The AEOCrawler scoring framework measures every page across 9 weighted dimensions that determine how likely AI search engines are to cite it: Answer Extraction, Citation Probability, Structured Data, Entity Authority, Query Coverage, Semantic Coverage, Conciseness, Readability, and Content Freshness. Each dimension is scored 0–100, and together they combine into a composite AI Visibility Score. A composite above 70 indicates strong AI citability. Below 50 means the content is effectively invisible to AI-generated answers regardless of its traditional search ranking.
Understanding how each dimension is calculated — and what signals drive it up or down — is the difference between guessing at AI optimization and executing it systematically. This article breaks down all 9 dimensions, explains the underlying signals, and shows what a professionally optimized page looks like at every score tier.
Why a Single AEO Score Isn't Enough
Before the framework, a note on why it exists in 9 dimensions rather than one.
A single aggregate AEO score tells you whether content is well-optimized or poorly optimized. It does not tell you why. A page that scores 58/100 could have that score for dozens of different reasons: absent answer blocks, inconsistent entity naming, thin schema coverage, low information density, or no coverage of adjacent queries. Each problem has a different fix. Without dimensional scoring, you know something is wrong but not what.
This is the same reason a doctor orders a panel of tests rather than a single "health score." The value is in the breakdown, not the headline number.
AEOCrawler's 9-dimension framework was built around the signals that retrieval-augmented generation (RAG) systems — the underlying architecture of ChatGPT, Perplexity, Google AI Overviews, and others — actually weight when selecting content to cite. Each dimension corresponds to a real mechanism in how these systems evaluate, index, and extract from web content.
For background on what AEO is and why it matters, start here.
Before the Dimensions: AI Engines Have to Find You First
Discovery is a prerequisite, not a scored dimension. Before any of the 9 dimensions matter, an AI engine has to locate, crawl, and retrieve your page at all. AEOCrawler's crawl layer checks this foundation during every analysis: indexation status, robots.txt and meta-robot directives, sitemap presence, canonical tag clarity, page load performance (AI agents time out on slow pages), and JavaScript-heavy rendering that AI crawlers cannot parse.
If discovery is blocked, no amount of content optimization helps. Technical SEO fundamentals are AEO fundamentals — the relationship between SEO and AEO is additive, not separate. The 9 dimensions below assume the engine can reach your content; they score what happens next.
Dimension 1: Answer Extraction (0–100) — weight 20%
What it measures: How easily an AI engine can extract a complete, citable answer from your content without needing to synthesize across multiple disparate sections.
This is the structural dimension, and it carries the highest weight in the framework. AI engines don't read pages the way humans do — they scan for extractable blocks. The easier you make it for the engine to isolate a complete answer, the more likely it is to use yours.
Signals that drive this score:
- Presence and placement of a direct answer block in the first 60 words (the single highest-impact factor in this dimension)
- Question-format headings (H2/H3 that mirror natural query phrasing)
- Paragraph length — blocks over 150 words are harder to extract cleanly
- Use of structured elements: bullet lists, numbered lists, comparison tables
- Section-level answer density — does each major section contain at least one citable fact?
- Presence of FAQPage schema with complete question-answer pairs
What a good score looks like: 75+ indicates content is structured for extraction throughout. Below 60 typically means no direct answer block, long paragraphs, or prose-heavy formatting that AI systems find difficult to parse.
Common fix: Most pages that score poorly on Answer Extraction are missing a direct answer block in the opening section. Write a 40–60 word paragraph immediately after the H1 that states the answer to the core question as a declarative fact. No preamble. No "In this article, we'll explore..." This single change often moves Answer Extraction scores by 15–25 points.
The AEO checklist covers the structural requirements in detail, including schema implementation and list formatting.
Dimension 2: Citation Probability (0–100) — weight 18%
What it measures: The likelihood that, once an AI engine has retrieved your content, it will choose to surface it as a citation in a generated answer.
This is arguably the most commercially important dimension. A page can be perfectly extractable and still never be cited — because citation is a competitive selection event. When an AI engine retrieves 10 candidate pages for a query, it cites 1 to 3. Citation Probability measures how your content stacks up in that selection, including the publisher-reputation signals AI engines use as trust proxies.
Signals that drive this score:
- Recency of content — AI engines prefer pages updated in the last 90 days (research shows AI-surfaced content is 25.7% fresher than standard search results)
- Author and publisher credibility signals (schema-declared authorship, About pages, author bios with credentials)
- Backlink profile quality as a proxy for source authority
- Outbound citations — pages that cite credible external sources are treated as more trustworthy
- Brand entity recognition — is the publisher a recognized entity in knowledge graphs?
