What does Claude say about your business, and how do you check it?

When a potential client asks an AI assistant to recommend an accounting firm in Mont Kiara, or a family dentist in Subang Jaya, the answer they receive may or may not include your business. Understanding what Claude says about your business — and building a repeatable check — is the practical starting point for any Malaysian SME owner who wants to improve their visibility with AI platforms.

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, property agencies. Disclosure: it comes from Geo One, a Generative Engine Optimisation (GEO) and Answer Engine Optimisation (AEO) agency based in Kuala Lumpur.

Quick answer: ask five buyer-style questions in fresh Claude chats, record which businesses are named, in what order and with which sources, check whether what it says about you is accurate, and repeat the same questions on a set schedule. Do not test only your own brand name.

How do you check what Claude says about your business?

Ask the question your buyer would ask, not your company name. Searching your own brand tells you almost nothing, because the question a customer types is "best accounting firm in Mont Kiara for a small Sdn Bhd", not "is Geo One any good".

A repeatable check takes about ten minutes:

1. Write down five questions a real buyer would type. Build each one around service, suburb and buyer type. At least half should include a constraint — the suburb, the company size, the specific problem — because those are the queries a small firm can realistically appear in.
2. Open a fresh conversation. Start a new chat for each question, to reduce the chance that past conversations or memory settings colour the answer.
3. Ask one question per chat. For each response, record which businesses get named, in what order, and which websites Claude cites as its sources, when it shows any. Claude does not always list sources.
4. Note what it gets wrong. If Claude names your business but gives an incorrect address, outdated service description or wrong phone number, that is a separate problem from not being named at all — and it needs a different fix.
5. Note the date. Answers shift over time, so a screenshot with no date is not a measurement.
6. Repeat on a schedule, same questions, same wording. Monthly is a sensible rhythm, or weekly while you are actively making changes.

Five example questions to adapt to your own business:

• "recommended family dentist in Subang Jaya that takes walk-ins"
• "accounting firm in Mont Kiara for a small Sdn Bhd"
• "veterinary clinic near Bangsar for a senior dog"
• "tuition centre in Petaling Jaya for IGCSE maths"
• "property agent in KLCC for first-time landlords"

To make the question-writing concrete: a dental clinic in Subang Jaya is more likely to appear when someone asks "affordable family dentist in Subang Jaya that takes walk-ins" than when they ask "best dental clinic in Selangor". The first query has a buyer in it — location, budget signal, intent. The second is a list question, and lists tend to be drawn from sources the engine already has on file, directories, roundups and publications that have been cited many times. A small practice may find it difficult to displace a roundup that already has significant authority behind it; it can, however, be specific enough that the roundup is not what the engine reaches for.

When you review Claude's responses, you are looking for four things: whether your business is named at all, what it says about you (and whether that is accurate), which competitors or directories it names instead, and which external sites it cites. Those four data points together tell you where the gap is.

For example, when we asked "best generative engine optimisation agency in Kuala Lumpur" on 6 October 2026, Perplexity cited cleverus.com, hashmeta.com, pctechmag.com, sortlist.com and mackyclyde.com: agency sites, a tech publication and a B2B directory. ChatGPT cited oblique.com.my and cleverus.com among others. Claude named businesses without listing any sources, which is why the check records sources wherever an engine shows them. We kept these answers in our scan log for that day. Results can differ by account, location, settings and date, which is why the check records the date.

If Claude names you but describes you incorrectly — wrong location, wrong services, outdated credentials — we have written separately on what to do when an AI assistant describes your business wrongly.

What your results are likely to show

A realistic result from this check is not that your business appears in every answer. On 6 October 2026 we ran Geo One's own check: 22 questions, each asked on ChatGPT, Claude, DeepSeek, Gemini and Perplexity, which makes 110 checks. DeepSeek returned no answers that day, so we recorded 88 answers. Geo One was named in 9 of them: 5 on Claude, 4 on ChatGPT, none on Gemini or Perplexity. This is our own first-party data, not a published third-party study. More than half of those mentions came from questions that named a buyer, such as "recommended AEO agency for small business in Kuala Lumpur"; no engine named us for the plain "best generative engine optimisation agency in Kuala Lumpur".

The difference between those two phrasings matters more than the hit rate. The first names a service, a buyer size and a city. The second is a superlative with no buyer in it, and superlative queries tend to be answered from whatever list-style pages the engine already trusts. So when you build your five questions, write them the way an anxious buyer actually types.

If a provider shows you a near-perfect hit rate across five engines, ask to see the question list and the dates before drawing conclusions.

Can Claude actually read your website?

Anthropic runs separate crawlers for different jobs. Its own documentation says that blocking Claude-User stops Claude from fetching your pages when someone asks it a question, and blocking Claude-SearchBot stops your content being indexed for its search results; both can reduce how visible you are in Claude's answers (Anthropic: how site owners can block the crawler). So it is worth confirming your site lets these through before drawing conclusions from your check. We have covered how to confirm this in Can Claude read your website at all? The first check for Malaysian SMEs.

What to do when Claude gets your business wrong

If you run the ten-minute check and Claude names your business but describes it incorrectly, the priority is to improve the signals AI platforms use to understand and verify your business. That means ensuring your business name, address, services and credentials appear consistently across authoritative third-party sources — relevant directories, industry publications, professional body listings and other sources appropriate to your category. If the information on those sources is accurate and consistent, AI platforms are better positioned to surface accurate information about you.

If Claude does not name your business at all for any of your five questions, the likely cause is that there is insufficient information for the platform to retrieve: pages it cannot fetch, facts it cannot confirm, or no mention of your firm anywhere outside your own website. That is slower work to address, and it involves building legitimate presence across the sources AI platforms use to understand businesses in your category.

Geo One runs this check across five AI engines — ChatGPT, Claude, DeepSeek, Gemini and Perplexity — every week or every month, depending on the plan, for the businesses we work with. If you would like to see what those engines currently say about your business, that is where we start: run a free AI visibility scan.