CiteSiteFree tool
Free GEO audit
Reads your site and scores it across six pillars — crawlability, entity clarity, citation readiness, content structure, structured data and authority — then ranks what to fix. Every subscore shows the observations behind it.
A readiness score, not a visibility score. Nothing here asks an AI engine a question — it measures whether a citation is possible.
Readiness is not visibility, and the difference is the point
This audit never asks an AI engine anything. It reads your pages, your robots.txt and your markup, and reports whether a citation is possible. That is a real and useful thing to know, and it is a weaker claim than the one most GEO scores make.
CiteSite’s own headline number is different: it combines measured Google positions with live answers from ChatGPT, Gemini and Perplexity. Giving both the same name is exactly how the two get conflated, so this one is called readiness and banded in its own language.
Why the weights are uneven
Crawlability carries a quarter of the score on its own, because every other pillar is conditional on it. A site with immaculate structured data behind a Disallow: / for PerplexityBot is not 80% ready for Perplexity — it is absent from it. A weighting that treated the six as peers would report that site as doing fine.
Authority carries only a tenth, for the opposite reason. Real authority is off-site — who links to you, who mentions you, who corroborates what you claim — and none of that can be read from a crawl of your own site. What can be read is whether the site cites anything and whether its pages connect, which is a genuine signal and a small one. Weighting it higher would be padding the score with a number this tool cannot support.
What each pillar is looking at
- AI crawlability — whether robots.txt shuts any answer engine out, whether a sitemap is declared, whether pages carry noindex, whether anything failed to fetch. The crawler checker goes deeper, agent by agent.
- Entity clarity — whether anything declares who the site belongs to, whether it says so consistently, and whether
sameAsties it to profiles elsewhere. The entity analyzer shows the claims one page at a time. - Citation readiness — the share of pages carrying passages that would survive being quoted alone, and the share that declare a date and an author. The citation checker breaks this into seven signals for one page.
- Content structure — one H1 per page, real sectioning, enough body text to be worth reading, and formats an engine can lift whole.
- Structured data — how much of the site carries JSON-LD, and whether the four types that do the most work appear anywhere. The schema checker reports whether a given page’s markup actually parses.
- Authority signals — whether the site cites other sources, whether its pages link to each other, and whether it publishes an llms.txt.
Where it is deliberately shallow
Fourteen pages, one host. That is enough for signals which are properties of how a site is built rather than of any single page, and templates repeat — but it will not find the one broken page on a large site, and it does not try to.
A few checks are honest approximations rather than measurements, and the report says which. Whether a page carries a list is inferred from the shape of its prose, because a crawl keeps the text and not the markup. Authority is read from outbound links, which is a fraction of what the word means.
And a high score is not a promise. Readiness is a precondition, not a cause: a legible site can go unmentioned because nothing on it answers the question being asked.
Questions
- What does the score actually measure?
- Readiness, not visibility. It is a measure of whether the site is in a state where an AI engine could find it, understand it and cite it. Nothing in this audit asks ChatGPT, Gemini or Perplexity a single question, so it cannot tell you whether you are being cited — only whether a citation is possible.
- Why are the pillars weighted unevenly?
- Because they are not equally consequential. Crawlability carries 25% because everything else is conditional on it — perfect markup behind a Disallow is worth nothing. Authority carries 10% because what a crawl of your own site can see about your authority is genuinely thin, and pretending otherwise would be padding the score with a number we cannot support.
- It read 14 pages. Is that enough to score a whole site?
- For these signals, mostly yes — they are properties of how a site is built rather than of any one page, and templates repeat. It is not enough to find the one broken page on a large site, and it is not trying to. If your site has genuinely different sections, run the single-page tools on a page from each.
- How is this different from the AI visibility checker?
- Same crawl, same signals, presented for a different job. This one is a score across six pillars with a ranked fix list — the thing you hand to a client or a manager. The visibility checker reads the same data as a chain of five preconditions and tells you where it breaks. Use that one when something is wrong and you want to know what to fix first.
- My score is high and I am still not cited anywhere.
- That is a coherent result and worth taking seriously. Readiness is a precondition, not a cause. A site can be perfectly legible and go unmentioned because nothing on it answers the questions being asked, or because competitors answer them better. Which of those it is can only be established by asking the engines.
Related
Generative engine optimization: the complete guide — the reasoning behind all six pillars, and what the evidence does and does not support.
A ready site is one that can be cited. Whether it is cited — and by which engine, for which question, against which competitor — is what CiteSite measures, on a schedule, with the provenance of every answer recorded.