AI Readiness

AI Readiness measures how well AI systems can understand, retrieve, cite, and act on your content. It is measurable, version-pinned, and distinct from SEO. The same signals it measures are the ones AI agents like ChatGPT, Claude, and Perplexity rely on when deciding what to cite.

The short definition

AI Readiness is not a metaphor. It is an AI Readiness Score, between 0 and 100, against a published rubric that defines and measures all four properties named above. The rubric is version-pinned, opinionated, and deterministic: the same content scored against the same rubric returns the same score, by design.

The dimension being measured is real and independent of the dimension humans evaluate. A reader visiting a documentation page sees prose, headings, code blocks, and design. An AI agent visiting the same page sees a parsed structure: schema markup, semantic HTML elements, agent-metadata files like llms.txt, and the raw HTML response before JavaScript executes. Both consumers can find the same page useful. Both can find it useless. The two evaluations are independent, and a site can score well on one and poorly on the other.

AI Readiness focuses on the agent's evaluation. It measures whether the structural signals an agent depends on are present, correct, and consistent across the site. It does not measure whether the prose is well-written or whether the page is beautiful — those are real properties of content, but they are not what determines whether an AI agent can extract a clean answer to cite.

The canonical definition, the rubric in detail, and the determinism property all live at /methodology.

What AI Readiness measures

The rubric carves the AI-agent loop into 8 categories. Each measures a distinct structural dimension; each is weighted; the 8 sum to 100.

  1. AI Crawlability & Access. Whether AI agents can reach the site at all. Robots-txt policy for AI bots, HTTP redirects, SSL configuration, response codes. The gate before everything else — a site agents cannot reach is a site agents cannot read.
  2. Documentation Patterns. Whether documentation is marked up in the formats AI agents recognize as documentation. TechArticle and APIReference schema, code-block markup that survives extraction, and the agent-metadata files (llms.txt, agents.md, .well-known/mcp.json) that signal a site is intentionally structured for agents. See /docs/documentation-patterns for the deep dive.
  3. Structured Data. Schema.org markup beyond the docs-specific types — Article, FAQPage, Organization, BreadcrumbList. Required fields filled correctly. Schema breadth across page types. JSON-LD that validates.
  4. AEO Readiness. Whether content is shaped for direct answers. Question-style headings, concise answer formatting, FAQ sections, structured Q&A patterns. Content shaped to be quoted, not just read.
  5. Content Structure. Whether the HTML uses semantic structure cleanly. Heading hierarchy that doesn't skip levels. Paragraph length that survives extraction. Lists used as lists. Semantic HTML5 elements (article, section, nav) used correctly.
  6. SSR & AI Rendering. Whether the content an agent receives in raw HTML matches what a user sees after JavaScript executes. Critical content in initial HTML, not deferred to a client-side render. Schema in raw HTML, not injected post-load.
  7. Site Architecture / Navigation. Whether an agent can traverse the site logically. Sitemap presence and depth. BreadcrumbList schema consistency. Internal-link density. Navigation menu present in raw HTML. URL hierarchy that reflects content hierarchy.
  8. Title & Identity. Whether each page identifies itself to a crawler. Page title quality. Canonical URLs. Organization schema. Meta description quality. The basic-hygiene category every page should pass without effort.

The canonical weights, the hard caps, and the per-category scoring details live at /methodology.

Why it differs from SEO

SEO and AI Readiness measure different things because they serve different consumers.

SEO scores content for a human ranking algorithm — primarily Google's — that decides which page deserves the click when a user searches for a query. Its signals reflect what predicts a satisfied human click-through: backlink quality, dwell time, Core Web Vitals, keyword salience, page authority, time-on-site. The downstream consumer is a human reading prose.

