with a price, location and developer name in one paragraph, while your own site may hide those facts in a PDF or image. Make your page the clearest answer and you give it a fair chance of being cited. #### Does strong Google SEO mean we will show up in ChatGPT? Not automatically. Good SEO helps, because both depend on crawlable, well-linked pages, but ChatGPT uses its own crawler and its own way of choosing sources. You still need to check that OpenAI's search crawler can reach you and that your pages state the facts buyers ask about. #### Can we put sales claims like 'best ROI in Klang Valley' on project pages to get mentioned? Avoid it. Property advertising in Malaysia is regulated under the Housing Development (Control and Licensing) Act 1966, and AI tools repeat claims they cannot verify poorly or not at all. Factual, checkable statements work better for buyers, regulators and AI answers alike. #### How often should we check our ChatGPT visibility? Monthly is enough for most developers, with an extra check after a major launch or rebrand. Use the same prompts each time, run each one several times, and compare trends rather than reacting to a single answer, because results vary between runs and users. ### Sources - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Organization structured data (Google Search Central)](https://developers.google.com/search/docs/appearance/structured-data/organization) - [KPKT urges homebuyers to check developers on TEDUH before buying (EdgeProp)](https://www.edgeprop.my/content/1917350/kpkt-urges-homebuyers-check-developers-teduh-buying) --- ## Indian hospital chains: AI visibility across Asia > Why multi-city Indian hospital chains get left out of, or misdescribed in, AI answers for patients in Bangladesh, Nepal and wider Asia, and how to fix it within NMC rules. URL: https://tryshubi.com/insights/india-hospital-chain-ai-visibility-asia/ Published: 2026-11-12 Updated: 2026-11-12 Author: Md Shariar Shanaz Shuvon Market: India Industry: medical Topics: AI visibility, AEO, medical, India, hospital chains, ChatGPT, Google AI Overviews ### Short answer Indian hospital chains lose AI visibility across Asia when their facts are scattered across city microsites, old brand names and facilitator portals. Fix it with one international patient hub, a clear page per hospital and centre, consistent doctor profiles, verifiable accreditation and visa process details, and claims that stay within the NMC's October 2026 advertising guidelines. ### Key points - Cross-border patients often ask an AI assistant which Indian city and hospital to choose before they ever contact an international patient desk. - AI engines build answers from what they can retrieve, so a chain whose facts are split across city sites and old brand names is described inconsistently or skipped. - The NMC's October 2026 guidelines, as reported, limit hospital advertising to factual, verifiable particulars, which is also the kind of content AI systems can quote safely. - Accreditation, visa letter processes and international desk contacts should be stated once, accurately, and repeated identically wherever the chain appears. - Measure mention rate, citation sources and factual accuracy with fixed prompts in English, Bengali and Nepali, and expect answers to vary between runs. A family in Dhaka or Kathmandu weighing cardiac surgery in Chennai, Kolkata or Delhi now often starts by asking ChatGPT or Gemini which hospital to choose. For Indian hospital chains, AI visibility across Asia decides whether your name appears in that first shortlist, and whether the answer gets your cities, specialities and accreditation right. Many chains find that a rival, a facilitator portal or a hospital in Malaysia or Thailand is named instead. This article is for international patient, digital and brand leads at multi-city Indian hospital groups. It explains why the gap happens, the prompts cross-border patients actually type, how to audit your position this week, and how to fix it without breaking the National Medical Commission's advertising rules. ### Why AI assistants skip or misdescribe Indian hospital chains AI assistants describe a hospital chain using whatever they can retrieve and reconcile about it, so a chain whose facts are fragmented gets a fragmented answer. The usual causes are structural, not a lack of clinical reputation. Google's documentation on [AI features](https://developers.google.com/search/docs/appearance/ai-features) says AI Overviews and AI Mode may use "query fan-out", issuing several related searches across subtopics to build a response. A prompt like "best hospital in India for liver transplant for a patient from Bangladesh" fans out into searches about the procedure, the cities, visas and costs. Your chain needs a clear, indexed page for each of those subtopics, or another source fills the gap. Google also says there are no special requirements for AI features beyond being indexed and eligible for a snippet, so ordinary crawlability problems carry straight through. ChatGPT search works through OpenAI's crawler. [OpenAI's crawler documentation](https://developers.openai.com/api/docs/bots) says OAI-SearchBot is used to surface websites in ChatGPT's search features, and that sites which opt out will not appear in ChatGPT search answers. Some hospital IT teams block every AI bot by default, which quietly removes the chain from those answers. Chains have specific weaknesses that make this worse: - **City microsites and separate domains.** One hospital sits on the group domain, another on a legacy domain from an acquisition, a third on a city subdomain. AI systems may treat them as unrelated organisations. - **Old and new brand names.** After a merger or rebrand, press coverage and directories keep the old name for years. Answers may mix the two or describe a hospital under a name it no longer uses. - **Doctors listed three times.** The same consultant appears on the group site, the hospital site and a booking platform, with different titles and departments. - **Facilitator portals with stronger pages.** Medical travel aggregators often publish tidy "top hospitals in India for X" pages. If your own pages are thinner, the portal becomes the cited source and its wording becomes your description. ### What patients across Asia actually ask AI Cross-border patients ask about cities, procedures, visas and trust before they ask about any single hospital. These prompts are typical of what international desks hear and what patients type, grouped by stage. **Early research** - "Which Indian city is best for heart bypass surgery for a patient from Bangladesh?" - "চেন্নাইতে বাইপাস সার্জারির জন্য কোন হাসপাতাল ভালো?" (Bengali: "Which hospital in Chennai is good for bypass surgery?") - "दिल्लीमा क्यान्सर उपचारका लागि राम्रो अस्पताल कुन हो?" (Nepali: "Which is a good hospital in Delhi for cancer treatment?") **Comparing options** - "Kidney transplant in India or Malaysia, which is better for a Sri Lankan patient?" - "Is [chain name] Kolkata NABH accredited, and does it have a paediatric cardiac unit?" - "Which hospitals in Hyderabad have international patient services for patients from the Maldives?" **Ready to act** - "How do I get an Indian medical visa letter from [hospital name]?" - "Can two family members travel with me on a medical attendant visa to India?" Notice how many prompts need practical, factual answers. The Government of India's [e-Visa page](https://indianvisaonline.gov.in/evisa/tvoa.html) lists e-Medical and e-Medical Attendant visas, says the e-Medical visa requires a letter from the hospital in India on its letterhead, and says only two e-Medical Attendant visas are granted against one e-Medical visa. A hospital that explains its own letter process clearly gives AI something accurate to quote. ### How to check where you stand You can run a useful baseline audit in a week with a spreadsheet and a fixed prompt list. Our [how we measure](https://tryshubi.com/how-we-measure/) page explains the method we use for repeatable testing. #### Step 1: Build a prompt set by country and speciality Take your top five specialities for international patients and your top four source countries. Write two or three prompts per pair, in English and in the language patients use. Keep the wording fixed so you can repeat it next month. #### Step 2: Run each prompt several times across engines Run each prompt in ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews, at least three times each. Answers vary between runs, users and locations, so a single