Why doesn't ChatGPT mention my business in Malaysia?

Usually for one of two reasons, and they are not the same problem. Either there is almost nothing about your business outside your own website for the model to work from, or there is plenty of material and it disagrees with itself. ChatGPT and Perplexity build an answer from the sources they can reach. If those sources are thin, you are not a candidate at all. If they contradict each other, a different trading name here, an old address there, two phone numbers and no way to tell which is current, then naming you is a risk, and naming a business whose details line up is safer.

Neither case is a penalty. Both are confidence problems. But the work that fixes one does almost nothing for the other, which is why "improve your website" is such unreliable advice here.

This is written for owners and practice managers of Klang Valley professional services firms with 2 to 50 staff: clinics, dental and veterinary practices, law and accounting firms, tuition centres and property agencies. Disclosure: it comes from Geo One, a Generative Engine Optimisation (GEO) and Answer Engine Optimisation (AEO) agency based in Kuala Lumpur.

Is this really happening, or does it just feel that way?

It's measurable. According to SOCi's 2025 Local Visibility Index, AI platforms recommend only 1.2% of business locations on ChatGPT and 7.4% on Perplexity, against 35.9% of locations appearing in Google's local 3-pack. On that reading, AI search is roughly 30 times more selective than ordinary Google search. A clinic that sits comfortably on page one can still be absent from every AI answer a patient reads.

The same index found business profile accuracy on ChatGPT and Perplexity sitting at 68%, with outdated details often repeated, against 100% on Gemini, which is grounded directly in Google Maps. That gap is the most useful diagnostic you have, and it is the basis of the test below. We've written separately on how Gemini and Perplexity do two different jobs and why you shouldn't read one as a proxy for the other.

Which of the two problems do you have?

Twenty minutes, three engines, two questions each. Ask ChatGPT, Perplexity and Gemini by name, "What do you know about [your firm], Kuala Lumpur?", and then the way a buyer would: "Recommended conveyancing lawyer in Bangsar." Read what comes back against these four patterns.

Gemini describes you accurately, ChatGPT says it has no information. Your Google Business Profile is doing its job, and nothing else is. ChatGPT doesn't read Google's data, so it has nothing to read. This is an absence problem, and no amount of further polishing on your own site or profile will move it.

Both engines name you, but the details are wrong or out of date. A contradiction problem. The superseded facts still exist somewhere a crawler can reach, and they are being repeated in preference to the current ones.

No engine recognises the name at all. Absence at its most basic. Either the crawlers can't fetch your site, or nothing third-party documents that you exist.

The buyer's question returns the same three competitors every time and never you. A category problem. You may be perfectly well documented and simply absent from the pages the model consults for that category in this city, association registers, practice directories, the roundup listicles that rank for "best X in PJ".

Our view, for firms of this size in the Klang Valley: absence is the more common diagnosis, not contradiction. Most 2-to-50-staff professional firms here built their online presence in an era when a good website and a well-kept Google Business Profile were genuinely sufficient, because Google read both. The directories were never needed, so they were never filled in. That choice was correct for fifteen years and is now the single largest reason ChatGPT has nothing to say about you.

Where do the inconsistencies actually come from?

If you do have a contradiction problem, it is rarely dramatic. These accumulate. The ones we see most in Klang Valley firms:

• The registered name and the trading name. The Sdn Bhd name on your invoices, a shorter trading name on your signage, and a third version on an old directory listing.
• A move you made years ago. The practice relocated from Damansara to Mont Kiara, the website was updated, three listing sites were not.
• Two phone numbers. A landline on the website, a WhatsApp number on Google, and neither one marked as the main line.
• Hours that changed and stayed changed. Raya or Chinese New Year hours posted once and never reverted.
• Services described differently everywhere. "Corporate advisory" on the website, "company secretarial" on a directory, "SSM compliance" in a brochure. A human reads all three as the same firm. A model may not.
• Practitioner names and credentials that appear on the website but nowhere a third party can verify them.

None of this breaks a Google ranking. All of it makes an AI engine less sure about you.

Why does ChatGPT read different sources from Google?

Because it doesn't have the same inputs. Research by Yext analysing AI citations found that nearly half of ChatGPT's local citations come from third-party sites such as Yelp, TripAdvisor and MapQuest, and that ChatGPT does not have direct access to Google Business Profile or Google Reviews.

That has a plain consequence for a Malaysian SME. If your entire online presence is a website plus a well-kept Google Business Profile, ChatGPT may have very little to work from. The tidy profile you've maintained for years isn't being read by that engine. What it reads instead is whatever third parties say about you, which for many firms is thin, old, or both.

Two other things sit underneath this. Whether OpenAI's crawlers are allowed to fetch your pages at all is set in your robots file and documented in the OpenAI Platform Docs. And whether your pages state your facts in a machine-readable form depends on structured data, the vocabulary for which is published at Schema.org. Both are checkable in an afternoon.

How do you audit your own facts?

Set aside an hour. You don't need a tool for the first pass.

1. Write the facts down once. Legal name, trading name, full address, main phone, WhatsApp number, email, hours, the list of services in the words you want used, and the practitioners with their credentials. This is your reference copy. Everything else gets checked against it.
2. Check your own site first. Footer, contact page, each service page. Sites drift internally more often than owners expect.
3. Check your Google Business Profile against the same list. Our six-check Google Business Profile audit walks through what matters for AI visibility specifically.
4. Search your business name plus Kuala Lumpur and open the first ten third-party results. Industry directories, association listings, old aggregator pages. Note every detail that disagrees, and note who owns each listing so you know whether you can edit it. If step 4 produces almost no results, that is not a clean bill of health, that is your absence problem, confirmed.

What should you do about it?

If the diagnosis is contradiction, fix the highest-authority listings first. Your own website and Google Business Profile take priority because they are under your control and because Gemini reads Google Maps directly. Third-party directories follow in order of how often they appear in AI citations, general business directories before niche ones. Use exactly the same wording you wrote in your reference copy. Partial corrections, fixing the address but leaving a service name inconsistent, reduce the problem without solving it.

If the diagnosis is absence, correcting things is beside the point; you need corroboration to exist in the first place. The most valuable entries are the ones a model can treat as independent verification: your professional body's public member register, the general business directories, and any local association or chamber listing. These outrank a tenth social profile, because they confirm rather than repeat.

Structured data on your own pages does a specific job in both cases: it lets crawlers read your name, address, phone, hours and services as labelled fields rather than inferring them from prose. A LocalBusiness schema block on your contact page is the minimum. A MedicalClinic, LegalService or AccountingService type, where one applies, is more precise.

The question of which AI engines can reach your site at all, and how to check, is covered in Can Claude read your website at all? The first check for Malaysian SMEs.

---

Geo One is a Generative Engine Optimisation (GEO) and Answer Engine Optimisation (AEO) agency based in Kuala Lumpur, serving professional services firms and solo practitioners across the Klang Valley. Its work focuses on AI citation visibility: making Malaysian businesses findable, accurately described and citable by ChatGPT, Perplexity, Gemini and Claude.