AI Readiness scores content for an LLM-powered agent that decides which page deserves the citation when a user asks an AI assistant a question. Its signals reflect what predicts a clean extraction: schema markup that names what the page is, semantic HTML that survives the parse, agent-metadata files that disclose what a site contains, code blocks that aren't decorated past recognition, and content visible in raw HTML rather than dependent on a JavaScript render. The downstream consumer is an agent extracting a snippet to cite.

The two scores are independent. A beautifully-designed marketing site with no schema markup, JavaScript-injected content, and no llms.txt can rank well on Google and score badly on AI Readiness — humans love it, agents miss it. The reverse is also routine: an unstyled MDN-shaped reference page with TechArticle schema everywhere, raw-HTML content, and version-pinned dates may rank poorly on Google search but get cited cleanly by every major AI assistant.

Both scores can be high; both can be low; one can be high while the other is low. Optimizing for one does not optimize for the other, and a site that needs both consumers needs to measure both dimensions separately. For the deep comparison — what each scores, where they overlap, where they diverge — see /docs/aeo-vs-seo.

Why it matters now

Developer-information-seeking behavior is shifting. A growing share of "how do I…" and "what is…" questions get asked of Claude, ChatGPT, or Perplexity before they get asked of Google. The agent reads documentation pages, extracts an answer, and cites the source. The cited source accumulates distribution: downstream readers — both human and agent — encounter it more often because it was chosen.

A site invisible to AI agents is invisible to that funnel. Not because anyone has decided the site is bad. Because the agent could not parse it cleanly — the likely outcome is that a competitor's page gets cited instead. The penalty is silent and structural: no error message, no rank drop. The structural signals AI agents read are accumulating; whether the citation-share shift shows up in the dashboards most teams watch is a separate question — most signal flows have lag.

The mechanics of the loop are mostly mechanical. The agent fetches HTML, parses structured data, extracts a snippet that matches the user's question, and emits a citation URL. Each step has a way to fail. Schema absent, no structured handle. Critical content in JavaScript, empty raw HTML. robots.txt blocking AI bots, no fetch at all. llms.txt absent, no top-level signal of what the site contains. Each failure is recoverable, and each is measurable. For the full mechanical loop and per-step failure modes, see /docs/how-ai-agents-read-your-docs.

In the scans we've run, most public developer documentation sites fall below 50 — not because their content is poor but because their structural signals were optimized for a different consumer. The fix is rarely a rewrite; it is a markup pass, an llms.txt, and a schema audit.

What you can do today

Three concrete steps:

  1. Run a free Lightning Scan of your site. It returns a single-page AI Readiness Score against the canonical rubric in roughly 30 seconds. The Lightning Scan covers a subset of the full Docs Readiness Audit's checks — single-page rather than multi-page — but the Score it returns is comparable to the full Docs Readiness Audit's Score and uses the same rubric.
  2. Read /methodology. It is the canonical definition of what AI Readiness measures, the eight-category rubric in full, and the determinism property explained.
  3. Read the per-topic articles in /docs. They cover specific structural moves — bot-control configuration, schema patterns, agent-metadata file formats — at depths the introduction here cannot reach.

The articles in /docs exist to make AI Readiness an addressable property. Each one explains a single decision, in a single sitting, with examples that copy-paste cleanly.

How Obaron measures it

This article is itself an example of high-AI-Readiness content.

The page wraps in TechArticle JSON-LD with explicit author attribution and an editable published-and-modified date pair. The breadcrumb hierarchy is schema-marked. The first paragraph standalone-answers the H1's implicit question, sized to the ~50-word citation window AI agents typically extract. Section structure is semantic. Internal links point at canonical URLs. Code blocks, when they appear, survive a copy-paste round trip without decoration loss.

The meta-point is intentional. Obaron's audit measures AI Readiness; Obaron's documentation should demonstrate it. If the rubric we publish does not score our own pages well, the rubric is the wrong rubric. The site is the live counterfactual.

The rubric this page is scored against, including the canonical definitions, weights, and hard caps, lives at /methodology.