run proves little. Record whether your chain is named, which hospital in the chain, and which sources are cited. #### Step 3: Check every fact the answers state For each mention, check the hospital name, city, departments, accreditation, contact route and visa guidance against your own records. Mark each answer accurate, partly wrong or wrong. Incorrect accreditation and outdated names are the errors with the most compliance risk. #### Step 4: List the sources that win Collect the URLs the engines cite. Group them: your own pages, facilitator portals, news, directories, competitors. This tells you whether you have a content problem (your pages are missing) or an authority problem (others are trusted more). #### Step 5: Check crawler access and indexing Ask IT for the robots.txt rules on every domain the chain uses. Check that OAI-SearchBot and Googlebot can reach the international patient pages, and that those pages are indexed, not hidden behind enquiry forms or scripts. ### How to fix it The fix is to make one accurate, factual version of the chain easy to find and repeat everywhere. Conveniently, this is also what Indian advertising rules now expect. #### Work inside the NMC's October 2026 guidelines The NMC issued guidelines on ethical advertising and public communication for doctors, hospitals and medical institutions in October 2026. [Medical Dialogues' report on the guidelines](https://medicaldialogues.in/health-news/nmc/nmc-issues-ethical-advertising-public-communication-guidelines-for-doctors-hospitals-details-180731) says clinical establishments may advertise only factual, objective and verifiable particulars such as name, location, departments, facilities, equipment, accreditation status and charges. It also reports prohibitions on testimonials and patient stories for promotion, advertised surgical results, unverifiable superiority claims such as "best" or "100% success", fake reviews or manipulated rankings, commissions tied to patient procurement, and AI-generated content that creates misleading claims or synthetic endorsements. Read the notice itself with your legal team before changing anything. The practical point for AI visibility is that the content you are allowed to publish, verifiable facts, is exactly the content AI systems can quote without distortion. Strip "No.1" language and outcome boasts from international pages, and review any facilitator arrangements that pay per referred patient. #### Build one international patient hub for the whole chain Create a single hub on the group domain that lists every hospital, its city, its main departments for international patients and the international desk contact for each. Use the same hospital names everywhere, including any former names stated once ("formerly known as..."). Link from each city site back to this hub and from the hub to each hospital page. #### Give each hospital and centre a page AI can quote Each hospital needs a page that opens with a plain summary: what it is, where it is, which departments it offers, its accreditation status with the date it was verified, and how international patients get in touch. Each centre, such as a transplant or oncology unit, needs its own page with the procedures offered, equipment and the consultants who lead it. Write the first paragraph so it can stand alone as an answer. #### Make accreditation verifiable AI answers often state accreditation, and get it wrong when chains describe it loosely. [NABH](https://nabh.co/) runs a public search of accredited healthcare organisations and a certificate verification tool. Name the exact accredited entity, link to the verification route, and remove accreditation claims for hospitals whose accreditation has lapsed. Apply the same discipline to any international accreditation you mention. #### Explain the visa and arrival process in your own words Publish a page per hospital on how to request the visa letter, what documents the desk needs, typical response time if you can commit to one, and what attendants need. Link to the official e-Visa site rather than restating rules that may change. This answers the "ready to act" prompts directly and keeps patients away from unofficial agents. #### Clean up doctor profiles and structured data Pick one canonical profile URL per doctor, with the qualifications, specialities and registration details the NMC guidelines treat as patient information. Redirect duplicates. Add schema.org markup such as `Hospital`, `MedicalOrganization` and `Physician`, and link each hospital to the parent organisation. Google's guidance says structured data should match the visible text on the page, so do not mark up facts you have not published. #### Correct the third-party sources AI already reads | Source type | Why AI may lean on it | What to do | | --- | --- | --- | | Facilitator and medical travel portals | Detailed, comparison-style pages that match patient prompts | Ask listed portals to correct names, departments and accreditation; avoid paid ranking or commission arrangements the NMC guidelines restrict | | Google Business Profiles | Used for city, address and contact facts | One profile per hospital, consistent names, international desk number where appropriate | | NABH directory | Independent confirmation of accreditation | Check the listed entity name matches what your pages say | | News and trade press | Explains mergers, new centres and rebrands | Issue factual announcements when names, ownership or centres change | | Country-language content | Answers prompts typed in Bengali, Nepali or Dhivehi | Translate the hub and key centre pages properly, reviewed by a native speaker | ### What to measure Track four numbers monthly against the same prompt set, and treat month-to-month movement of a few points as noise until it persists. - **Mention rate:** the share of prompt runs where any hospital in the chain is named, by country and speciality. - **Citation share:** the share of cited URLs that are your own pages rather than portals or competitors. - **Accuracy:** the share of mentions with correct name, city, departments, accreditation and contact route. - **Framing:** whether answers describe you in factual terms or repeat unverifiable claims from third parties. Run the full set monthly and a short set of ten priority prompts weekly during big changes such as a rebrand. Record the engine, date, language and location for every run, because answers vary between runs and users and you will need that context to explain a change. ### Common mistakes - Publishing patient stories or "success rates" on international pages because AI "likes stories", which conflicts with the NMC guidelines as reported. - Letting each hospital in the chain describe its accreditation, departments and international desk differently. - Blocking all AI crawlers by default, then wondering why ChatGPT search never cites the chain. - Testing only English prompts from an Indian location, when patients type in Bengali or Nepali from their own country. - Leaving acquired hospitals on legacy domains with no visible link to the parent brand. - Treating a single good answer as success instead of tracking results over many runs. ### How Shubi helps Shubi tracks how AI assistants mention and describe brands across Asia, using fixed prompt sets in the countries and languages your patients use, and reports mention rate, citations and factual accuracy over time. See [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) for how the measurement works, and [content](https://tryshubi.com/services/#content) and [technical](https://tryshubi.com/services/#technical) services for the fixes. We do not promise mentions or rankings, and we keep recommendations within NMC advertising rules. