AI SEARCH READINESS FRAMEWORK
Is Your Site Ready to Be Cited by AI Search?
AI search readiness is how ready a website is to be found, understood and cited by ChatGPT, Perplexity, Gemini and Google AI Overviews. This is the 10-signal framework we use to score it, and you can run it on your own site.
A methodology you can apply, not a benchmark you have to wait for. The figures here are scoring definitions and illustrative examples, not published statistics.
IN SHORT
What Is AI Search Readiness?
AI search readiness is how ready a website is to be found, understood and cited by AI search engines such as ChatGPT, Perplexity, Gemini and Google AI Overviews. This framework scores readiness across 10 signals in three groups: whether AI crawlers can reach the content, whether machines can parse its structure and meaning, and whether the content is substantive enough to be worth citing. Each signal is scored 0 to 10 for a total out of 100. The framework is designed to be run on any site by anyone. The figures below are scoring definitions and illustrative examples, not published benchmark results.
10
Signals scored
3
Signal groups
100
Points possible
The numbers above describe the framework's structure. They aren't measured results.
THE FRAMEWORK
Which 10 Signals Determine AI Search Readiness?
Ten signals in three groups. Access decides whether AI crawlers can read the content. Machine readability decides whether they can parse it. Substance decides whether it's worth citing. Each signal scores 0 to 10, for a total out of 100.
Access
signals 1 to 3Can AI crawlers reach and read the content at all? If they cannot, nothing else matters.
AI crawler policy in robots.txt
0 to 10 points
Explicit rules for GPTBot, PerplexityBot, ClaudeBot, Google-Extended and other AI crawlers. This decides whether AI engines are even allowed to read the content. Block them and citation is impossible.
Scoring: 0 = blocks all AI crawlers, 3 = mixed signals, 7 = allows most, 10 = deliberate, documented allow policy
Each crawler publishes its own user-agent. OpenAI, Google and others document how to control access via robots.txt.
llms.txt presence
0 to 10 points
A dedicated /llms.txt file giving language models a curated, plain-text map of the most important pages. An emerging convention rather than a ratified standard, so treat it as a bonus signal, not a requirement.
Scoring: 0 = absent, 5 = present but thin, 10 = present and points to the pages that matter
Proposed convention (llmstxt.org). Not yet an official standard, so weight it lightly.
Feed and freshness discoverability
0 to 10 points
A discoverable RSS or Atom feed, plus visible published and updated dates. AI systems weight recency when choosing which source to cite, and a feed lets crawlers spot new content without re-scanning the whole site.
Scoring: 0 = no feed and no dates, 5 = one of the two, 10 = autodiscoverable feed and clear dates
Freshness as a ranking factor is documented in Google Search guidance and observed across AI answer engines.
Machine readability
signals 4 to 7Can a machine parse the structure and meaning of the page? The accessibility tree is how screen readers and AI crawlers both read a page, by structure and meaning rather than pixels.
Organization schema and entity identity
0 to 10 points
Organization JSON-LD declaring name, logo, contact and sameAs links to authoritative profiles. This is what ties a site to a recognised entity in the knowledge graph, so an AI engine knows who it is quoting.
Scoring: 0 = absent, 5 = partial, 10 = complete with sameAs to authoritative profiles
Organization structured data is documented by Google. Entity linking via sameAs is a recognised knowledge-graph signal.
Article or BlogPosting schema
0 to 10 points
Article or BlogPosting JSON-LD on published content, carrying headline, author, date published and date modified. It tells a crawler what the page is, who wrote it and when, without guessing from the layout.
Scoring: 0 = no schema, 5 = present but missing required fields, 10 = complete and valid
Article structured data and its required fields are documented by Google Search Central.
FAQ and question-shaped markup
0 to 10 points
FAQPage structured data, and headings phrased as the questions a reader would actually ask. AI engines lift question-and-answer pairs directly into responses, so a clean FAQ block is one of the most citable structures on a page.
Scoring: 0 = absent, 5 = questions in prose but no schema, 10 = valid FAQPage schema on question-shaped content
FAQPage structured data is documented by Google. Question-phrased content matches how generative engines retrieve and quote sources.
Accessibility tree and layout stability
0 to 10 points
A clean accessibility tree (semantic headings, landmarks, alt text) and a stable layout. This is the machine-readability core: a page that is hard for a screen reader to parse is hard for an AI crawler to parse. Sucks for screen readers, sucks for language models.
Scoring: 0 = broken structure or unstable layout, 5 = partial semantics, 10 = clean accessibility tree and stable layout
The accessibility tree is how assistive tech and automated agents both read a page. Layout stability (CLS) is a measured Core Web Vital.
Substance
signals 8 to 10Is the content actually worth citing? Access and markup get a page read. Substance is what earns the citation.
Named author and expertise signals
0 to 10 points
An identifiable human author with a real bio and credentials, linked from the content. Attribution is a trust signal, and generative engines lean toward sources that show who is speaking and why they would know.
Scoring: 0 = no author, 5 = generic or corporate byline, 10 = named author with a linked bio and credentials
Author and expertise signals sit under Google E-E-A-T guidance and are echoed in how AI engines weight source trust.
Content depth and specificity
0 to 10 points
Substantive, specific, well-sourced content rather than thin promotional copy. Generative engines reward pages that answer the question completely, cite their own sources, and use concrete figures over vague claims.
Scoring: 0 = thin or promotional only, 5 = moderate depth, 10 = specific, sourced, genuinely useful
The GEO study (arXiv 2311.09735) found that adding citations, quotations and statistics measurably raised how often generative engines surfaced a source.
Direct-answer structure
0 to 10 points
The page states its answer near the top, before the supporting detail. AI engines extract self-contained answer spans, so content that leads with the answer is easier to quote cleanly than content that buries it.