- Content depth relative to competing pages on the same topic
What a good score looks like: 70+ indicates strong selection candidacy against average competition. In highly competitive queries, 80+ may be needed to win citations consistently.
Common fix: Citation Probability is often dragged down by missing authorship schema, no visible publication/update dates, or lack of external citations in the content body. Adding Article schema with author, dateModified, and publisher fields is frequently the fastest single improvement.
Dimension 3: Structured Data (0–100) — weight 12%
What it measures: How well your page's machine-readable markup supports AI understanding and attribution — the architecture beneath the content.
If Answer Extraction scores the quality of your extractable content blocks, Structured Data scores the quality of the container. A page can have excellent content that AI engines struggle to use because the schema is malformed, the heading hierarchy is broken, or the markup is inconsistent.
Signals that drive this score:
- Schema markup validity and completeness (validated against schema.org spec, no errors)
- Schema type coverage: Article, FAQPage, Organization, BreadcrumbList where appropriate
- Heading hierarchy: logical H1 → H2 → H3 progression with no skipped levels
- HTML semantic correctness: proper use of
<article>,<section>,<aside>,<header> - Table markup: comparison data in proper
<table>elements, not images or CSS-faked layouts - List markup:
<ul>and<ol>elements rather than dash-separated paragraphs
What a good score looks like: 75+ indicates clean, semantically correct HTML with valid schema throughout. Below 55 often reveals broken schema (common on WordPress sites with conflicting plugins), invalid heading hierarchies, or data presented in non-parseable formats.
Common fix: Validate your schema using Google's Rich Results Test and Schema.org's validator. Fix all errors — not just warnings. Then audit your heading structure: the H1 should appear exactly once per page, H2s should be the primary section dividers, and H3s should be subsections. Never skip levels (no jumping from H1 to H3).
Dimension 4: Entity Authority (0–100) — weight 12%
What it measures: The clarity and consistency of the named entities in your content — your brand, products, people, concepts, and the connections between them — as perceived by AI knowledge graph processing.
AI engines build and maintain knowledge graphs about brands, products, people, and topics. When your content contains entity signals that are consistent, well-defined, and machine-readable, you get credited toward topical authority. When entities are fragmented — your brand name appears in three different spellings, your product names conflict across pages — the AI sees multiple weak entities instead of one strong one.
Signals that drive this score:
- Entity consistency: the same name used for the same thing everywhere (page content, schema markup, meta tags, alt text, anchor text)
- Schema-declared entities: Organization, Person, Product, SoftwareApplication markup
- Entity co-occurrence: does your brand name appear alongside the correct category terms?
- Knowledge graph cross-references: Wikipedia, Wikidata, Google Knowledge Panel presence
- Mention density: does your primary entity appear enough times to establish subject matter?
- Entity disambiguation: does the content clarify which "entity" is being discussed when names could be ambiguous?
What a good score looks like: 70+ indicates well-defined entities with consistent expression. Below 50 often reveals inconsistent brand naming, absent schema, or content so generic that no strong entity signals emerge.
Common fix: Conduct an entity audit. Search your site for every variation of your brand name and key product names. Standardize them. Then add Organization schema to your homepage and SoftwareApplication schema (for SaaS products) to your product pages. These two steps alone frequently produce 10–20 point gains in Entity Authority.
Dimension 5: Query Coverage (0–100) — weight 10%
What it measures: The breadth of search intent your content addresses — how many variations of the core query does the page answer?
When a user asks ChatGPT "What is AEO?", they're not asking a single question. Beneath that surface query is a set of sub-intents: How does AEO work? Why does AEO matter? How does AEO differ from SEO? What tools are used for AEO? AI engines tend to surface content that satisfies the full intent cluster, not just the headline question.
Signals that drive this score:
- Coverage of the core query with a direct answer
- Coverage of natural follow-up questions (What, How, Why, When, Who, Which)
- Comparison coverage: does the content address alternatives or competitive context?
- Use-case coverage: does the content explain when/why the concept applies?
- Risk/limitation coverage: does the page acknowledge what doesn't work or when exceptions apply?
- Related-entity coverage: does the content connect the main topic to surrounding concepts?
What a good score looks like: 70+ indicates broad intent coverage with the core and major sub-intents addressed. Below 55 typically means the content is thin — it answers the headline question but leaves natural follow-ups unanswered.
Common fix: Use question-format subheadings to systematically cover sub-intents. For every page, ask: "What would a reader naturally ask after learning the headline answer?" Create a section for each. This improves Query Coverage and also makes the content more useful to human readers.
