Last updated July 2026.
Most pages fail to get cited not because the information is wrong, but because the structure is invisible to the retrieval layer. AI engines do not read pages the way humans do. They look for a clean, extractable answer near the top, a logical hierarchy of headings to navigate the rest, and structural signals (lists, schema, outbound citations) that tell the model the page can be trusted.
The scorecard below turns those signals into a numbered checklist with weights. Run it on any page before it ships. Fix the gaps. Publish with confidence.
How the scorecard works
Each of the 12 items below carries a point weight. Add up your points. The total lands you in one of three verdict tiers.
| Score | Verdict | Action |
|---|---|---|
| 70-100 | Ship | Structural citability is in place. Publish. |
| 50-69 | Fix first | At least one high-weight signal is missing. Fix before publishing. |
| 0-49 | Rewrite | The page lacks the structural foundation for AI citation. Rebuild. |
Use the printable table at the end of this post to score each item and tally your total.
The 12-point AI Citability Scorecard
1. Direct-answer opening (40-60 words)
Rule: The page opens with a self-contained, 40-60 word answer to the target query. That answer is complete on its own: a reader who sees only that paragraph walks away knowing the answer.
Why engines weight it: According to Kevin Indig’s 2026 analysis of 1.2 million ChatGPT responses (reported by Search Engine Land), 44.2% of ChatGPT citations were drawn from the first 30% of a page’s content. The retrieval layer grabs the first clean passage that looks like an answer. Pages that bury the conclusion get paraphrased out of the citation.
Scoring:
- 40-60 words, self-contained, first paragraph: 14 points
- Present but longer than 60 words or positioned below fold: 7 points
- Missing or buried below other content: 0 points
2. Sequential heading hierarchy
Rule: Headings run in logical order: one H1, then H2 sections, then H3 subsections. No skipped levels. No decorative H2s that exist purely for visual styling.
Why engines weight it: According to AirOps’ July 2025 study of 12,000+ URLs, 68.7% of pages cited by ChatGPT used a sequential heading structure (H1 to H2 to H3), compared to just 23.9% of Google’s top-ranked pages for the same queries. The heading tree is how AI engines build their internal map of a page.
Scoring:
- Sequential H1 to H2 to H3, no skipped levels: 10 points
- Minor inconsistency (one skipped level, one decorative heading): 5 points
- Flat structure or multiple H1s: 0 points
3. Stat density
Rule: The page includes at least two verifiable, attributed statistics. Each stat names the source and the year. Stats are placed in the first half of the page where possible.
Why engines weight it: Factual density is a core citability signal. A peer-reviewed GEO study by researchers at Princeton, Georgia Tech, IIT Delhi, and the Allen Institute for AI (presented at ACM KDD 2024) found that adding statistics to content improved AI visibility scores by roughly 40% on their Position-Adjusted Word Count measure. An extractable passage with a named source and a number is exactly what a retrieval layer prefers.
Scoring:
- 2 or more attributed stats with source and year: 10 points
- 1 attributed stat: 5 points
- No stats or unnamed “studies show” references: 0 points
4. Schema markup
Rule: The page implements at least one of these schema types: Article, FAQPage, HowTo, or Author. Schema is implemented as valid JSON-LD in the page head.
Why engines weight it: Schema helps AI engines parse page structure and intent. Note the important caveat: an Ahrefs study of 1,885 pages (published May 2026) found no statistically significant uplift in AI citations from adding schema to pages already receiving citations. Schema is necessary for structured parsing, not a shortcut to citation. It belongs on every page, but it does not compensate for thin prose.
Scoring:
- 2 or more schema types (Article + FAQPage, or Author + HowTo): 8 points
- 1 schema type, valid JSON-LD: 4 points
- No schema: 0 points
5. Freshness signal
Rule: The page displays a visible last-updated date. High-priority pages are reviewed and updated at least every 90 days.
Why engines weight it: According to ConvertMate’s 2026 AI Visibility Study (which analyzed 80 million citations across 10,000+ domains), 76.4% of ChatGPT’s most-cited pages had been refreshed within the prior 30 days. A visible date both signals recency to engines and builds reader trust. Note: the ConvertMate study is a single-vendor finding and the underlying methodology has not been independently verified.