Scoring: 0 = no clear answer span, 5 = answer present but buried, 10 = clear answer-first summary near the top
Answer-first structure matches how retrieval-augmented engines select and quote passages. This page follows its own rule, above.
Signals 1 to 9 can be checked with a browser, a structured-data validator and a page-speed tool. Signal 10 needs a human read. Total possible score: 100 points. These are scoring definitions, not measured values for any site.
READING THE SCORE
What Does an AI Search Readiness Score Mean?
Totals map to five bands. Each band is a different level of preparedness for AI-powered search. The band definitions are part of the framework, not a distribution measured across real sites.
AI search readiness bands (scoring rubric)
| Score | Band | What it means |
|---|---|---|
| 0 to 20 | Invisible | AI crawlers are blocked, the markup is missing, or both. The site is effectively unreadable to AI search and will not be cited. |
| 21 to 40 | Basic | Some signals exist but coverage is patchy. A crawler can reach the site but struggles to parse or trust most of it. |
| 41 to 60 | Developing | The foundation is in place. Core schema is deployed and content is published, but AI-specific signals and depth are uneven. |
| 61 to 80 | Advanced | Most signals are active across access, readability and substance. The site is positioned to be cited across its core topics. |
| 81 to 100 | Engineering-grade | Readiness is built into how the site is run. Every signal is active, content is authoritative, and the site is designed to be cited. |
Illustrative band definitions. Not a measured distribution of any company set.
WORKED EXAMPLE
How Does a Score Come Together?
A made-up example to show how the three groups add up to a total and point at the fastest fix. This is illustrative. It isn't an audit of any real company.
Illustrative scorecard for a fictional site
| Signal group | Example score | Why |
|---|---|---|
| Access (signals 1 to 3) | 18 / 30 | Crawlers allowed and dates visible, but no feed and no llms.txt |
| Machine readability (4 to 7) | 22 / 40 | Organization and Article schema present, FAQ schema missing, accessibility tree clean |
| Substance (8 to 10) | 15 / 30 | Named authors, decent depth, but answers buried below the fold |
| Total | 55 / 100 | Band: Developing. Fastest lift: add FAQ schema and move answers up. |
Illustrative only. Invented figures shown to demonstrate the method, not a real measurement.
USING THE FRAMEWORK
Who Uses an AI Search Readiness Score, and How?
For CMOs
Run the framework on your own domain and your closest competitors. The gaps show where you're invisible to AI search and which signals give the fastest lift. It's a repeatable audit, not a one-off report.
For PE partners
Score a portfolio the same way and compare like for like. A low readiness score flags a company that the next generation of search can't see. A high one flags a durable organic moat worth protecting.
For growth leaders
Use the 10 signals as a quarterly checklist. Score, fix the lowest group first, re-score. Because most signals are technical and templated, one fix can lift every page at once.
WHAT THIS IS BUILT ON
Where Does the Framework Come From?
The signals draw on public research and platform documentation, not invented benchmarks. The core sources:
GEO: Generative Engine Optimization
Aggarwal et al., arXiv 2311.09735. Found that adding citations, quotations and statistics measurably raised how often generative engines surfaced a source. Underpins the Substance group.
Google structured-data documentation
Google Search Central defines Organization, Article and FAQPage structured data and their required fields. Underpins the Machine readability group.
The accessibility tree as a machine-readability signal
The accessibility tree is how screen readers and automated agents both read a page, by structure and meaning. A page that's hard for assistive tech to parse is hard for an AI crawler to parse.
Pew Research on AI Overviews
Pew Research Center reporting on the reach of Google AI Overviews across US searches. Grounds why readiness for AI-generated answers now matters, without importing any specific figure as our own claim.
FREQUENTLY ASKED QUESTIONS
AI Search Readiness, Answered
What is AI search readiness?
AI search readiness is how ready a website is to be found, understood and cited by AI search engines such as ChatGPT, Perplexity, Gemini and Google AI Overviews. It covers three things: whether AI crawlers can reach the content, whether machines can parse its structure and meaning, and whether the content is substantive enough to be worth quoting. This framework scores each of those across 10 signals for a total out of 100.
How do you score AI search readiness?
Score each of the 10 signals from 0 to 10 using the rubric on this page, then add them up for a total out of 100. The signals split into three groups: Access (crawler policy, llms.txt, feeds and freshness), Machine readability (Organization, Article and FAQ schema, plus the accessibility tree), and Substance (named authors, content depth, and answer-first structure). Most of the signals can be checked with a browser, a structured-data validator and a page-speed tool in an afternoon.
How is AI search readiness different from SEO?
Traditional SEO optimises for ranked links on a results page. AI search readiness optimises for being cited inside a generated answer. The overlap is real, since good structured data and depth help both, but the target differs. AI engines quote self-contained answer spans, weight structured data and named authorship heavily, and read a page through its accessibility tree rather than its visual layout. A site can rank well and still be hard for an AI engine to quote.
Do the numbers on this page come from a benchmark study?
No. This page is a framework, not a published benchmark. The point totals are scoring definitions, and the worked example is illustrative to show how a score is built, not a real audit of any company. It draws on public sources including the GEO study (arXiv 2311.09735) on what makes content visible to generative engines, the Google structured-data documentation, and Pew Research on the reach of Google AI Overviews.
PUT A NUMBER ON IT
Want Your Site Scored Against This Framework?
We'll run the 10 signals across your domain and your competitors, hand you the scorecard, and rank the fixes by lift. No benchmark to wait for.
Score My SiteSTAY INFORMED
Get the Full Scoring Guide
Drop your email and we'll send you the full guide: how to check each signal, the tools that do it, and a scorecard template you can reuse every quarter.
Get Notified When Published
No spam. The scoring guide and template, once.