Understanding how AI engines select which content to cite helps explain why intent breadth matters so much in the selection process.
Dimension 6: Semantic Coverage (0–100) — weight 10%
What it measures: The topical depth and vocabulary completeness of your content — whether it uses the full range of terms, concepts, and relationships that experts and AI engines associate with the topic.
Semantic Coverage is related to Query Coverage but distinct. Where Query Coverage measures whether you answer enough questions, Semantic Coverage measures whether you use the right vocabulary and conceptual framework. AI engines trained on vast corpora have a strong sense of which terms should co-occur when discussing a given topic. Content that's missing key semantic neighbors signals shallow expertise.
Signals that drive this score:
- Presence of core vocabulary for the topic domain (synonyms, related terms, category labels)
- Use of technical terminology appropriate to the content type
- Conceptual completeness: are the important sub-concepts of the topic present?
- Semantic co-occurrence with established authoritative content on the topic
- Content length relative to semantic expectations for the query type (shallow topics need less; complex topics need more)
- Absence of off-topic dilution: does the content stay on-topic throughout?
What a good score looks like: 65+ indicates solid topical vocabulary. Very high scores (85+) indicate the content reads like domain expert writing — complete vocabulary, proper conceptual coverage, no semantic gaps.
Common fix: Read the top-cited content for your target query and note which terms appear consistently that are absent from your page. These are semantic signals you're missing. Add them naturally — not as a keyword-stuffing exercise, but as proper integration of expert vocabulary. A content audit tool or AEOCrawler's semantic gap analysis can identify missing terms automatically.
Dimension 7: Conciseness (0–100) — weight 7%
What it measures: The information density of your content — how much extractable substance it delivers per word, and how little filler stands between the reader (or the AI engine) and the answer.
AI engines work within retrieval and synthesis budgets. When two candidate pages contain the same answer, the one that states it economically is easier to quote cleanly and more likely to survive the engine's summarization step intact. Conciseness is not about being short — a 3,000-word guide can score high — it is about every paragraph earning its place.
Signals that drive this score:
- Front-loaded substance: how much complete, extractable information lands in the first 150 words of the page body
- A direct-answer opening: first words that signal an authoritative statement ("The...", "Yes...", "According to...") rather than a wind-up
- Sentence economy in the opening: a handful of complete, declarative sentences rather than fragmented or meandering prose
- Substantive vocabulary: concrete, specific wording scores above vague filler
- Sufficient body content: density is only measurable when there is enough substance to measure — very thin pages are penalized
Filler, redundancy, and hedge words all hurt these signals indirectly: every padded sentence in your opening pushes real substance out of the measured window.
What a good score looks like: 70+ indicates dense, economical writing where answers are front-loaded. Below 55 usually means long wind-up intros, repetitive sections, or marketing copy padding out informational content.
Common fix: Cut the first paragraph of most drafts — it is usually preamble. State the answer first, then explain. Keep one claim per paragraph, and keep links and CTAs outside your direct answer blocks so the quotable text stays clean.
Dimension 8: Readability (0–100) — weight 6%
What it measures: How reliably your sentences can be parsed — by humans and by the language models that extract, summarize, and quote your content.
Complex, nested sentences are where extraction errors happen. When an AI engine quotes a clean 18-word declarative sentence, the citation survives intact. When it has to untangle a 45-word sentence with three subordinate clauses, it either paraphrases (losing your attribution strength) or skips your page for a clearer source.
Signals that drive this score:
- Average sentence length — 15–20 words is the extraction sweet spot; long, multi-clause sentences drag the score down fastest
- Vocabulary weight — average syllables per word; dense polysyllabic prose parses less reliably
- Balance in both directions — oversimplified prose is penalized too, because grade-school simplicity reads as low substance
AEOCrawler computes a Flesch Reading Ease value from sentence length and vocabulary weight, then maps it to the Readability score:
| Flesch Reading Ease | Readability score |
|---|---|
| 60–70 (standard) | 100 |
| 70–80 (fairly easy) | 90 |
| 50–60 (slightly difficult) | 80 |
| Over 80 (too simple) | 70 |
| 40–50 (difficult) | 60 |
| Under 40 (very difficult) | 40 |
The target band is the same for every content type — technical pages do not get a complexity allowance. If your documentation needs heavy terminology, explain it in short sentences rather than long ones.
What a good score looks like: 70+ indicates clear, parseable prose. Below 55 typically reveals academic-style writing, legalese, or translated content that was never edited for flow.
