AI visibility checker
Paste a domain. It drafts the questions your buyers would ask before they know you exist, you edit them, then each one goes to an AI assistant with live web search. You get back how often you are cited, who is cited instead, and which of your pages are doing the work.
Your questions
These are the questions a buyer might type before they know you exist. Change them. You know what your customers ask better than any model does. Delete the ones that miss, rewrite the rest, add your own.
Competitors (optional)
Running
What this measures, and what it cannot
Two different things get counted for every question, and confusing them is how most reporting on this goes wrong.
- Retrieved means your domain turned up in the search results the assistant saw while writing the answer.
- Cited means the answer actually leaned on it. This is the one that matters, because a reader only ever sees what was cited.
A site can be retrieved on every question and cited on none. When that happens, the problem is not crawling, indexing or robots.txt. The pages are being found and passed over, which is a content problem: the page does not answer the question directly enough to be worth quoting. That distinction decides what you do next, so the report keeps the two apart instead of averaging them into a score.
Six questions tells you whether you are in the game. It does not tell you whether you improved. With a handful of yes-or-no results, a site whose real citation rate is 30 percent can show up anywhere from 1 in 6 to 4 in 6 on any given run, and AI answers are not deterministic, so the same question can pull different sources an hour later. A zero is a clear signal. A move from 2 to 3 is noise. Tracking a trend needs a bigger question set run repeatedly, which is the paid version of this and not something a free page should pretend to do.
What it cannot do: this is one model with one web search tool. It is not ChatGPT, not Google AI Overviews, not Perplexity, and it will not agree with them, or with itself next month. Those systems retrieve differently and change without notice. What this gives you is a consistent measurement, taken the same way every time, so two runs are comparable to each other. Anything sold to you as "your rank in AI" is overstating what can be known.
Why the questions matter more than the score
The questions are generated from your own site, and never contain your brand name. Asking an assistant "who is Acme" tells you nothing, because Acme's own site is obviously cited. The useful question is whether you appear when somebody describes their problem and has never heard of you. That is where the category is actually won.
What to do with a bad result
Run the AEO audit on the pages that should have been cited and were not. The checker tells you that a page is losing. The audit agent reads that page and tells you why: no direct answer under the heading, schema that does not match the content, no author or dates, crawlers blocked. Measurement then diagnosis, in that order.
If you want this tracked over time across a real question set, with the fixes made and measured, that is the AEO work I do. Start there.
If you want the questions tracked over time, with the fixes made and the movement measured, that is the paid version and it is the work I do. This page is the free sample, and it is deliberately the size of a sample.