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Can we publish patient success stories so AI has something to quote? Not for promotion. Reporting on the NMC's October 2026 guidelines says testimonials, patient stories, before-and-after photos and advertised surgical results are prohibited for promotional purposes. Give AI systems factual material instead: departments, procedures offered, equipment, accreditation status and how the international patient process works. #### Should each hospital in the chain have its own website? It can, but each one should point clearly to the parent brand and use the same names, accreditation wording and international desk details. Separate sites with different facts are a common reason AI assistants describe one hospital correctly and its sister hospitals wrongly or not at all. #### Do we need to block or allow AI crawlers? That is a policy choice, but blocking search crawlers limits where you can appear. OpenAI's documentation says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, and that this is separate from GPTBot, which relates to model training. #### Which countries should our prompt tests cover? Start with the countries your international patient desk already serves, such as Bangladesh, Nepal, Sri Lanka, the Maldives and Myanmar, plus any corporate or insurer markets you are pursuing in Malaysia or Singapore. Test in the languages those patients actually type in, not only English. ### Sources - [NMC issues ethical advertising, public communication guidelines for doctors, hospitals (Medical Dialogues, October 2026)](https://medicaldialogues.in/health-news/nmc/nmc-issues-ethical-advertising-public-communication-guidelines-for-doctors-hospitals-details-180731) - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Overview of OpenAI crawlers (OpenAI)](https://developers.openai.com/api/docs/bots) - [National Accreditation Board for Hospitals & Healthcare Providers (NABH)](https://nabh.co/) - [e-Visa categories, including e-Medical and e-Medical Attendant (Government of India)](https://indianvisaonline.gov.in/evisa/tvoa.html) --- ## KL jeweller: Perplexity keeps citing the same rivals > Why Perplexity keeps footnoting the same competitors when Klang Valley buyers research jewellers, and a step-by-step plan to make your own pages citable. URL: https://tryshubi.com/insights/kl-jeweler-perplexity-rivals/ Published: 2026-11-12 Updated: 2026-11-12 Author: Md Shariar Shanaz Shuvon Market: Malaysia Industry: jewelry Topics: AI visibility, AEO, jewelry, Malaysia, Perplexity ### Short answer Perplexity searches the web live and footnotes the pages it can quote. If a KL jeweller keeps losing to the same rivals, those rivals usually own the cited listicles, forum threads and clear guide pages. Audit which URLs get cited, publish plain buying guides and branch facts, and earn mentions on the sources Perplexity already trusts. ### Key points - Perplexity shows its sources as numbered citations, so you can see exactly which URLs your rivals win with. - Most citations for KL jewellery research prompts come from listicles, forums and guide pages, not brand lookbooks. - Branch facts such as outlet location, parking, hours, services and appointment rules need one consistent version everywhere. - Neutral education on 916 and 999 gold, workmanship charges and trade-ins is more quotable than campaign copy. - Purity and pricing claims must follow Malaysia's Trade Descriptions rules, enforced by KPDN. - Track cited URLs monthly across English, Malay and Chinese prompts, because answers vary between runs and users. You run digital for a KL jeweller with outlets across the Klang Valley. Your team types "best jeweller in KL for engagement rings" into Perplexity and the answer names the same three rivals, with the same handful of footnotes underneath. Your group, with more outlets and a longer history, is either missing or mentioned once without a link. This article explains why Perplexity keeps citing those rivals, gives you the prompts Klang Valley buyers actually type, and walks through an audit you can run this week. It then covers the fixes that tend to move citations for jewellery groups: branch facts, buying guides Perplexity can quote, structured data and the third-party pages it already trusts. ### Why Perplexity keeps citing the same KL jewellery rivals Perplexity cites the pages it finds and can quote at the moment of the question, so the rivals in your footnotes are the ones whose pages answer buyer questions most directly. It is a contest between URLs, not between showrooms. Perplexity's help centre says it searches the internet in real time and that every answer includes numbered citations linking to the original sources. That is different from a chatbot answering mostly from training data. When someone asks about jewellers in KL, Perplexity runs searches, reads a set of pages and builds its answer from passages on those pages. The brands named in the answer are mostly the brands named on the pages it read. For a jewellery research prompt, the pages that win tend to be: - **"Best jewellers in KL" listicles** on lifestyle, wedding and city guide sites, which name five to fifteen shops in one quotable list. - **Forum and community threads** where couples compare engagement ring shops or ask where to buy 916 gold without a high workmanship charge. - **Guide pages** that explain one thing well, such as the difference between 916 and 999 gold, or how lab-grown and natural diamonds compare. - **Brand pages** only when they state facts plainly: which outlet does resizing, whether you need an appointment, what certificate comes with a diamond. If your rivals appear on more of those listicles, or if their sites have a clear "how to choose an engagement ring" guide and yours has a lookbook, Perplexity has more material to quote about them. Once a few pages repeat the same short list of names, the pattern reinforces itself across related prompts. Crawl access matters too. Perplexity's documentation says PerplexityBot is designed to surface and link websites in its search results and is not used to crawl content for training foundation models. If your robots.txt or security settings block it, your own pages are less likely to be read and cited, even when they are the best answer. ### What buyers in KL actually ask AI Klang Valley jewellery buyers mix English, Malay and Chinese, and they ask specific, practical questions. These are the kinds of prompts to test, grouped by buying stage. **Early research** - "Best jeweller in KL for engagement rings under RM10,000" - "Kedai emas yang dipercayai di Kuala Lumpur untuk beli cincin tunang" - "吉隆坡哪里买916金首饰比较好" (where in KL is good for buying 916 gold jewellery) **Comparing options** - "916 vs 999 gold for a wedding gift, which is better and where to buy in Klang Valley" - "Lab-grown vs natural diamond engagement ring, which KL jewellers sell both with certificates" - "Which gold shops in KL have the lowest workmanship charge (upah)" **Ready to visit** - "Jeweller in Mid Valley or Bangsar that does ring resizing same day" - "Kedai emas dekat Petaling Jaya yang terima trade-in emas lama" Notice how many of these turn on facts that only you can supply accurately: which outlet resizes, whether trade-ins are accepted, what certificates you issue. Those are the gaps where your own pages can compete with listicles. ### How to check where you stand You can measure your Perplexity position in a few hours because Perplexity shows its sources. The audit below gives you a list of URLs to win, not just a feeling that rivals are ahead. For more on why we sample prompts repeatedly, see [how we measure](https://tryshubi.com/how-we-measure/). #### Step 1: Build a prompt set Write 20 to 30 prompts across the three stages above, in English, Malay and Chinese. Include outlet areas your group serves, such as Jalan Masjid India, Bukit Bintang, Mid Valley, Petaling Jaya, Subang and Cheras. Keep the wording close to how customers speak to your staff. #### Step 2: Run each prompt more than once Answers vary between runs and users, so run each prompt at least three times, ideally logged out and from a Malaysian connection. Record whether your brand is named, where it appears in the answer, and how it is described. #### Step 3: Log every cited URL Copy each numbered source into a spreadsheet with the prompt, date and which brands that page names. After a full run you will usually see a short list of URLs doing most of the work. Those are your real competitors in Perplexity, often more than the rival brands themselves. #### Step 4: Read the cited pages For each frequently cited URL, note why it is quotable. Does it have a clear list? A table of outlets? A direct answer in the first paragraph? Is your brand missing, or mentioned with an old outlet or wrong service? This tells you whether to fix your own site, correct a third-party page, or both. #### Step 5: Check your own site's crawlability Confirm robots.txt does not block PerplexityBot, that your outlet and guide pages load without heavy scripts or login walls, and that key facts are in text rather than inside images. Many jewellery sites put purity, hours and services inside banner graphics that no crawler can read. ### How to fix it The fix is to give Perplexity better pages to quote, on your site and on the sources it already reads. Work through these