How the AEO Score works

Why citation-readiness matters

AI answer engines built on large language models increasingly answer questions directly, so classic search engine optimization no longer guarantees you are seen. The score measures readiness for that shift. The data, each figure linked to its source:

• 800M+More than 800 million people use ChatGPT every week as of October 2025. (TechCrunch)
• ~2BGoogle's AI Overviews reached roughly 2 billion monthly users by mid-2025, across 200 countries. (Alphabet / TechCrunch)
• 750MGoogle's Gemini app surpassed 750 million monthly active users as of Q4 2025. (Alphabet / TechCrunch)
• 8% vs 15%When Google shows an AI summary, people click a traditional search link only 8% of the time, versus 15% when there is no summary (per Pew, 2025; Google disputes the size of the effect). (Pew Research)
• +693%U.S. retail traffic from generative-AI sources grew roughly 693% year over year over the 2025 holiday season (Adobe Analytics). (Adobe Analytics)
• -58%AI Overviews are linked to a roughly 58% lower click-through rate for the #1 organic result (Ahrefs, 2026; a correlation, not proven cause). (Ahrefs)
• ~1%AI referrals still account for only about 1% of all website traffic as of 2025 - fast-growing, but small. (Digiday / Similarweb)
• 0.66 vs 0.22In a 2025 Ahrefs study of 75,000 brands, unlinked brand mentions correlated with AI Overview visibility at 0.66, versus just 0.22 for backlinks - mentions tracked visibility more closely than links. (Ahrefs)
• ~12%For third-party AI assistants like ChatGPT and Gemini, only about 12% of cited URLs rank in Google's top 10 for the original query (Ahrefs, 2025). (Ahrefs)

Deterministic where we can, judged where we must

Most of the score runs as plain code. We parse the page's HTML, schema, robots rules, and dates and apply the same checks every time, so the structural score is reproducible - the same page always scores the same, with no randomness and no token cost. A few dimensions need judgment (answerability, factual density, topical coverage), so they are graded by an AI model at temperature 0 with structured output. The free score here runs the deterministic checks; the AI-judged deep analysis is part of the full product.

• Answerability & Q&A structure

  Hybrid

  Whether the page is shaped like answers: question-style headings, answer-first paragraphs, TL;DR and summary blocks, FAQ sections, and self-contained passages a model can lift verbatim.

  Why it matters: AI engines retrieve and quote discrete passages. Content that answers a question directly, near the top of a section, is far more likely to be excerpted into an AI answer.

  Code-checked structure blended with an AI judgment.

• Structured data / schema

  Code-checked

  Presence, validity, and relevance of JSON-LD / schema.org markup (FAQPage, HowTo, Article, Product, Organization, Author, Review, Breadcrumb).

  Why it matters: Schema gives engines an explicit, machine-readable map of your content's meaning and key facts, making it easier to parse, trust, and cite.

  Run entirely in code. Reproducible, no token cost.

• Extractable claims & factual density

  AI-judged

  Count and quality of discrete, verifiable, self-contained factual statements: stats with sources, definitions, and specific numbers and dates.

  Why it matters: Engines prefer to cite concrete, checkable facts. Dense, specific, self-contained claims are the raw material of AI answers.

  Graded by an AI model at temperature 0 with structured output.

• Citations, sourcing & E-E-A-T

  Hybrid

  Outbound links to authoritative sources, a credentialed author with a bio, clear publish/updated dates, and original data or research.

  Why it matters: Experience, expertise, authoritativeness and trust signals make engines more confident citing you over an anonymous, unsourced page.

  Code-checked structure blended with an AI judgment.

• Semantic clarity & chunkability

  Hybrid

  Readability, short paragraphs, a clean heading hierarchy, low ambiguity, and sections that each survive retrieval chunking on their own.

  Why it matters: Retrieval splits pages into chunks. If a chunk only makes sense with the rest of the page, it won't be retrieved or quoted accurately.

  Code-checked structure blended with an AI judgment.

• Topical coverage & query fan-out match

  AI-judged

  How fully the page covers the cluster of sub-questions an engine fans a topic into - breadth, depth, related entities, and comparisons.

  Why it matters: Engines decompose a query into many sub-queries. Pages that answer the whole cluster get cited across more of the generated answer.

  Graded by an AI model at temperature 0 with structured output.

• Freshness & maintenance

  Code-checked

  Published and updated dates, recency of the facts cited, and explicit 'last updated' signals.

  Why it matters: Engines bias toward current information. Visibly maintained content is more likely to be trusted and cited for time-sensitive queries.

  Run entirely in code. Reproducible, no token cost.

• AI-crawler accessibility

  Code-checked

  Whether robots.txt allows AI crawlers (GPTBot, PerplexityBot, Google-Extended, ClaudeBot, CCBot, etc.), whether content is in server-rendered HTML, and load speed. (llms.txt is detected but not scored - no major engine is confirmed to consume it.)

  Why it matters: If AI bots are blocked or your content only exists after JavaScript runs, engines may never read it - and uncited content can't be cited.

  Run entirely in code. Reproducible, no token cost.

• Entity & brand clarity

  Hybrid

  Clear entity definitions, consistent brand naming, a clear About, and alignment with knowledge graphs (Wikipedia / Wikidata).

  Why it matters: Engines reason over entities. When your brand is an unambiguous, well-defined entity, it's easier to associate facts with you and cite you by name.

  Code-checked structure blended with an AI judgment.

We score the page you give us, in context

Citation-readiness is not one-size-fits-all. An editorial article genuinely needs a named author and a publish date; a brand homepage or a product page does not. We detect the page type and apply the right expectations, so a homepage is not marked down for missing a byline it was never meant to have, and a product page is asked for Product schema rather than Article schema.

What we do not do

We do not claim to see the real prompts people type into AI engines - nobody can. We probe representative prompts, measure AI-referral traffic, and mine question proxies. We are not affiliated with any engine. When the tool cannot fully read a page (for example, one that renders its content with JavaScript), we tell you rather than guess.

Engine coverage, stated plainly

The tracker probes engines through their official APIs and licensed data, labeled per engine, and we publish exactly what is live versus planned. We never scrape the ChatGPT consumer interface or any engine's logged-in UI: it violates their terms, it breaks without warning, and results from it cannot be honestly reproduced. If an engine is not live yet, we say so here rather than sell it as if it were.

This table is updated as engines come online. Every plan includes every live engine at no per-engine surcharge, and results are always labeled with the engine and method they came from.

Common questions

What is an AEO Score?

It is a 0 to 100 estimate of how citation-ready a web page is for AI answer engines like ChatGPT, Perplexity, and Gemini, built from nine weighted dimensions covering structure, schema, sourcing, freshness, crawlability, and entity clarity.

Is the score reproducible?

The deterministic dimensions are pure functions of the page, so the same page always gets the same structural score. The AI-judged dimensions run at temperature 0 with structured output to keep them stable.

Do you see the real prompts people ask AI?

No. Nobody can see the real prompts users type into AI engines. We probe representative prompts, measure AI-referral traffic, and mine question proxies. We never claim to capture real user prompts.