No black box. Here is every check behind the number, how much each one counts, and why the same site always scores the same.
The GEO Readiness Score is a 0 to 100 rating of how ready a website is to be cited by AI search engines like ChatGPT, Perplexity, Gemini and Google AI Overviews. It is the weighted sum of eight structural checks run live against your site. No AI model scores it, so the number is deterministic: the same site produces the same score every time, and it costs nothing to run.
Eight categories, weighted by how much each one affects whether an AI engine can reach, understand and quote your pages. The points add up to 100.
Can the AI citation crawlers reach the site at all? We read robots.txt for a global block (Disallow: /) and for any rule that shuts out GPTBot, ClaudeBot, anthropic-ai, PerplexityBot, Google-Extended, CCBot, OAI-SearchBot or Bytespider. A declared Sitemap earns the last points.
Does the site publish an llms.txt at the root, and does it follow the convention? We check for an H1 title, a summary blockquote, `##` sections and enough sectioned links to orient an AI crawler. Missing file scores zero here.
We parse every JSON-LD block and reward Organization or WebSite, then Article or BlogPosting, then FAQPage or QAPage, then BreadcrumbList, Product, HowTo or Service. Rich, valid structured data is what lets an engine understand the entity behind the page.
A title of the right length (roughly 10 to 70 characters), a meta description, a canonical link, and a full set of Open Graph tags (og:title, og:description, og:image). These frame how the page is quoted and shared.
One H1, two or more H2 sub-headings, lists, a comparison table, concrete numbers, and at least five links. Engines lift structured, fact-dense passages far more readily than walls of prose.
Schema sameAs pointing to social profiles, an About page link, and a Contact page link. These are the knowledge-graph signals that connect a page to a recognised brand entity.
An html lang attribute, visible freshness or last-updated signals, an RSS or Atom feed, and hreflang for international audiences.
A machine-readable discovery endpoint at /.well-known/ai.json or /.well-known/ai.txt. When only llms.txt is present it earns partial credit here.
The total lands in one of four bands.
Structurally ready for citation.
Solid foundation with gaps to close.
Core pieces exist, real work needed.
Crawlability or structure is blocking visibility.
The eight categories measure structure. Citability is a second 0 to 100 read of the on-page signals that actually lift AI citation. The weights reflect published GEO research, and the total is capped at 100, so a page does not need every signal to score well.
Quotable statements and blockquotes, the kind AI lifts verbatim (research links this to roughly a 41% citation lift).
Concrete stats, percentages and figures (roughly a 33% citation lift).
A short, self-contained line near the top carrying a concrete number or price.
Outbound links to authoritative sources.
Questions answered directly, through FAQ blocks or question-style headings.
A visible last-updated date.
A named author or byline.
Lists and tables an engine can lift cleanly.
A short video with timestamped chapters and a transcript.
Google's Search Quality Rater Guidelines are explicit that Trust is the most important member of E-E-A-T: Experience, Expertise and Authoritativeness all exist to support it, and a page that cannot be trusted is rated Low however expert it looks. So we grade it, out of 100, using only signals we can actually see on the page. Nothing here is a model's opinion of your credibility, which means every point lost can be traced to a specific missing signal.
A real byline, so a reader can see who stands behind the page.
Outbound links to authoritative sources, so claims can be traced.
An identifiable publisher: sameAs links, an About page and a way to make contact.
The same facts stated machine-readably, so engines can attribute the page.
The reader can tell how current the page is without guessing.
Experience, the first E, is the one we deliberately do not score. Whether an author has genuinely used the thing they are writing about is not something a crawler can verify, and inventing a number for it would be the opposite of a trust signal. It stays a judgement for you and your reviewers.
Patterns that AI and search engines actively distrust are surfaced as warnings rather than folded silently into the number, so a false positive can only add a line to review. We flag three: hidden blocks of keyword-stuffed text, a single word repeated far past natural use, and prompt-injection text aimed at AI crawlers (for example, "ignore previous instructions"). Thresholds are kept high to stay clear of legitimate repetition.
Every input is parsed from real, public files: your robots.txt, your llms.txt, your page HTML, and your /.well-known/ discovery files. No language model is asked to judge the page, so there is no variance between runs and no per-audit cost. Run it a hundred times and an unchanged site returns the same score a hundred times. That is what makes the number worth putting in front of a client and worth tracking month over month.
It is a single 0 to 100 rating of how ready a website is to be discovered and cited by AI search engines such as ChatGPT, Perplexity, Google Gemini and AI Overviews, and Microsoft Copilot. It is the weighted sum of eight structural checks run live against the site.
Crawlability (robots.txt) and llms.txt carry 18 points each, Schema 16, Meta tags 14, Content structure 12, Brand and entity 10, Signals 6, and AI discovery 6. The points add up to 100.
No. The score is computed by parsing real signals from your site: robots.txt, llms.txt, the page HTML, and the /.well-known/ discovery files. No language model scores it, so the same site always produces the same number and no tokens are spent.
Out of 100, from signals visible on the page: a named author (25), cited sources (20), brand and entity clarity (20), author and organisation schema (20), and a visible last-updated date (15). Trust is weighted most heavily because Google's Search Quality Rater Guidelines treat it as the centre of E-E-A-T. Experience is not scored, because a crawler cannot verify whether an author has genuinely used what they write about.
The eight-category score measures structural readiness. The citability score is a separate 0 to 100 read of on-page signals that lift AI citation: quotations, statistics, a quotable answer line, cited sources, direct answers, freshness, a named author, extractable structure, and chaptered video.
Spam signals are surfaced as warnings rather than folded silently into the number. We flag hidden blocks of keyword-stuffed text, a single word repeated far past natural use, and prompt-injection text aimed at AI crawlers. Thresholds are kept high to avoid false positives.
Run any URL through the free GEO Readiness Checker. It returns the eight category scores, the citability read, the content-hygiene warnings, and a prioritised list of the biggest gaps to fix first.
Run any URL through the free checker for the full category breakdown, the citability read, and the gaps to fix first.
Prefer a hands-on list? Work through the 250-point AEO & GEO checklist or learn the fundamentals in the free Academy.