in order of what your audit showed. #### Publish one clear page per outlet Each outlet needs its own page with the full address, mall level and lot number, parking and nearest LRT or MRT station, opening hours, phone or WhatsApp line, and the services available there: resizing, engraving, repairs, custom design, trade-in, appointment-only viewing. Write these as plain sentences and short lists. If resizing happens only at the flagship workshop, say so and say how long it takes. Perplexity answers "same-day resizing in Mid Valley" questions from exactly this kind of detail, and if you do not state it, a forum thread will state it for you, correctly or not. #### Write buying guides Perplexity can quote Lookbooks and campaign pages rarely get cited for research prompts. Neutral education does. Good candidates for a KL jewellery group: - 916 vs 999 gold: durability, colour and what each is typically used for - How gold jewellery pricing works: gold price, workmanship charge, and how trade-in values are calculated at your outlets - Choosing an engagement ring: certificates, the 4Cs, lab-grown vs natural - Caring for gold and diamond pieces in Malaysian humidity Open each guide with a two or three sentence direct answer, then the detail. Use your gemmologist or senior goldsmith as a named author where compliance agrees. Keep promotional lines to a short note at the end. #### Keep purity and price claims compliant Every purity statement must be accurate. The Trade Descriptions (Articles Made of Precious Metals) Regulations 1994, made under Malaysia's trade descriptions law, set the fineness standards and marking rules for gold articles; for example, they require gold of 916.6 fineness to be marked as 916. The Trade Descriptions Act 2011 now sits above them, and KPDN enforces it. Enforcement is active in KL. In January 2026 Malay Mail reported that KPDN seized more than 3,500 pieces of gold stamped "AU 999.9" from premises in Kuchai Lama after testing found much lower gold content. Buyers now ask AI how to check purity, which is an opening for a clear, honest guide from an established group. It is also a reason to have compliance check every comparison and pricing page before you publish. #### Add structured data for every outlet Google Search Central recommends defining each business location with the LocalBusiness type, using the most specific subtype available, and supports `openingHoursSpecification` and a `department` property for distinct departments with their own hours or phone numbers. Use `JewelryStore` for each outlet page. Structured data is documented by Google rather than Perplexity, so treat it as a way to keep machine-readable facts consistent across search engines, not a direct switch for Perplexity citations. Its real value is forcing one source of truth for hours, addresses and phone numbers. #### Earn mentions on the sources Perplexity already cites Your audit gives you a list of the listicles and threads behind your rivals' citations. Work through it honestly: | Source type | Why Perplexity uses it | What to do | | --- | --- | --- | | "Best jewellers in KL" listicles | One page answers the whole question with names | Send editors accurate outlet facts and a short note on what you are known for; ask for corrections where you are listed wrongly | | Wedding planning sites and vendor directories | Specific to engagement and wedding band prompts | Keep your vendor profile complete, current and consistent with your site | | Forums and community threads | Real buyer experiences in plain language | Respond openly as the brand where rules allow; never post as a fake customer | | Business press and mall directories | Trusted, factual, often cited for "who is" prompts | Keep mall tenant listings correct and pitch genuine stories, such as a new workshop or heritage milestone | | Google Business Profiles | Hours, address and reviews per outlet | Keep each outlet's profile identical to its page on your site | Do not try to game these sources. Undisclosed paid placements and planted reviews can be removed, and when a buyer notices them they undermine the trust a jeweller depends on. #### Match the languages your buyers use If a large share of your customers ask in Malay or Chinese, publish your key guides and outlet pages in those languages too, written by people who speak them, not machine-translated word for word. Use the terms customers use: "emas 916", "upah", "cincin tunang", "金饰", "回收旧金". A Chinese-language prompt is more likely to be answered from Chinese-language pages. #### Keep every outlet consistent Multi-outlet groups often have three versions of the same facts: the website, Google Business Profiles and mall directories, each updated at a different time. Assign one owner for outlet data and update all channels on the same day when an outlet opens, moves, closes or changes hours. Inconsistent facts are a common reason AI answers describe a group wrongly or leave it out. ### What to measure Measure which URLs Perplexity cites and how often your brand appears, using the same prompt set every month. Expect movement over months, not days, and expect noise between runs. - **Mention rate:** the share of runs in which your brand is named, by prompt stage and language. - **Citation share:** the share of cited URLs that are your own pages, and the share of cited third-party pages that name you. - **Accuracy:** whether outlets, services, hours and purity information are described correctly. One wrong outlet address can send a customer to a closed shop. - **Description and sentiment:** how Perplexity characterises you compared with rivals, for example "heritage goldsmith" versus "budget option". - **Source movement:** which of the top cited URLs you have been added to or corrected on since the last run. Share a short monthly summary with outlet and regional managers. They hear the questions customers bring in and often spot wrong AI answers before head office does. ### Common mistakes - **Treating it as a Google ranking problem.** A page can rank well on Google and still not be quoted if it does not answer the question directly. - **Publishing lookbooks instead of answers.** Beautiful imagery does not give Perplexity sentences to quote. - **Blocking AI crawlers by default.** Security or CDN rules sometimes block PerplexityBot without anyone deciding to. - **Hiding facts in images.** Purity, hours and services inside banners are invisible to crawlers. - **Letting outlet data drift.** Old addresses on mall directories and listicles get repeated in answers long after an outlet moves. - **Chasing every listicle.** Focus on the URLs your audit shows Perplexity actually cites. ### How Shubi helps Shubi runs the prompt and citation tracking described above across Perplexity and the other major AI assistants, in English, Malay and Chinese, so your team can see which URLs drive rivals' citations and which facts about your outlets are wrong. We then help you build the guide and outlet pages and the third-party corrections that give AI engines better material to quote. We cannot guarantee Perplexity citations, but we can show you clearly where you stand each month. See [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) and [content for AI answers](https://tryshubi.com/services/#content), or [Message Shubi on WhatsApp](https://wa.me/601137634679). ### Frequently asked questions #### Is getting cited by Perplexity the same as ranking on Google? Not quite. Perplexity runs its own live search and shows the pages it used as numbered citations, so a page that ranks well on Google can still be skipped if it does not answer the question in a quotable way. Good SEO helps, but the test is whether a passage on your page states the answer clearly. #### Should we block PerplexityBot to protect our designs? Perplexity's documentation says PerplexityBot is used to surface and link websites in its search results, not to train foundation models. Blocking it in robots.txt reduces your chances of being cited. Protect design imagery with watermarks and image licensing rather than by hiding your guide and branch pages. #### Should we ask listicle writers to add our brand? Yes, if you approach it honestly. Send the editor accurate, verifiable facts about your outlets and services and let them decide. Paying for undisclosed placements or posting fake forum reviews breaks platform rules and can damage trust when it is found out. #### How long before our pages show up in Perplexity citations? There is no fixed timeline and no one can guarantee it. Perplexity