Your GEO Score, category by category

Content Quality (25 of 100 points)

Is there enough substance on the page to synthesise an answer from?

An answer engine has to lift a complete thought out of your page. Thin pages, or pages where the structure does not signal what each section is about, get skipped in favour of a competitor who spelled it out.

• Page depth — 10 pts: Average words per crawled page. Thin pages across the site cost the most of any single check on the rubric.
• Cornerstone long-form page — 5 pts: At least one page of 1,500+ words covering your core topic end to end.
• Heading hierarchy — 5 pts: One H1 per page, H2/H3 used in order, headings that describe the section rather than sell it.
• No placeholder copy — 3 pts: No lorem ipsum, no 'Your text here', no template stubs left in production.
• Site breadth — 2 pts: More than a single page of real content for a crawler to work with.

The second version contains extractable facts: a service, a coverage area, a response time, a duration, a price. That is what a model quotes.

• Find your three most important pages and take each past 800 words of specifics, not adjectives.
• Publish one cornerstone page of 1,500+ words on the topic you want to own.
• Give every page exactly one H1, and make each H2 a sentence a person would actually ask about.
• Search your site for placeholder text and remove it.

Citation Worthiness (20 of 100 points)

Is anything on the page liftable as a quote?

Models cite the sentence they can reuse without rewriting. Verifiable numbers, one-line definitions and clean tables are far easier to reuse than marketing prose.

• Factual density — 8 pts: Verifiable claims per page: statistics, dates, prices, measurements, named specifics.
• Quotable definitions — 3 pts: Plain 'X is Y' sentences a model can lift verbatim.
• Comparison tables — 3 pts: At least one real HTML table comparing options, plans, or specifications.
• Outbound citations — 3 pts: Links to sources outside your own domain, supporting the claims you make.
• Freshness signal — 3 pts: A visible, recent published or updated date on substantive pages.

Numbers, dates and a named source. Each one is a claim a model can attribute to you rather than paraphrase away.

• Replace three superlatives per page with a number you can defend.
• Open each key page with a one-sentence definition of the thing it is about.
• Turn at least one 'option A vs option B' passage into a real HTML table.
• Cite your sources with outbound links, and show a real last-updated date.

Structured Data (18 of 100 points)

Can a machine read who you are and what this page is, without guessing?

JSON-LD is the only part of your page that is unambiguous. Everything else is prose a model has to interpret.

• Organization / WebSite schema — 5 pts: A JSON-LD Organization node with name, url, logo and contact details, plus a WebSite node.
• Schema health — 4 pts: The JSON-LD you ship parses cleanly and uses valid @type values.
• Metadata and social tags — 4 pts: Title, meta description in the useful length range, Open Graph and Twitter card tags.
• Article schema on posts — 3 pts: Blog and news pages carry Article or BlogPosting with author and dates.
• Breadcrumb schema — 2 pts: A BreadcrumbList node so engines understand where a page sits in the site.

Five lines of JSON-LD answer 'who is this' with no inference. Missing Organization schema alone is 5 of the 18 points in this category.

• Ship Organization and WebSite JSON-LD sitewide, in the head of every page.
• Add BlogPosting schema with an author and dates to every article.
• Add BreadcrumbList to nested pages.
• Validate your JSON-LD — a block that fails to parse scores as if it were absent.

Authority Signals (17 of 100 points)

Is there a real, accountable organisation behind this content?

Answer engines carry the reputational cost of a bad citation, so they lean towards sources that name a company, an author and a way to be contacted.

• Author bylines — 4 pts: Named human authors on substantive content, not 'Admin' or 'Team'.
• About or Team page — 3 pts: A crawlable page describing the organisation and the people in it.
• Social proof — 3 pts: Named testimonials, case studies, client logos or verifiable reviews.
• Privacy policy and terms — 3 pts: Both present and linked from the footer.
• Founding date — 2 pts: A visible founding or 'since' year.
• Contact information — 2 pts: A real postal address, a phone number in a recognisable format, and an email.

A name, a role and a date turn anonymous copy into attributable expertise.

• Put a named author and a date on every article.
• Publish an About page with real people, and a founding year.
• Show your full contact block: address, phone, email.
• Link privacy policy and terms from the footer on every page.