Scoring:
- Visible last-updated date, page reviewed within 90 days: 8 points
- Visible publish date only, no update date: 4 points
- No date visible anywhere on page: 0 points
6. Author schema and byline
Rule: The page carries an Author schema block with name, credentials, and a link to an author bio. The byline is visible in the page body.
Why engines weight it: E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is the framework Google’s quality guidelines use to evaluate sources. A named author with credentials is one of the clearest E-E-A-T signals available. Engines trained on quality signals weight author-credentialed content over anonymous pages, particularly for advice-giving queries.
Scoring:
- Author schema (name + credentials + bio link) and visible byline: 8 points
- Visible byline only, no Author schema: 4 points
- No byline or author attribution: 0 points
7. Structured list presence
Rule: The page includes at least one structured list (bulleted or numbered) of three or more parallel items. Lists are used for genuinely list-worthy content, not as decorative formatting.
Why engines weight it: According to AirOps’ July 2025 analysis of 12,000+ ChatGPT-cited URLs, nearly four out of five pages cited by ChatGPT include at least one structured list, compared to just 29% of Google’s top-ranked pages. Lists give engines a clean, extractable chunk that can be reproduced verbatim in an answer.
Scoring:
- 2 or more structured lists on the page: 8 points
- 1 structured list: 4 points
- No lists, or lists with fewer than 3 items: 0 points
8. Outbound citations
Rule: The page cites at least two external, named sources with a publication name and year (and a link where possible). Sources are authoritative: peer-reviewed papers, government databases, established industry research, or named company studies with disclosed methodology.
Why engines weight it: Pages that cite others are structurally similar to pages that get cited. Outbound citations signal that the author did research, not just content generation. An AI engine’s retrieval layer treats a source-cited page as more credible than an uncited one because it can trace the claim back to a verifiable origin.
Scoring:
- 2 or more named, attributed external sources: 8 points
- 1 named, attributed external source: 4 points
- No external citations, or only unnamed references: 0 points
9. Extractable passage length
Rule: The page contains at least one passage of 40-120 words that answers the target query completely, on its own, without requiring surrounding context. This is the passage an engine will quote verbatim.
Why engines weight it: Generative engines do not quote whole articles. They extract a passage. If no passage on the page is the right length (too short to be complete, too long to be used whole), the engine paraphrases instead of cites. A paraphrase rarely credits the source. A direct quote almost always does.
Scoring:
- 1 or more 40-120 word passages that are self-contained answers: 10 points
- A passage exists but is under 40 words (incomplete) or over 150 words (unwieldy): 5 points
- No extractable self-contained passage: 0 points
10. Cross-platform entity presence
Rule: The brand, author, or site is mentioned on at least two platforms AI engines already trust (Wikipedia, Reddit, LinkedIn, G2, Capterra, established editorial sites, or industry publications). These mentions are verifiable and consistent.
Why engines weight it: According to Profound’s analysis of 100,000 prompts across ChatGPT and Perplexity, only 11% of cited domains overlap between the two platforms. An engine cannot cite a brand it has never encountered in its training or retrieval data. Third-party mentions on trusted domains build that recognizability. A brand cited only on its own website has a much narrower citation footprint than one discussed on G2, an industry blog, and a journalist’s roundup.
Scoring:
- Verified mentions on 3 or more trusted third-party platforms: 8 points
- Mentions on 1-2 trusted third-party platforms: 4 points
- No verifiable third-party mentions found: 0 points
11. Entity clarity
Rule: The page states clearly what the product or service is, who it is for, and what problem it solves, within the first two paragraphs. Brand name, product category, and primary use case appear in the page’s title tag and meta description.
Why engines weight it: AI engines resolve ambiguity by skipping ambiguous sources. If a page could be about three different things, the model moves on to a clearer one. Entity clarity is the GEO equivalent of keyword clarity in traditional SEO: a page that knows exactly what it is about, and says so immediately, is easier to retrieve and attribute correctly.