searches the web at the time of each question, so new or updated pages can be picked up once they are crawled, but whether they are cited depends on the question and the competing sources. Track it monthly rather than daily. #### Can we publish gold prices and purity comparisons? You can, as long as every purity and pricing statement is accurate and matches what is stamped on the item. Malaysia's Trade Descriptions Act 2011 and the precious metals regulations govern these claims, and KPDN enforces them. Have compliance review comparison content before it goes live. ### Sources - [How does Perplexity work? (Perplexity Help Center)](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work) - [Perplexity Crawlers (Perplexity documentation)](https://docs.perplexity.ai/guides/bots) - [Local business (LocalBusiness) structured data (Google Search Central)](https://developers.google.com/search/docs/appearance/structured-data/local-business) - [Trade Descriptions (Articles Made of Precious Metals) Regulations 1994 (WIPO Lex)](https://www.wipo.int/wipolex/en/legislation/details/3134) - [KL domestic trade enforcers seize over 3,500 pieces of gold marketed as 'pure' (Malay Mail, 18 January 2026)](https://www.malaymail.com/news/malaysia/2026/01/18/what-you-see-isnt-always-what-you-get-kl-domestic-trade-enforcers-seize-over-3500-pieces-of-gold-marketed-as-pure/205885) --- ## Singapore luxury hotel AI citations: how to earn them > Why AI assistants cite booking sites and rival collections when planners ask about Singapore luxury hotels, and what a hotel group can do about it. URL: https://tryshubi.com/insights/singapore-luxury-hotel-ai-citations/ Published: 2026-11-12 Updated: 2026-11-12 Author: Md Shariar Shanaz Shuvon Market: Singapore Industry: luxury Topics: AI visibility, AEO, luxury, Singapore, Gemini ### Short answer Singapore luxury hotels earn AI citations by publishing the facts planners and guests ask about as plain text on their own sites: ballroom capacities, room counts, renovation dates, transit access and how each property in a collection differs. Keep Google Business Profile details current, earn trade press coverage, and sample real prompts monthly, accepting that answers vary. ### Key points - AI assistants cite booking sites when hotel websites hide key facts in PDFs, images or booking engines that are hard to quote. - Event planners ask for capacities, pillar-free ballrooms and walking time to the MRT, so those facts belong in HTML on the property page. - A collection page that compares each Singapore property by use case helps AI tell sister hotels apart instead of merging them. - Google says no special optimisation is needed for AI Overviews, but pages must be indexed and eligible to show a snippet. - Track mention rate, citation share and factual accuracy each month across Gemini, ChatGPT and Perplexity, knowing answers vary between runs. A travel manager in Raffles Place asks Gemini for a five-star hotel near Marina Bay with a pillar-free ballroom for 600 guests. The answer names three properties and cites a booking site, a listicle and a rival collection. Your hotel has the ballroom, the location and the loyalty members, yet it is not in the answer. That is the problem with **Singapore luxury hotel AI citations**: the engines quote whoever publishes the facts in a form they can lift, and that is often not the hotel. This guide is for marketing, brand and digital leaders at luxury hotels and multi-property collections in Singapore. It explains why AI assistants favour online travel agencies (OTAs) and aggregators, what planners and guests actually type, how to audit your position this week, and the concrete fixes that give AI a reason to cite your own pages. ### Why do OTAs and rivals win Singapore luxury hotel AI citations? AI assistants cite the sources that answer the question most directly in text they can retrieve and quote. Booking platforms publish room types, amenities and neighbourhood details in the same tidy format for every hotel, while many hotel sites hide the same facts in PDFs, image sliders and booking widgets. Most AI answers about hotels now come from live retrieval rather than memory alone. Google's documentation on [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features) says AI Overviews and AI Mode may issue "multiple related searches across subtopics and data sources" to build a response. A question about ballrooms near Marina Bay can fan out into searches about capacity, MRT access and recent reviews. Whichever page answers each sub-question cleanly has a chance to be linked. Google also says there are no special optimisations needed to appear in these features. A page must be indexed and eligible to show a snippet, and structured data should match the visible text. In practice, that means the banquet capacity table locked inside a downloadable kit is invisible to the part of the system that writes the answer. Other engines work along similar lines. Perplexity's help centre says it searches the internet in real time and that [each answer includes numbered citations](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work) linking to the original sources. OpenAI's [crawler documentation](https://developers.openai.com/api/docs/bots) states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Some hotel groups block AI crawlers at the corporate firewall or CDN without marketing knowing, which removes them from those answers entirely. Multi-flag collections have an extra problem. When a group runs a heritage hotel, a business hotel in the CBD and a resort on Sentosa, AI can merge facts across them or describe only the property with the most press coverage. Without a page that states plainly how the hotels differ, the engine fills the gap with whatever a travel blog wrote. ### What buyers in Singapore actually ask AI Buyers ask specific, practical questions, often in English mixed with Mandarin or Bahasa Indonesia, and usually with a date, a headcount or a neighbourhood attached. These are the kinds of prompts worth sampling, grouped by stage. **Shortlisting** - "Best luxury hotels in Singapore for a corporate offsite of 80 people, walking distance to an MRT station" - "Which five-star hotels near Marina Bay have a pillar-free ballroom for 500 to 700 guests?" - "新加坡适合办婚宴的五星级酒店，哪家有大宴会厅？" (Which five-star hotels in Singapore suit a wedding banquet and have a large ballroom?) **Comparing** - "Compare [Hotel A] and [Hotel B] for a regional leadership meeting during Singapore Grand Prix week" - "Hotel mewah di Orchard Road yang dekat mal dan cocok untuk keluarga" (Luxury hotels on Orchard Road close to the malls and good for families) - "Is [Hotel] still under renovation, and which restaurants are open?" **Deciding** - "Does [Hotel] have a club lounge, and what is included?" - "How long does it take from Changi Airport to [Hotel] by taxi?" Notice how many of these turn on facts the hotel controls: capacities, renovation status, lounge access, travel times. That is good news, because those are the facts you can publish. ### How to check where you stand Run a structured sample before changing anything, so you can see what improves. The steps below take a few hours across a small team. Our page on [how we measure](https://tryshubi.com/how-we-measure/) explains the wider method we use. #### Step 1: Build a prompt set from sales and events Ask your MICE, wedding and corporate sales teams for the questions they hear most from planners and couples. Turn them into 20 to 30 prompts across the three stages above, in English and in the other languages your guests use. Keep the wording natural, the way a busy planner would type it. #### Step 2: Run each prompt in each engine Run the set in Google (AI Overviews and AI Mode where shown), Gemini, ChatGPT with search and Perplexity. Use a clean browser session, note the date, and record every hotel named and every source cited. Answers vary between runs and users, so run important prompts more than once. #### Step 3: Score accuracy, not just presence For every mention of your hotels, check the facts: room count, ballroom capacity, opening hours of outlets, renovation status and neighbourhood. A citation that says your ballroom is closed for refurbishment can do more harm than no mention at all. #### Step 4: Map the cited sources List the domains AI cites for your category: OTAs, review sites, travel magazines, wedding directories, venue-finding platforms and rival hotel pages. This becomes your map of where AI already looks for Singapore luxury hotel