SEO and AI marketing specialist for fitness studiosDaniyal Sherafat

Guide

What an AI visibility checker actually checks

What an AI visibility checker does, how to judge whether one is honest, and how to check by hand in five minutes if you would rather not use a tool at all.

People have started asking assistants the questions they used to type into a search box. Not “pilates rotterdam” but “which pilates studio in Rotterdam would suit a complete beginner”. The answer comes back as a paragraph naming two or three places, and the businesses that were not named never find out they were in the running.

An AI visibility checker exists to answer that one question: when someone asks, are you named?

What a checker does

Three things, and the third is the one that separates a useful tool from a novelty.

It asks. It puts questions to one or more assistants, the way a customer would phrase them, and records the answer word for word. Both OpenAI and Anthropic offer web search through their APIs, and both return the sources used, which is what makes this measurable rather than anecdotal.

It reads. It fetches your public pages the way an assistant’s crawler would and notes what is there in plain text: what you do, where you are, what a first session costs, how to book. An assistant cannot cite what it cannot read, and a surprising amount of what a studio would want quoted lives inside an image, a booking widget or a PDF.

It repeats. This is the part that matters and the part most tools skip. Ask an assistant the same question twice and you can get two different answers. One answer is an anecdote. The useful output is a rate: named in four of twelve answers, cited in one.

How to judge whether a checker is honest

Four questions to put to any tool before you trust its number, this one included.

Does it say which assistants it asked, and how many times? “AI visibility: 62%” is not a measurement unless you know what was asked, of whom, and how often. A tool that will not show you the questions is showing you a score it invented.

Does it show the answers, or only a grade? You should be able to read what was actually said about you. That is the evidence. A grade without the transcript cannot be checked or argued with.

Does it separate being named from being cited? These are different and both matter. Being named means the assistant said your name. Being cited means it linked your website as a source. A business can be named from a directory listing without its own site being read at all, which is worth knowing, because it means the assistant is describing you using someone else’s words.

Does it say what it cannot see? No checker can see your Google Business Profile, your booking system, or what a specific person was shown yesterday. A tool that claims otherwise is guessing.

How to check by hand, in five minutes

You do not need a tool for a first look, and doing it once by hand teaches you more than any report.

  1. Open a fresh chat, signed out if you can, so your own history does not colour the answer.
  2. Ask the question a stranger would ask. Not your business name. Something like “which yoga studios in Utrecht are good for beginners?”
  3. Read who gets named, and in what order.
  4. Ask the same question twice more. The variation between the three is the real picture.
  5. Now ask directly: “what do you know about [your studio name] in [your city]?” If the answer is vague, wrong, or built from a directory rather than your own site, that is your finding.

If your own website is never among the sources, the assistant is not reading you. That is a different problem from not being recommended, and it is the one you can actually fix.

What this site’s checker does, and does not

The check here asks ChatGPT and Claude four fixed questions: two about the kind of studio you are, in your city, with no mention of your name, and two about your studio directly. The unbranded ones run three times each because a single answer is not evidence. That is sixteen answers.

Alongside that it reads your public pages and reports what the assistants could and could not find in plain text, and it runs Google’s own page speed test.

It does not measure Google rankings, traffic, or bookings. It cannot see your Google Business Profile. It does not promise that fixing anything will get you recommended, because nobody can promise that: the selection rules are not published, and the answers move.

The method page sets out the questions, the models, the number of runs and the limits in full. The crawler identifies itself and can be blocked.

A free AI visibility checker, or an AI visibility audit?

The two words get used for the same thing, but they are different jobs.

A free AI visibility checker is a snapshot. It asks a fixed set of questions once, on one day, and shows you what came back. That is enough to answer the first question, whether you are named at all, and to see which facts the assistants could not find on your site. It is not enough to tell a trend from noise, because the answers move between runs.

An AI visibility audit is the longer version, usually done by a person and usually paid. It asks more questions, in the words your customers use, repeats them over weeks, compares you with the businesses that do get named, looks at which sites the assistants cite, and ends in a list of fixes with an owner for each. It is worth paying for once the snapshot shows a gap and you want to close it properly.

Whichever you use, ask the same four things of it: which questions were asked, how many times, where the raw answers are, and on which date. A score without those four is not a measurement. The checker on this site is the free kind: a snapshot with the answers attached, not an audit.

What to do with the result

If you were not named, the first question is not “how do I rank” but “is there anything to cite”. Most of the time the fix is unglamorous: put the things a customer asks about into plain text on your own pages. What you offer, who it suits, where you are, what the first session costs, how to book. Not in a graphic. Not behind a widget. In words, on a page, in the language your customers use.

That is the whole of it. There is a fuller version of the argument in the guide to AI search visibility for studios, and if the answer to “why am I not there” is what you actually came for, there is a page on why ChatGPT may not recommend you.

Sources

  1. OpenAI, Web search tool: citations and user location (checked 2026-09-21)
  2. OpenAI, Overview of OpenAI crawlers (OAI-SearchBot, GPTBot, ChatGPT-User) (checked 2026-09-21)
  3. Anthropic, Web search tool (checked 2026-09-21)
  4. Google Search Central, AI features and your website (checked 2026-09-21)

FAQ

Questions studio owners ask

What is an AI visibility checker?

A tool that asks one or more AI assistants questions a customer would ask, and reports whether a given business was named in the answer and whether its website was cited as a source. The useful ones also read the site itself, because being named depends on there being something readable to name.

How is that different from a rank tracker?

A rank tracker records a fixed position in a list of results. An assistant does not produce a list of results; it produces a paragraph that varies between runs and between assistants. So the honest unit of measurement is not a position but a rate: out of this many answers, you were named in this many.

Can any checker guarantee I get recommended?

No, and a checker that implies otherwise is worth leaving. The selection rules are not published by any provider, and answers vary between runs of the same question. What a checker can do is show what was said, what your site made readable, and what is missing.

Do I need a tool, or can I just ask ChatGPT myself?

You can ask yourself, and you should at least once. The section above sets out how. A tool is worth it for two things a person doing it by hand tends not to do: asking the same question several times, because a single answer is a coincidence rather than a measurement, and recording the result so next month's answer can be compared with this one.

What does the check on this site ask?

Four fixed questions, put to ChatGPT and Claude. The two unbranded ones are asked three times each and the two branded ones once, which is sixteen answers in total. The method page sets out the questions, the models and the limits, including what the check cannot see.

Is a free checker reading my site safe?

It depends on the checker, and it is fair to ask before using one. A reasonable checker reads only pages that are already public, identifies its crawler by name so you can block it, and tells you what it stores and for how long. This site publishes all three.