Common fix: Break long sentences at the conjunctions — sentence length is the single biggest lever in this score. Swap heavy words for lighter ones where nothing is lost. Define every technical term the first time it appears, in a short sentence of its own — this also feeds Semantic Coverage, since definitions are among the most-quoted content blocks in AI answers.
Dimension 9: Content Freshness (0–100) — weight 5%
What it measures: How recently your content was created or updated — and whether AI engines can detect that recency through machine-readable signals.
Content Freshness is the newest dimension in AEOCrawler's framework, and we're not aware of another AEO tool that scores it as a discrete, weighted algorithm. Research shows that pages updated within the last two months receive approximately 28% more AI citations than older content. AI engines treat date signals as trust indicators — stale content gets deprioritized in citation selection, even if the underlying information is still accurate.
This dimension is also the most actionable in the entire framework. Unlike building external authority (which depends on signals that accumulate over time), freshness can be fixed in minutes: update your content, add a dateModified schema tag, and display a visible "Last updated" line.
Signals that drive this score (4 detection methods):
- Schema.org
dateModified/datePublished— The most authoritative machine-readable freshness signal. AEOCrawler extracts ISO 8601 dates from Article, BlogPosting, and WebPage schema. - HTTP
Last-Modifiedheader — A server-level signal that tells AI crawlers when the document was last changed at the file level. - Open Graph
article:modified_time— Social sharing metadata that many AI engines also parse for recency signals. - In-text date patterns — Visible "Last updated: 2026-05-19" or "Published: April 2026" strings detected via pattern recognition in the page body.
Scoring curve:
| Content age | Score |
|---|---|
| 0–30 days | 100 |
| 31–90 days | 85 |
| 91–180 days | 65 |
| 181–365 days | 45 |
| Over 1 year | 25 |
| Unknown (no date detected) | 50 |
What a good score looks like: 85+ indicates content updated within the last 90 days with at least one machine-readable date signal. Below 50 means the content is either over a year old or has no detectable date signals at all — both of which reduce citation probability significantly.
Common fix: Add dateModified to your Article schema markup and keep it synchronized with actual content updates. Display a visible "Last updated" date on the page. Establish a content refresh cadence — quarterly for evergreen content, monthly for competitive topics. Even minor updates (adding a new statistic, updating a comparison table, revising an example) justify a fresh dateModified and improve this score.
Why it matters for proactive AEO: Content Freshness fits perfectly into AEOCrawler's pre-publication workflow. Before publishing, check that your schema includes dateModified, your page displays a visible date, and your content references current data. This is something you can verify and fix in your pre-publication workflow — no external dependencies, no waiting for backlinks or brand recognition.
How the 9 Dimensions Combine Into the AI Visibility Score
The 9 dimensions are not weighted equally. AEOCrawler applies a weighted composite model — the result is your AI Visibility Score, the single 0–100 number that summarizes AI citation readiness:
| Dimension | Weight | Rationale |
|---|---|---|
| Answer Extraction | 20% | Most directly influences whether content can be used in a generated answer |
| Citation Probability | 18% | Determines selection among competing candidates |
| Structured Data | 12% | Machine-readable structure is how AI engines understand and attribute content |
| Entity Authority | 12% | Entity clarity is foundational for AI knowledge graph attribution |
| Query Coverage | 10% | Broad intent coverage significantly increases citation frequency |
| Semantic Coverage | 10% | Vocabulary depth signals expertise to language model evaluators |
| Conciseness | 7% | Dense, economical content survives AI summarization intact |
| Readability | 6% | Clean sentences parse and quote reliably |
| Content Freshness | 5% | Pages updated within 2 months get ~28% more AI citations |
The weightings reflect a key insight: you can have perfect technical structure and still score poorly overall if your content doesn't contain extractable answers. Answer Extraction and Citation Probability are the highest-leverage dimensions for most pages. Content Freshness carries a lower weight because it is a modifier rather than a structural factor — but its impact compounds: a fresh page with strong Answer Extraction and Entity Authority will consistently out-cite an identical but stale page.
Score Tiers: What Each Range Means
| Score Range | Tier | AI Citation Behavior |
|---|---|---|
| 85–100 | Elite | Cited consistently; competitive in high-difficulty queries |
| 70–84 | Strong | Cited regularly; competitive in most queries |
| 55–69 | Moderate | Cited occasionally; loses to better-optimized competitors |
| 40–54 | Weak | Rarely cited; requires significant structural improvement |
| 0–39 | Invisible | Not cited; fundamental content or technical issues |
The practical benchmark for most content teams: aim for 70+ across all 9 dimensions before publishing. Pages below 70 on any dimension have a meaningful weakness that a better-optimized competing page will exploit.