information. #### Step 5: Check crawler access Ask IT or your web agency to confirm that robots.txt and any CDN or bot-protection rules allow Googlebot and OAI-SearchBot to reach your property pages. Many hotel sites sit behind brand-level platforms where these settings are decided centrally. ### How to fix it The fix is to make your own site the clearest primary source for each fact buyers ask about, then reinforce those facts in places AI already trusts. These are the actions that matter most for a Singapore luxury hotel or collection. #### Publish a Singapore collection page that compares your hotels If your group runs more than one property in Singapore, publish one page that states, in plain sentences, what each hotel is best for. For example: which suits a board retreat, which suits a 700-guest wedding banquet, which is closest to Suntec or Marina Bay Sands Expo, which is on Sentosa. Link each line to the property page. This gives AI a single passage it can quote when someone asks about your group, and it reduces the chance of facts bleeding between sister hotels. #### Put meeting and banquet facts in HTML Move capacities out of PDFs and into a text table on each property's meetings page: room name, size in square metres, ceiling height, pillar-free or not, and capacity by layout (theatre, banquet, classroom, cocktail). Add a short paragraph above the table that answers the common question directly, such as "The Grand Ballroom seats up to 600 guests for a banquet and is pillar-free." That is the sentence an AI answer can lift. #### Write neighbourhood and access paragraphs Planners ask about walking time to the MRT, taxi time from Changi, and how close a hotel is to Suntec, the CBD or Orchard Road. State these on the property page in plain text, using the station and road names people actually search. Avoid vague phrases like "steps from the city's best attractions", which no AI can quote as an answer. #### Keep renovation and reopening dates current Singapore's luxury hotels refresh wings, ballrooms and restaurants regularly, and AI answers can repeat old news for a long time. Publish a dated "what's open" note on each property page, update it the day something reopens, and make sure the same date appears in press releases and on your Google Business Profile. #### Keep Google Business Profile hotel details accurate Google's help page on [managing your hotel's details](https://support.google.com/business/answer/9177958?hl=en) says owners with a verified profile can edit hotel attributes under Hotel Details, and that Google shows a summary of amenities in Search and Maps. It also notes that hotel attributes cannot be updated through the spreadsheet upload. Assign one owner per property to check these details monthly, especially after outlet changes. #### Earn third-party sources AI already reads Your audit in Step 4 shows which outside sources the engines cite. Use the table below to decide where to spend effort. | Source type | Why AI tends to use it | What to do | | --- | --- | --- | | OTAs and booking platforms | Consistent, text-based facts for every hotel | Keep content, amenities and photos accurate and identical to your own site | | Travel trade and MICE press | Named, dated coverage of openings and venues | Pitch stories that name both the group and the property in the first paragraph | | Wedding and venue directories | Answer banquet and capacity questions directly | Supply capacities and packages that match your meetings page | | Review platforms | Fresh guest language about service and location | Respond to reviews and correct factual errors politely | | Official tourism and events bodies | Authority on Singapore as a destination | Make sure your venue facts are listed correctly wherever you are featured | The Singapore Tourism Board launched a [global campaign positioning Singapore as the "World's Best MICE City"](https://www.stb.gov.sg/about-stb/media-publications/media-centre/singapore-tourism-board-launches-global-mice-campaign-to-position-singapore-as-the--world-s-best-mice-city-/) in March 2024. Business events are a priority for the destination, so planners researching venues will keep asking AI for shortlists. Being described accurately in that context is worth the effort. #### Add structured data that matches the page Mark up each property with schema.org Hotel or a related type, including name, address, telephone, star rating where appropriate and amenities. Google's AI features documentation says no special schema is required, but structured data that matches the visible text helps engines understand which hotel a page describes. Treat it as support for clear writing, not a substitute. ### What to measure Measure four things monthly with the same prompt set: mention rate, citation share, accuracy and sentiment. Expect movement over months, not weeks, and expect noise between runs. - **Mention rate:** the share of prompts in which any of your hotels is named. - **Citation share:** how often your own domain is cited compared with OTAs and rivals. - **Accuracy:** the share of mentions with correct capacities, dates, outlets and location. - **Sentiment and framing:** whether AI describes the hotel the way you position it (for example "heritage" versus "business"). - **Group versus property split:** for collections, whether AI names the group, the property, or both. Report these by engine and by buyer stage. A rise in accuracy is often the first sign of progress, before mention rate moves. Because AI answers vary between runs and users, look at trends across several months rather than reacting to a single result. ### Common mistakes - Keeping banquet capacities and floor plans only in downloadable PDFs or images. - Letting the brand platform block AI search crawlers without marketing knowing. - Writing every property page in the same luxury adjectives, so AI cannot tell sister hotels apart. - Leaving "reopening soon" notices online months after a wing or ballroom has reopened. - Measuring only English prompts when wedding couples and regional guests search in Mandarin and Bahasa Indonesia. - Treating one AI answer as the truth instead of sampling the same prompts over time. ### How Shubi helps Shubi runs structured prompt sampling across Gemini, ChatGPT, Perplexity and Google's AI features for hotel groups, then shows where each property is named, cited or described wrongly. Our [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence) work turns that into a fix list, and our [content](https://tryshubi.com/services/#content) and [technical](https://tryshubi.com/services/#technical) teams help publish the pages and access settings AI needs. We cannot promise citations, but we can make sure the facts are there to be found. [Message Shubi on WhatsApp](https://wa.me/601137634679) ### Frequently asked questions #### Why do OTAs get cited instead of our hotel website? Booking platforms publish room types, amenities and location details in consistent, text-based layouts across thousands of hotels, which is easy for retrieval systems to quote. Many hotel sites keep the same facts in brochures, image carousels or booking widgets. Publishing those facts in plain text on your own pages gives AI a primary source to cite. #### Is Gemini the only AI engine a Singapore hotel should track? No. Sample Google AI Overviews and AI Mode, Gemini, ChatGPT and Perplexity, because corporate travel managers, wedding couples and MICE planners use different tools. Each engine retrieves and cites sources in its own way, so a hotel can be cited in one and missing from another. #### Do we need separate AI visibility work for each flag in our collection? Start with the group or collection page and the facts shared across properties, then fix property pages one by one. Clear links between the group entity and each hotel help AI describe sister properties correctly rather than confusing them. #### Can a hotel pay to be cited in AI answers? Organic citations in AI answers are not something a hotel can buy, and no one can guarantee them. What a hotel controls is whether accurate, quotable facts exist on its own site and in credible third-party sources the engines read. #### How often should we update renovation and reopening information? Update the property page on the day a wing, ballroom or restaurant reopens, and again when the work is announced. AI answers that still describe a hotel as under renovation can push planners towards competitors for months. ### Sources - [AI features and your website (Google Search