How to check whether ChatGPT can actually see your business

People increasingly ask an assistant rather than a search engine. “Find me a dentist in Vaughan” now returns a short list of named businesses instead of ten blue links. If your business is not one of the names, you are invisible in that conversation. Here is how to check, using nothing but a browser.

Check one: can the crawlers reach you at all

Open yourdomain.com/robots.txt. You are looking for whether the AI crawlers are permitted. The ones that matter most:

• GPTBot and OAI-SearchBot, OpenAI. The second is the one used for live ChatGPT search retrieval.
• ClaudeBot, Claude-User and Claude-SearchBot, Anthropic.
• PerplexityBot, Perplexity, which retrieves close to real time.
• Google-Extended, which governs Gemini rather than ordinary Google Search.

A blanket Disallow: / is an obvious problem. The subtler one is silence: if none of these agents are named at all, you are relying on each operator’s default behaviour, and those defaults differ and change. Naming them explicitly removes the ambiguity.

Check two: are you a machine-readable business

Run your homepage through Google’s Rich Results Test or the Schema Markup Validator. You are looking for an Organization or LocalBusiness block containing your name, address, phone and hours.

Without it, an assistant reading your site sees prose it has to interpret. With it, it sees a declared business entity. That difference is most of the reason some local businesses get named in AI answers and near-identical competitors do not.

If you have an FAQ on the page, check it is marked up as FAQPage. Question and answer pairs are the format answer engines find easiest to lift.

Check three: ask the assistants directly

Open ChatGPT, Perplexity and Google’s AI mode, and ask the question a customer would actually ask. Not your business name, which proves nothing, but the buying question:

• “Who are the best [your service] providers in [your city]?”
• “I need [your service] near [landmark or intersection]. Who should I contact?”
• “How much does [your service] cost in [your city]?”

Record whether you appear, and which competitors do. The competitor list is as useful as your own position, because it tells you which sources the assistant trusts for your category.

Do this from a logged-out session. Assistants personalise on conversation history, and your own account has been talking about your business.

Check four: does your site answer questions directly

Answer engines lift passages. A paragraph that answers a question in its first sentence can be quoted on its own. A paragraph that spends four sentences building to the point cannot, because the useful part does not survive extraction.

Read your service pages and ask of each paragraph: if this were lifted out and shown alone, would it still make sense and still be useful? That single test changes how a page gets written.

Check five: is your information consistent everywhere

Assistants cross-reference. If your website says one set of hours, your Google profile says another and a directory lists an old phone number, the model has conflicting information about you and tends to resolve that by mentioning something else instead.

What this exercise will not tell you

It is a sample, not a measurement. Asking ten questions tells you whether you surface for those ten. It does not give you a citation share, it will not be identical tomorrow, and results vary by phrasing and location.

Treat it as a directional check you can repeat monthly with the same questions. Consistency of method matters more than the absolute number, because what you are watching for is movement.

Frequently Asked Questions

Can ChatGPT actually see my website?

Only if your robots.txt permits the relevant crawler and your content is readable without JavaScript. Different assistants use different agents, and allowing one does not allow the others. A site can be perfectly visible to Google and invisible to an assistant.

Why does an AI assistant give wrong information about my business?

Usually because it is assembling an answer from stale directory listings, inconsistent address details, or a competitor page that mentions you. If your own site does not state something plainly, the model will take it from wherever it can.

Does blocking AI crawlers hurt my business?

It depends what you want. Blocking them protects your content from being used in training, and it also removes you from the answers those assistants give. For most local service businesses, being findable is worth more than being withheld.

What is llms.txt and does it matter yet?

It is a plain text file at your site root that summarises your content for language models. Adoption is not universal and no major assistant guarantees it uses one. It costs very little to publish, which is why we do, but it is not the thing that decides whether you are cited.

What actually makes an assistant cite my business?

Specific, checkable statements it can lift and attribute. Prices, locations, hours, service definitions and direct answers to real questions. Vague marketing copy is never quoted because there is nothing in it to quote.