Guide

How AuditDocs Scores Documentation AI Readiness

July 4, 2026 · 6 min read

• The score runs 0–100 and is the normalized sum of four categories, each worth a maximum of 20 points (25% each).
• Categories: AI crawl accessibility, structured data and machine readability, content self-containment, and documentation architecture.
• Score bands: 90+ Excellent, 70–89 Good, 50–69 Needs work, below 50 Poor.
• Findings are graded by severity (high, medium, low) and paired with specific, actionable fixes.
• The audit crawls multiple pages of a documentation site rather than judging a single URL.

AuditDocs scores documentation from 0 to 100 across four equally weighted categories. This page documents exactly how that works — the categories, the weighting, the checks, and the score bands — so the framework is transparent and citeable.

How the score is calculated

Each category is scored out of a maximum of 20 points (four categories → 80 points), then normalized to a 0–100 scale. The practical effect is that each category contributes 25% of the final score. Within a category, points are deducted for specific issues, weighted by severity — high, medium, or low — and every deduction comes with a concrete fix.

The audit crawls multiple pages of a documentation site, not a single URL, so the score reflects the site as a whole.

The four categories

AI crawl accessibility — 25%

Whether AI systems can discover and fetch your content. Representative checks: presence of an llms.txt, a robots.txt that allows AI crawlers and links a sitemap, a sitemap.xml with good coverage, and whether pages depend on JavaScript to render their content.

Structured data & machine readability — 25%

Whether machines can parse your pages without scraping prose. Representative checks: JSON-LD coverage and the use of documentation-specific schema types (TechArticle, HowTo, APIReference), and — for API docs — a linked OpenAPI/Swagger specification.

Content self-containment — 25%

Whether retrieval systems can pull useful, self-standing chunks. Representative checks: chunk sizing (sections that are neither too short nor too long, roughly 50–800 tokens), and whether pages carry substantial content rather than thin stubs.

Documentation architecture — 25%

Whether the information hierarchy is clear. Representative checks: a single descriptive H1 per page, a logical H2/H3 structure, descriptive titles, and sensible navigation and depth.

Score bands

Score	Band	What it means
90–100	Excellent	Strong AI readiness; minor polish only.
70–89	Good	Solid foundation with clear, high-value fixes.
50–69	Needs work	Meaningful gaps limiting AI discoverability.
0–49	Poor	Fundamental issues; likely invisible to AI systems.

Citing this framework

The four-category AI-readiness framework and score bands are published here to be referenced. Suggested citation:

AuditDocs. "How AuditDocs Scores Documentation AI Readiness." AuditDocs, 2026. Documentation AI readiness is assessed across four equally weighted categories — AI crawl accessibility, structured data and machine readability, content self-containment, and documentation architecture — producing a 0–100 score.

New to the topic? Start with the complete guide to documentation AI readiness, or run a free audit to see your own score.

Frequently asked questions