Central)](https://developers.google.com/search/docs/appearance/ai-features) - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) - [How does Perplexity work? (Perplexity Help Center)](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work) - [Manage your hotel's details (Google Business Profile Help)](https://support.google.com/business/answer/9177958?hl=en) - [Singapore Tourism Board launches global MICE campaign to position Singapore as the World's Best MICE City](https://www.stb.gov.sg/about-stb/media-publications/media-centre/singapore-tourism-board-launches-global-mice-campaign-to-position-singapore-as-the--world-s-best-mice-city-/) --- ## Specialist clinics left off Copilot shortlists in Singapore > Why Singapore specialist clinic groups go missing when HR teams and insured employees ask Microsoft Copilot for a shortlist, and how to fix it within MOH advertising rules. URL: https://tryshubi.com/insights/singapore-specialist-clinic-copilot-shortlists/ Published: 2026-11-12 Updated: 2026-11-12 Author: Md Shariar Shanaz Shuvon Market: Singapore Industry: medical Topics: AI visibility, AEO, medical, Singapore, Microsoft Copilot ### Short answer Microsoft Copilot builds specialist shortlists by sending short Bing queries and reading the pages Bing returns. Singapore specialist clinics drop out when their doctors' registered specialties, panel status and locations are vague or buried in PDFs. Publish plain, factual, verifiable pages that follow the Healthcare Services (Advertisement) Regulations, keep Bing listings current and track citations in Bing Webmaster Tools. ### Key points - Copilot grounds web answers in Bing, so a clinic that is thin or outdated in Bing's index is easy to leave off a shortlist. - Copilot sends Bing a few-word query built from the prompt, so pages must answer specialty, location and panel questions in plain text. - Use the specialty names recognised by the Specialists Accreditation Board so AI can match your doctors to the right referral. - Singapore's advertisement regulations bar superlatives, quality comparisons and reproduced testimonials, so AI-facing copy must stay factual. - Bing Webmaster Tools' AI Performance report shows how often your pages are cited in Copilot answers, which gives you a baseline. A benefits manager in a CBD office opens Copilot in Teams and types: "Shortlist orthopaedic specialist clinics near Raffles Place for our staff panel." An employee with a knee injury asks much the same thing on her phone. Your specialist clinic group has the surgeons, the panel contracts and the branches, yet the Copilot shortlist names three other clinics and an aggregator page. For a large Singapore specialist clinic network, that is a lost referral you never see. This article explains why Microsoft Copilot leaves established specialist clinics off its shortlists, what Singapore employers and patients actually ask it, and how to audit and fix your position without breaking the Healthcare Services (Advertisement) Regulations. It is written for marketing and business development leads at clinic groups and hospital-affiliated specialist centres that depend on corporate and insurer panels. ### Why Copilot leaves your specialist clinic off the shortlist Copilot builds web answers from Bing results, so it can only shortlist clinics that Bing returns for a short, specific query. If your specialty, doctors, locations and panel facts are vague, out of date or locked in PDFs, another page answers the query instead. Microsoft's documentation is clear on the mechanics. When web search is enabled, Copilot and Copilot Chat parse the prompt, identify terms where web information would help, and send the Bing search service a generated query that "consists of a few words informed by the user's prompt". The Bing results are then used to compose the reply. The long, detailed question your buyer typed becomes something like "orthopaedic clinic Raffles Place", and Copilot works from whatever ranks for that. Three things follow for specialist clinics in Singapore: - **Directories often win the query.** Insurer and third-party administrator panel directories, hospital doctor finders and listing sites tend to be text-heavy and well structured. If your own pages are thin, Copilot may cite them rather than you, and they may hold an old roster. - **The prompt can mix company data with the web.** Microsoft's examples show Copilot reasoning over an internal document the user references, then searching the web on themes from it. An HR team that uploads its panel list and asks Copilot to compare clinics will get web facts about each name on that list. If your facts are missing, you look weaker than you are. - **Some tenants have web search off.** Microsoft lets IT admins enable or disable web search across a tenant, and users can switch off a Web content toggle. Answers therefore vary between companies, between users and between runs. Treat any single result as a sample, not a verdict. Microsoft's Bing team also says, in its AI Performance announcement, that clear headings, tables, FAQ sections, regular updates and current Bing Places for Business listings help AI systems reference information accurately. That is the brief for your pages. ### What buyers in Singapore actually ask AI Singapore buyers ask Copilot in English, with local shorthand: "ortho", "ENT", "panel clinic", "Novena", "IP" for Integrated Shield Plan, and the names of their insurer or TPA. Some patients and family members will ask in Mandarin. **HR and benefits teams setting up or reviewing a panel** - "List specialist clinic groups in Singapore with orthopaedic, ENT and gastroenterology under one roof, near Tanjong Pagar." - "Compare these five clinics from our panel file: specialties, locations, weekend hours." - "Which specialist clinics in the CBD accept direct billing for corporate insurance?" **Employees and dependants looking for care** - "Sports medicine doctor near Jurong East for an ACL injury, on my company panel." - "ENT specialist Novena for my child's recurring ear infections, can I see one without a GP referral?" - "新加坡 骨科专科 诊所 Orchard 附近" (orthopaedic specialist clinic near Orchard). **Referring GPs and case managers** - "Which private gastroenterologists in Singapore do colonoscopy at a day surgery centre in the east?" - "Find a rheumatologist with weekday evening clinics in Punggol." Notice what these prompts need: a recognised specialty, a neighbourhood, a panel or billing answer, hours and a procedure. None of them asks who is "best", which matters for the rules below. ### How to check where you stand You can check your Copilot visibility in a week with a fixed prompt list, a clean test account and a simple sheet. The aim is a baseline you can repeat, not a one-off screenshot. Our [how we measure](https://tryshubi.com/how-we-measure/) page explains the wider method. #### Step 1: Build a prompt set from real demand Write 20 to 30 prompts across the three groups above, one row per specialty and area you serve. Use the language your call centre and corporate sales team hear, including Mandarin where relevant. Keep the list fixed so month-on-month results compare. #### Step 2: Run the prompts in Copilot Chat with web search on Use a work account where web search is allowed, and record the date and whether the Web content toggle was on. Microsoft notes that Copilot Chat shows the exact web search queries it sent to Bing in the citation section, kept in the thread for 24 hours. Copy those queries: they tell you what you need to rank for. #### Step 3: Record the shortlist and the citations For each prompt, log whether your clinic appears, its position, which of your locations is named, and every cited URL. Note when the citation is a panel directory, a hospital doctor finder or a rival. #### Step 4: Check accuracy line by line Compare what Copilot says against your current roster, hours, branches and panels. Common errors include doctors who left a year ago, a closed branch, or a specialty attached to the wrong doctor. Check each doctor against the Singapore Medical Council register, which lists specialists registered after Specialists Accreditation Board accreditation. #### Step 5: Look at Bing's own data Bing Webmaster Tools' AI Performance report, in public preview since February 2026, shows total citations, average cited pages and page-level citation counts across Microsoft Copilot, AI summaries in Bing and some partner integrations. It also shows a sample of grounding queries. Microsoft is clear that these figures do not show placement or why a page was chosen, so use