Pages between 55–69 are in the most dangerous zone: they're good enough to feel optimized but weak enough to consistently lose citation competitions to slightly better content. This is where most "we're not getting cited despite good content" problems originate.
Proactive Scoring vs Post-Publish Analysis
The 9-dimension framework is most powerful when applied before publication. Once a page is live, underperforming on AI citation is hard to diagnose — you only know you're not getting cited, not why. The scoring framework tells you why, and it tells you before publication when fixing costs nothing.
This is the core distinction between proactive and reactive AEO. Reactive monitoring tells you your AI visibility is low. Proactive scoring tells you which of the 9 dimensions is weak and exactly what to change.
Content teams that integrate AEOCrawler into their pre-publication workflow — scoring drafts, improving weak dimensions, and publishing above 70+ — consistently see faster AI citation gains than teams that publish first and monitor second. If you want to understand why traditional SEO content often scores poorly on these dimensions, why your SEO content is invisible to ChatGPT explains the structural gap directly.
For definitions of the technical terms used in this framework, the AEO glossary covers all key vocabulary with context for why each term matters.
Start scoring your content before AI engines make the decision for you.
Score your content across all 9 dimensions before publishing →
Frequently Asked Questions
What are the 9 dimensions of AEO scoring?
AEOCrawler evaluates content across 9 dimensions: Answer Extraction (how easily AI can extract a citable answer), Citation Probability (selection likelihood among competing pages), Structured Data (schema and markup quality), Entity Authority (clarity and consistency of named entities), Query Coverage (breadth of intent addressed), Semantic Coverage (topical vocabulary depth), Conciseness (information density), Readability (sentence parseability), and Content Freshness (recency of creation or update, detected via 4 methods). Each dimension is scored 0–100 and combines into a composite AI Visibility Score.
What AEO score should I aim for before publishing?
Aim for 70 or above on every dimension before publishing. Pages below 50 on any dimension are unlikely to be cited by AI engines. Pages between 50–70 will be cited occasionally but will lose citation competitions to better-optimized competitors. Pages above 70 across all 9 dimensions are strong candidates for consistent AI citation. Pages above 85 are in the elite tier.
Which AEO dimension has the biggest impact on AI citation?
Answer Extraction carries the highest individual weight (20%) in AEOCrawler's composite scoring model, because it most directly determines whether AI engines can use your content in a generated answer. A page can be technically excellent and still be ignored if it contains no extractable answer block. Citation Probability (18%) is next, followed by Structured Data and Entity Authority at 12% each.
How is AEO scoring different from SEO scoring tools?
SEO scoring tools evaluate factors that influence ranking in a list of links: keyword density, backlink count, page speed, crawlability. AEO scoring evaluates factors that influence citation in AI-generated answers: answer clarity, entity consistency, schema structure, intent coverage, and information extractability. There is overlap — schema markup and crawlability matter for both — but the primary signals are different. Use both together for a complete optimization picture.
Can a page score well on SEO and poorly on AEO?
Yes, frequently. A page can rank #1 on Google for its target keyword while scoring 35/100 on AEO dimensions — meaning it captures traditional search traffic but receives no citations from ChatGPT, Perplexity, or Google AI Overviews. This happens when the page is well-optimized for keyword relevance but poorly structured for answer extraction, lacks schema markup, or doesn't address enough query intent variants. Why SEO content is often invisible to AI engines is explained in more detail in our SEO vs AEO comparison.
How often should I rescore my content?
Rescore content when you make significant changes (adding new sections, updating statistics, revising the answer block). For evergreen pages, a quarterly AEO audit is a reasonable cadence. For high-competition queries where competitors are actively optimizing, monthly scoring may be appropriate. AEOCrawler's monitoring layer tracks dimension changes over time so you can detect when competitors' improvements are affecting your relative citation position.
Does every page need to hit 70+ on all 9 dimensions?
Practically speaking, yes — for any page where you want consistent AI citation. For supporting pages (deep technical documentation, internal reference pages, narrow FAQs) a lower score is acceptable because they're not the primary citation candidates. Focus your optimization energy on your pillar content, category pages, and any page you want cited for high-value queries.
What is the hardest dimension to improve quickly?
Citation Probability is the slowest to move because part of it depends on external authority signals — backlink quality, brand entity recognition, press coverage — that accumulate over time. New sites should maximize the dimensions they fully control first: Answer Extraction, Structured Data, Conciseness, Readability, and Content Freshness can all be improved within hours. Answer Extraction is typically the fastest win: adding a direct answer block and question-format headings can move the score 15–25 points in a single editing pass.