Scoring:
- Product/service, audience, and use case stated within first 2 paragraphs, consistent with title and meta: 6 points
- Entity is clear but positioned later on page or inconsistent with meta: 3 points
- Entity is ambiguous or not stated explicitly: 0 points
12. Internal links to supporting pages
Rule: The page links to at least two related pages on the same domain that go deeper on subtopics the page introduces. Anchor text is descriptive, not generic (“learn more”).
Why engines weight it: Internal links tell the retrieval layer that the page exists within a web of topically related content, not in isolation. Topical authority, built through a cluster of interlinked pages, is a meaningful signal in how AI engines weight sources for authoritative queries. A standalone page with no internal connections looks thin even if the prose is excellent.
Scoring:
- 2 or more descriptive internal links to related pages: 2 points
- 1 internal link or generic anchor text: 1 point
- No internal links: 0 points
Printable scorecard
Use this table to score any page before it ships. Fill in the “Your score” column. Total at the bottom.
| # | Signal | Max points | Your score |
|---|---|---|---|
| 1 | Direct-answer opening (40-60 words) | 14 | |
| 2 | Sequential heading hierarchy | 10 | |
| 3 | Stat density (2+ attributed stats) | 10 | |
| 4 | Schema markup (FAQPage, Article, HowTo, Author) | 8 | |
| 5 | Freshness signal (visible date, updated within 90 days) | 8 | |
| 6 | Author schema and byline | 8 | |
| 7 | Structured list presence | 8 | |
| 8 | Outbound citations (2+ named sources) | 8 | |
| 9 | Extractable passage (40-120 words, self-contained) | 10 | |
| 10 | Cross-platform entity presence (3+ trusted platforms) | 8 | |
| 11 | Entity clarity (stated in first 2 paragraphs) | 6 | |
| 12 | Internal links (2+ descriptive, related) | 2 | |
| Total | 100 |
Verdict:
- 70-100: Ship.
- 50-69: Fix the items scoring 0 on items 1, 2, 9 first, then reassess.
- 0-49: Rewrite before publishing.
Tools that automate the check
Running this scorecard manually is fast for one page. At scale, you need a tool.
Temso (from $89/mo) is the all-in-one AI SEO platform that covers the full loop: it tracks citation share across 8 engines, surfaces which pages are and are not being cited, identifies the structural gaps, and guides you through fixing them. For teams that want to go from zero citation visibility to active citation growth without switching tools, it is the obvious starting point. It does not replace a dedicated keyword-rank tracker for traditional SEO, but for GEO it handles monitoring, diagnosis, and execution in one subscription.
Profound goes deeper on citation source attribution at the enterprise level. Its citation maps show exactly which URLs AI engines pull from for your target queries, which makes it easier to diagnose why a page scores well on the scorecard but still does not appear in answers. Entry price for meaningful engine coverage is $399/mo.
Surfer integrates a citability-oriented Content Score directly into the writing editor, so writers can optimize for AI citation signals (NLP terms, heading structure, content completeness) while drafting. Useful when the bottleneck is writing volume, not analysis. From $99/mo.
Otterly.AI runs a GEO Audit against 20+ on-page citation-readiness factors on any URL, and its prompt-level citation tracking shows which engines cite you and for which queries. A solid option for freelancers and small agencies. From $29/mo (Lite, 15 prompts).
The full ranking of GEO tools with pricing, engine coverage, and head-to-head comparisons lives at /rankings/geo-tools. Key terms used in this post are defined in the /glossary.
One step after the scorecard
The scorecard tells you whether a page is structurally ready to be cited. What it does not tell you is whether the engines are actually citing it. Those are different questions.
After you publish a page that scores 70 or higher, set a tracking prompt in your GEO monitoring tool. Run the target query across ChatGPT, Perplexity, Gemini, and Google AI Overviews. If your page does not appear within six to eight weeks, the structural foundation is in place and the gap is likely entity authority, not on-page signals. That points to the third-party mention work in item 10.
For the full methodology behind how we evaluate GEO content readiness, see /methodology.