Search is being replaced by AI answers, and the new goal is to be the page that gets cited. This free AI visibility checker scores how likely answer engines like ChatGPT, Gemini, and Perplexity are to quote a page — based on the structural signals they reward, from FAQ schema to extractable data. Your score appears below, with every failing signal named and the fix for each one. One check is a snapshot: save it to a free account to re-run it and get told when something slips.
How the AI visibility checker works
Enter a public URL and the tool fetches the page the way a crawler would, then checks it against eighteen structural signals that correlate with being cited in AI answers. It reports two scores and a prioritised list of what is failing. It measures on-page citability — it does not query live AI models — so pair it with real share-of-model tracking for the complete picture.
The two scores, and what each one means
Agent readiness — can AI find and parse you. This grade combines the discovery and understanding signals: whether AI crawlers are allowed in, whether the page is reachable and canonical, and whether its structure is explicit enough for a machine to extract the facts without guessing. A low agent-readiness score usually means the engines never get a clean read of the page at all, so nothing downstream can help.
Citation score — will AI quote you. This grade reflects the content and trust signals: depth, specific data worth quoting, named authorship, and entity links that let an engine attribute the page to a real organisation. A page can be perfectly readable and still never be cited because it gives the engine nothing concrete to lift.
Both scores are graded A to F. Treat anything below a B as a to-do list rather than a verdict; every failing signal links to the tool that fixes it.
The eighteen signals, grouped by what they decide
Discovery — can the engines reach the page?
- AI crawlers allowed. If robots.txt blocks GPTBot, OAI-SearchBot, PerplexityBot or ClaudeBot, the page cannot be read by those engines regardless of anything else on this list. This is the single most common reason a well-built site is invisible to AI.
- llms.txt file. A plain-text index at the site root that points language models to your most important pages. Not every engine reads it yet, but it costs nothing and removes ambiguity about what matters on your site.
- XML sitemap. The canonical map of what exists. Engines that crawl rather than search still rely on it.
- Canonical tag. Tells the engine which version of a duplicated or parameterised page is the real one, so authority is not split across copies.
- HTTPS. Baseline. Several crawlers deprioritise or skip insecure pages.
Understanding — can the engines parse what the page says?
- JSON-LD structured data. Any schema at all signals that the page describes itself in machine-readable form.
- Organization schema. Names the entity behind the page, so a citation can be attributed to a brand rather than a bare URL.
- FAQ schema. Question-and-answer pairs map directly onto how answer engines format responses. Pages with FAQPage markup are disproportionately quoted.
- Open Graph tags. A clean title, description and image for the page as a shareable object; engines use them as a fallback summary.
- Descriptive title. A title tag that states what the page is about in plain language, not a brand slogan.
- Meta description. A one-sentence summary an engine can use verbatim.
- Single H1. One unambiguous statement of the page topic. Multiple H1s read as multiple topics.
- Clear subheadings. At least three descriptive H2 or H3 headings. Engines extract answers section by section; headings are the section labels.
- Lists or tables. Structured elements an engine can lift whole. A comparison table or a numbered list is far more quotable than the same facts in prose.
Trust and content — will the engines choose to quote it?
- Enough depth. Roughly 600 words or more of real content. Thin pages rarely contain anything an engine considers worth attributing.
- Stats or data. Specific numbers, dates and measurements. Answer engines favour pages that give them a concrete figure to cite.
- Author and provenance. A named author and a visible publication or update date. Engines increasingly weight who said it and when.
- Entity sameAs links. Links from your Organization schema to your official profiles elsewhere, so the engine can confirm you are who the page says you are.
What the free check is, and what it is not
The check is free with no account, no card and no trial clock — run it on as many pages as you like. It sits alongside 69 other free tools that do not expire. It is the step before monitoring: it tells you whether the engines can read and quote a page, and exactly what to fix if they cannot.
What it does not do is ask the engines anything. It will not tell you whether ChatGPT, Perplexity, Gemini, Claude or Copilot actually name your brand today, or which competitors they name instead. That is a different measurement — share of model — and it requires querying the engines on a schedule. The Prompt Observatory does that weekly for $49 a month, and the natural order is: fix what this check finds first, then start tracking whether the fixes moved the answers.
Frequently asked questions
How do AI answer engines decide what to cite?
They favor content that is clearly structured and easy to extract: concise answers, headings, lists, tables, statistics, and schema markup, backed by trust signals like author identity and a credible domain. Research also shows a strong correlation between ranking in traditional search and being cited in AI answers.
What is share-of-model?
Share-of-model is how often your brand is mentioned or cited across AI answer engines, the AI-era equivalent of search ranking. This tool measures the on-page factors that influence it; tracking actual mentions requires an AI-visibility monitoring service.
How can I improve my chances of being cited?
Lead with clear answers, add FAQ or HowTo schema, include original statistics and data, use descriptive headings and lists, and strengthen author and entity signals. Strong traditional SEO remains the foundation.
Is the checker really free, and do I need an account?
The check itself is free and needs no account — run it as often as you like. An account is only useful if you want the result kept: saving it lets you re-run the same URL on a schedule, compare scores over time, and get an email when a signal you had passing starts failing.
What counts as a good score?
An A or B on both scores means the page is readable and has something worth quoting; the remaining work is competitive, not structural. A low agent-readiness grade with a high citation grade is the most common and most fixable pattern: good content the engines cannot reach, usually because of robots.txt or a missing canonical. The reverse — readable but nothing to quote — is a content problem, not a technical one.
Which AI engines do these signals matter for?
All of the major ones. ChatGPT, Perplexity, Gemini, Claude, Copilot and Google AI Overviews each crawl or search the web and each rewards the same basics: reachable, well-structured, specific, attributable pages. The crawler-access signal is engine-specific — each has its own user agent — which is why the checker names which bots your robots.txt is blocking.
Does adding schema guarantee a citation?
No. Schema makes a page unambiguous to the systems deciding what to quote; it does not make the content worth quoting. Pages that pass every structural signal but say nothing specific still go uncited. Schema is the floor, not the ceiling.
How often should I re-check a page?
After any change to the page, its template, robots.txt or your schema — those are the moments a passing signal quietly starts failing. Otherwise monthly is enough for the structural signals, since they change only when you change them. Whether the engines actually cite you is the thing that moves week to week, and that is what the Observatory tracks.