How to check if ChatGPT knows your business (and what to do if it doesn't)

Test each assistant with the same direct questions

Ask ChatGPT, Claude, Perplexity and Gemini the same set of questions: "What does [business name] do?", "Where is [business name] located?", "Who founded or runs [business name]?", "What do people say about [business name]?". Use your exact business name plus your industry and city if the name is generic. Run each question in a fresh chat with no prior context, so the model answers from what it already knows rather than what you just told it. Write down three things for each answer: what it got right, what it got wrong, and what it left out entirely. That gap map is what you fix.

Training data vs browsing changes what you're testing

Not every assistant answers the same way. Models like base ChatGPT or Claude without browsing enabled answer purely from training data, which has a cutoff date, so anything about your business that changed recently, such as a rebrand, a new location or a closure, may be missing or wrong. Perplexity and browsing-enabled ChatGPT or Gemini retrieve live web pages before answering, so their result reflects your current site and recent mentions, not a frozen snapshot. Test both modes where you can. A wrong answer from a training-data-only model tells you what the historical record says about you. A wrong answer from a retrieval-enabled one tells you your current web presence is failing right now.

Why a model might not know you, or get you wrong

Several things can produce it. If your business is genuinely new, it may not exist in older training data yet. If your web presence is thin, such as a single page with no reviews and no directory listings, there is not enough source material for a model to learn from. If your name, address, industry or founder details are stated inconsistently across your site, socials and directories, the model has no single confident version to repeat. If you have few or no mentions on sites a model treats as authoritative, such as industry publications, review platforms or local directories, there is nothing reinforcing who you are. Any one of these alone can produce a blank or wrong answer.

Blocked crawlers and the hallucination problem

Two more causes are worth checking separately. First, some sites block AI crawlers such as GPTBot, ClaudeBot, PerplexityBot and Google-Extended in robots.txt, deliberately or by accident, which stops retrieval-enabled assistants reading your site at query time even if the page itself is fine. Second, when a model has a gap, it does not always say it doesn't know. It can generate a plausible-sounding but invented answer, guessing your location, services or founding year from patterns in similar businesses. This is hallucination, not deception, and it is the most damaging outcome of all because it looks confident and specific while being wrong.

Fix the public signals models learn from and retrieve

You cannot edit what a model already knows, but you can fix what it finds next time it retrieves, and what future training runs will learn from. Make your business name, address, phone number, industry and key facts identical across your site, Google Business Profile, LinkedIn and major directories. Add structured data, specifically Organization and LocalBusiness schema, so the facts are machine-readable and not just visible text. Publish an llms.txt file that summarises who you are in plain language for AI systems to read. Check that robots.txt allows the major AI crawlers rather than blocking them by default. Earn genuine mentions on sites with real authority in your industry, since repetition across independent sources is what builds model confidence.

One scan checks the machine-readable side of this

Fixing everything above is worth doing properly, but it is hard to audit by eye which pieces are actually missing from your live site. AuditHQ's free scan runs in about 60 seconds, and its AI Visibility suite checks the machine-readable signals that AI assistants and their crawlers rely on: structured data, llms.txt presence, AI crawler access in robots.txt, and entity clarity. It will not tell you what ChatGPT says about you today, but it will tell you whether the underlying signals are actually there to be found, so you can work from a clear list of gaps instead of guesswork.

Frequently asked questions