them for direction. ### How to fix it Fixing Copilot visibility for a specialist clinic means making your facts easy for Bing to find and safe for you to publish. Every change below must pass the Healthcare Services (Advertisement) Regulations 2021, which cover licensees' advertising on the internet as well as in print. #### Use recognised specialty names, then add the words patients use The Specialists Accreditation Board recognises 36 specialties and 10 subspecialties, including Orthopaedic Surgery, Otorhinolaryngology/ENT, Gastroenterology and the subspecialty of Sports Medicine. Use these exact titles on doctor and service pages, then add patient wording in the body ("ENT doctor", "slipped disc", "knee scope"). This lets Copilot connect a casual prompt to a registered specialty. Never imply specialist status a doctor does not hold on the register. #### Give every doctor a text profile Portrait-only team grids give Bing nothing to read. Each specialist needs their own page with name, registered specialty, conditions and procedures they treat, languages spoken, clinic locations, session days and how to book. Add Physician and MedicalClinic structured data that matches the visible text exactly. #### Publish panel and billing facts in plain text Corporate prompts ask "is this clinic on our panel?" and "do they do direct billing?". Publish a plain HTML page listing the insurers and TPAs you work with, what patients need to bring, and whether a GP referral letter is required. Keep it dated and current. Then ask each insurer and TPA to update their directory entries, because Copilot may cite those instead. #### Keep branch data identical everywhere Name, address, unit number, phone and hours should match on your site, Bing Places for Business, Google Business Profile, hospital directories and insurer listings. A Novena suite on one listing and an old Orchard address on another is exactly the kind of mismatch Copilot repeats. Use IndexNow, which Bing recommends, to notify it when pages change. #### Write AI-facing copy that stays inside the rules Regulation 5 requires advertising to be factually accurate, capable of being substantiated, and free of superlatives, laudatory statements, quality comparisons with other licensees and before-and-after images. That rules out "Singapore's leading sports clinic" and "faster recovery than other clinics". Regulation 14 restricts reviews and testimonials to ones given directly on your own site or social account and not reproduced. Regulation 13 limits how you show accreditations and awards, and regulation 16 stops you linking from your site to pages that break these rules. | Content | Why Copilot can use it | What is allowed in Singapore | | --- | --- | --- | | Doctor's registered specialty and qualifications | Matches the specialty in the query | State factually, matching the SMC register | | Conditions and procedures offered | Matches symptom and procedure prompts | Describe scope; no promised outcomes | | Locations, hours, languages | Matches area and access prompts | Fine if accurate and current | | Panel and billing details | Matches HR and insured employee prompts | Fine if accurate; date the page | | "Best", "leading", "top" claims | Rarely helps grounding | Not allowed (laudatory or superlative) | | Reproduced patient reviews | Often ignored or misread | Restricted under regulation 14 | #### Strengthen the third-party sources Copilot already cites List every external page that appeared in your audit. Ask hospital doctor finders, insurer panels and professional society directories to correct rosters and specialties. Accurate, consistent third-party pages make it more likely that Copilot describes you correctly, even when it does not cite your own site. ### What to measure Track four numbers monthly on the fixed prompt set: mention rate, citation share, accuracy and placement. Expect movement over months rather than weeks, and expect noise between runs. - **Mention rate:** share of prompts where your clinic appears in the shortlist. - **Citation share:** share of cited URLs that are your own pages, against directories and rivals. Bing's AI Performance report adds a view of how often your pages are cited. - **Accuracy:** share of answers with the right doctors, specialties, branches, hours and panels. - **Location coverage:** whether each branch appears for its own neighbourhood prompts. Re-run the set after a roster change, a new branch, or a new panel contract. Report ranges and trends to leadership, not a single figure. ### Common mistakes - Writing "best" or "leading" into service pages to attract AI, which breaches regulation 5 and does not help grounding. - Leaving specialist rosters in brochure PDFs while the HTML page shows only photos. - Using internal service-line names ("Joint Preservation Centre") without the recognised specialty title that buyers and AI match on. - Updating your own site but not the insurer and TPA directories Copilot often reads first. - Checking only consumer ChatGPT and ignoring Copilot inside the Microsoft 365 tools your corporate clients use daily. - Judging success from one screenshot, when Microsoft's own controls mean answers differ by company and user. ### How Shubi helps Shubi runs the prompt audits and Bing citation tracking for specialist clinic groups, then works with your compliance team on factual, regulation-safe pages and directory corrections. We cannot guarantee that Copilot will shortlist any clinic. Our job is to make sure your clinic is described accurately when it does. See [AI Visibility Intelligence](https://tryshubi.com/services/#intelligence), [content](https://tryshubi.com/services/#content) and [technical](https://tryshubi.com/services/#technical) work, or [Message Shubi on WhatsApp](https://wa.me/601137634679). ### Frequently asked questions #### Does Microsoft Copilot use our website when it recommends specialists? It can. Microsoft's documentation says that when web search is enabled, Copilot sends a short generated query to the Bing search service and uses what comes back to compose the answer. If Bing returns a directory or a rival's page instead of yours, that is what Copilot reads. #### Can we tell AI tools we are the best orthopaedic or ENT clinic in Singapore? No. The Healthcare Services (Advertisement) Regulations 2021 say advertisements must be factually accurate and capable of being substantiated, and must not contain laudatory statements, superlatives or quality comparisons with other licensees. Describe scope, specialist registration, locations and panels instead. #### Can we put patient reviews on our specialist pages so AI picks them up? Only within narrow limits. Regulation 14 allows a licensee to show a review or testimonial on its own website or social media account only if the person gave it directly to the licensee and it is not reproduced. Check with your compliance team before publishing any. #### What if a client's IT team has switched off web search in Copilot? Then Copilot answers from the organisation's own data and cannot read your site. Microsoft lets admins turn web search on or off for the tenant, so employees at some firms will see web-grounded answers and others will not. This is one reason results vary between companies. #### How often should we check Copilot answers? Monthly is a sensible rhythm for a fixed prompt set, with a re-check after any roster, panel or branch change. Answers vary between runs and users, so look at trends across many prompts rather than a single screenshot. ### Sources - [Data, privacy, and security for web search in Microsoft Copilot and Microsoft Copilot Chat (Microsoft Learn)](https://learn.microsoft.com/en-us/copilot/microsoft-365/manage-public-web-access) - [Introducing AI Performance in Bing Webmaster Tools Public Preview (Bing Webmaster Blog)](https://blogs.bing.com/webmaster/2026/2/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview/) - [Healthcare Services (Advertisement) Regulations 2021 (Singapore Statutes Online)](https://sso.agc.gov.sg/SL/HSA2020-S1033-2021) - [List of recognised specialties (Specialists Accreditation Board)](https://sab.healthprofessionals.gov.sg/specialist-training/list-of-recognised-specialties/) - [Specialist registration (Singapore Medical Council)](https://www.smc.gov.sg/becoming-a-registered-doctor/register-of-medical-practitioners/specialist-registration) ## Contact - WhatsApp: +60 11-3763 4679 (https://wa.me/601137634679) - Contact page: https://tryshubi.com/contact/ - Free AI visibility snapshot: https://tryshubi.com/snapshot/