AI visibility checker: what a real one actually measures
There is a shortcut going around in this category. A tool reads your homepage, counts some tags, and hands you an AI visibility score. It is quick, it feels rigorous, and it is measuring the wrong object.
The difference that decides whether the number means anything
Readiness is your ceiling. Outcome is where you sit under it. They move independently, and a checker that collapses them into one figure hides the only thing you needed to know.
We can show you exactly how far apart they get, because we measured it. Across 42 well-known sites, the average technical score was 81 out of 100 and the average AI visibility score was 52. Two rows carry the point: [Figure withdrawn 26 Aug 2026 pending re-measurement, why.]
- GoCardless scored 90 on technical checks and 11 on AI visibility. It was named in 0 of 6 buying questions. Stripe took 100% of those same answers.
- Canva scored 43, the worst technical score in the set, and 65 on AI visibility. Its page was set to noindex, its content only appeared after JavaScript ran, and it had no sitemap, no schema and no meta description. It was named in 6 of 6.
A page-only checker would have told GoCardless it was in excellent shape and told Canva it was broken. Both readings would have been useless, and one of them dangerously so.
What a real check runs
These are the checks worth running, in the order that matters. Access first, because nothing below it counts if an engine cannot get in.
- Crawler access. Request the homepage as GPTBot, ClaudeBot, PerplexityBot and ChatGPT-User. Anything other than a 200 and you are invisible to that engine no matter what the rest of the page says. This is the single heaviest check we run, and it is the one that caught our own site at zero.
- robots.txt. One managed security setting can write Disallow rules for every AI bot without anyone choosing it.
- Rendering. Many AI fetchers read raw HTML and do not run JavaScript. If your sentences are not in the source, they may not exist.
- Identity. Organization schema, a consistent name and description across your site and your listings, and an llms.txt. Engines cross-check facts, and facts they cannot verify get left out.
- Something quotable. A clean two-sentence answer to what you do, for whom. Engines quote statements, not atmosphere.
- The answers themselves. Generate the real questions a buyer asks in your category, put them to a live web-grounded engine, and count who gets named. This is the outcome half, and it is the only half a buyer experiences.
What no checker can tell you
Worth saying plainly, because the limitations are usually left off these pages.
A single run is one measurement, on one day, against one engine. AI answers vary between runs, between engines, and as models update. Roughly 85% of the sources engines cite are third-party pages rather than your own site, so a large part of your outcome score is decided by material you do not control and cannot edit today. And a checker cannot tell you your number will improve, only what is currently in the way.
Anyone selling you certainty in this category is selling you something that does not exist yet. What a check gives you is a baseline and a ranked list of what is broken. That is genuinely worth having, and it is all it is.
Our own numbers, published
The fastest way to judge any checker is to ask what its own score is. Ours is readiness 100, taken there from 0, with every measurement since published daily at our live score history, because a company selling AI visibility that hides its own number has answered the question.
The reason it is 0 is the same reason it is 0 for most young companies: off-site presence takes months, and no amount of work on our own pages substitutes for it. That is the honest mechanism, and it is why the readiness half moved to 100 while the outcome half has not moved at all.
After the number
A score you do not act on is trivia. The order that works is the same one the checks are listed in: fix access, then identity, then what is quotable, then spend the long money on getting mentioned somewhere engines already cite. Re-measure monthly, because a score from March means very little in July.
If you want the full method rather than the score alone, it is written up at how to improve your AI search visibility, and the manual version of these checks is at the 12-check list.
Questions, answered
Is there a free AI visibility checker?
Yes. Ours is free, needs no signup or card, and takes about a minute. It scores two halves separately: readiness, meaning whether AI can reach, read and verify your site, and outcome, meaning whether AI actually names you when a buyer asks a question in your category. Most free tools in this category only grade your page, which is the readiness half alone.
How do I check if ChatGPT recommends my business?
Ask it the questions your buyers ask, not your brand name. Searching your own name tells you almost nothing, because an engine will usually find you if it is handed your name. The useful test is a category question such as best accountants in Leeds or who should I use for payroll software, run several times, counting how often you are named and who is named instead. Repeat across ChatGPT, Claude, Perplexity and Gemini, because the overlap between them is smaller than people expect.
Why does my site score well on SEO but badly on AI visibility?
Because they measure different things. Technical SEO is largely about your own site, and AI visibility is largely about everyone else's. Roughly 85% of the sources AI engines cite are third-party pages. In our own study of 42 well-known sites, GoCardless scored 90 on technical checks and was named in 0 of 6 buying questions, while Canva had the worst technical score in the set and was named in 6 of 6.
How often should I re-check my AI visibility?
Monthly is enough for most businesses. AI answers move as models update and as competitors publish, so a single measurement ages faster than a traditional rank check. If you cannot name your score and the date you measured it, you are effectively unmeasured.
What is a good AI visibility score?
There is no established benchmark yet, and anyone quoting one firmly is guessing. For context, across the 42 well-known sites we measured, the average was 52 out of 100 and only 5 scored above 70. Treat your first number as a baseline to move rather than a grade to pass.
Run the check on your own site
It asks the real buyer questions in your category against a live engine, records which businesses get named, and scores both halves separately. Free, about a minute, no signup and no card.
Free score in under a minute. Live AI answers, not estimates. No signup, no card.