FAQ & Q&A Format (12 of 100 points)

Do you answer questions in the shape a model is asking them?

AI search is question-shaped. A page that already pairs a question with a short, complete answer needs no reformatting to be used in an answer.

• Dedicated FAQ page — 4 pts: A crawlable FAQ page, or a full FAQ surface on the scanned page.
• FAQ depth — 3 pts: Enough distinct questions to cover the topic, deduplicated across templates.
• FAQPage schema — 3 pts: Valid FAQPage JSON-LD with real Question and acceptedAnswer entries.
• Question-format headings — 2 pts: Headings written as the question a visitor would type.

The heading matches the question, and the answer is complete on its own without the surrounding page.

• Write down the 15 questions your sales team answers weekly, and publish them.
• Answer each in 40–80 words, complete without context.
• Wrap the visible FAQ in FAQPage JSON-LD. Schema without visible answers does not count.
• Rewrite section headings as questions where it reads naturally.

Crawler Accessibility (8 of 100 points)

Can the crawler see the page at all?

The smallest category by weight, and the one that ends the conversation when it fails. If the raw server response is empty, nothing else on the rubric can be earned.

• Indexable pages — 3 pts: No noindex on the homepage or on the majority of live pages.
• Mobile viewport meta — 2 pts: A viewport meta tag in the head.
• Server-rendered content — 2 pts: The main content is in the raw HTML, not injected only after JavaScript runs.
• Canonical tag — 1 pts: A self-referencing canonical URL on each page.

Some engines render JavaScript and some do not. Content present in the first response is content every engine can use.

• Fetch your own page with JavaScript disabled and check the text is there.
• Remove stray noindex tags from live pages.
• Add the viewport meta tag and a self-referencing canonical.
• Publish robots.txt, a sitemap and llms.txt — reported as findings, not deductions.

What your total means

• 80–100 — A: EXCELLENT
• 65–79 — B: GOOD
• 50–64 — C: MODERATE
• 35–49 — D: POOR
• 0–34 — F: CRITICAL

Two things override the total

• Homepage, or most live pages, set to noindex caps the score at 30. If you have told search engines not to index the site, nothing else can lift the score above 30 until that is removed.
• Content Quality below 8 of 25 caps the score at 60. A site with almost no readable content cannot score in the top bands on the strength of schema alone.

Quick Scan vs Site Analysis

A Quick Scan reads a small number of pages, so these site-level checks cannot be proven and are removed from the total rather than failed: Dedicated FAQ page, About / Team page, Cornerstone long-form page (1,500+ words), Site breadth (multi-page coverage), Site-wide FAQ depth.

Questions about the score

What is a good GEO Score?

80 and above is the top band (EXCELLENT). 65–79 is GOOD, 50–64 MODERATE, 35–49 POOR, and 0–34 CRITICAL. The bands are the same ones used everywhere in the product, including the badge and the PDF report.

Why is Content Quality worth more than Crawler Accessibility?

Weights reflect how much each category can move a citation outcome, not how often it fails. Crawler Accessibility is only 8 points because it is close to binary: it either works or it caps everything else. Content Quality is 25 because it has the widest range of outcomes between a thin page and a comprehensive one.

Can a perfect rubric score guarantee citations?

No. The GEO Score measures whether your pages can be read, understood and quoted. Whether a specific engine cites you for a specific prompt also depends on competition, off-site authority and the engine's own retrieval choices. That is why AI Visibility testing is reported separately from the structure score.

Why did my Quick Scan and Site Analysis give different scores?

A Quick Scan crawls a handful of pages, so five site-level checks (dedicated FAQ page, About page, cornerstone page, site breadth, site-wide FAQ depth) cannot be proven. Those points are removed from the denominator instead of being scored as failures, and the remainder is normalised to 100. A Site Analysis crawls the whole site and scores the full rubric.

What is the score cap and why did it hit my site?

Two conditions override the category total: a noindexed homepage or majority of live pages caps the score at 30, and Content Quality below 8 of 25 caps it at 60. Both are reported with the specific remedy, and the projected score shows where you land once the cap is cleared.

Which scoring version is this?

GEO Score v3.6. Category weights and point values are frozen per version — a change to any of them requires a new version, so your score history stays comparable.