CiteSiteFree tool
AI visibility checker
Five things have to go right, in order, before an AI engine can name your site in an answer. This reads your pages and finds the first one that does not.
Five things have to go right in order. This finds the first one that does not — deliberately not a score, because averaging a chain hides the break.
The five steps
- Can I fetch these pages at all? robots.txt decides this, and it is absolute.
- Is the text here, or does a browser have to build it? A page whose body arrives empty is an empty page as far as most crawlers are concerned.
- What is this site, and is it the same thing I have seen elsewhere? Whether anything declares the entity, says so consistently, and links it to itself elsewhere.
- Is there a passage here I could lift? Whether any paragraph still means something quoted alone.
- Can I say who wrote this, and when? Author and date, declared rather than rendered.
Why the order matters more than the verdicts
Each step is a precondition for the one after it. A site with immaculate markup behind a Disallow: / is not partly visible — it is absent, and improving its markup changes nothing.
That is the whole reason this is not a score. Average five steps where the first gates the rest and you get a number that says “60%” about a site nothing can read, which a reader will act on by working on step four. So the output is one sentence instead: here is where it stops. Everything below the break is shown, and dimmed, because a reader needs to see that there are four more things and that none of them matters yet.
What it cannot tell you
Whether ChatGPT, Gemini or Perplexity actually mention you. That answer costs a paid API call per question, per engine, and a page like this one — public, unauthenticated, linked from search results — would be spending it for strangers at a rate set by whoever links here. Every tool on this site is built on our own crawling for that reason, and the constraint is stated rather than hidden.
It is worth being suspicious of a free tool that claims otherwise. The honest question to ask is what it actually queried.
When all five steps are clear, the infrastructure is not what is holding you back, and the remaining questions are about substance and competition — whether your pages answer what people ask, and whether someone else answers it better. Those need the engines themselves, which is what a full analysis does, recording for every answer whether the model searched the web or answered from memory.
If you want the score instead
The GEO audit reads the same crawl and presents it as a weighted score across six pillars, with a ranked fix list. Same data, different job: that one is for handing to someone else, this one is for working out what to do next.
Questions
- Does this tell me whether ChatGPT mentions my brand?
- No, and no free tool honestly can — that answer costs a paid API call per question, per engine. This checks the five things that have to be true before a mention is possible, and finds the first one that is not. If a tool claims to tell you what ChatGPT says about you at no cost, ask what it actually queried.
- Why a chain rather than a score?
- Because the five steps are not independent — each is a precondition for the next. Averaging them produces a comfortable-looking number for a site that cannot be read at all, and a reader acts on it by improving step four. What they need is one sentence: here is where it stops.
- How is this different from the GEO audit?
- Same crawl, same signals, different reading. The audit is a score across six pillars with a ranked fix list — the thing you hand to a client. This is a diagnosis: five preconditions in order, with the first break highlighted and everything after it dimmed, because none of it matters yet.
- Step two says my pages are nearly empty and they are not.
- Then the body is being assembled in the browser. Everything here is read from the HTML as it arrives, which is how most crawlers read it — so what this saw is roughly what they see. That is the finding, and it outranks everything below it.
- Everything passes. Now what?
- Then the infrastructure is not what is holding you back, and the remaining questions are about substance and competition: whether your pages answer what people ask, and whether someone answers it better. Those are measured by asking the engines rather than by reading your site.
Related
AI visibility: how to measure it — what can honestly be measured once the chain is clear, and how to read the numbers without over-reading them.
A clear chain means a citation is possible. CiteSite measures whether one happens — across ChatGPT, Gemini and Perplexity, on a schedule, with the provenance of every answer recorded.