Fix any AI readability gap.
AI readability is whether ChatGPT, Perplexity, Copilot, and other AI answer engines can crawl, parse, and cite your site. This is a plain-English guide to the 18 checks AccessKnight scores it on and exactly how to fix each one.
Fixing accessibility too? See the WCAG rule library — the Shared Core pillar below overlaps with it directly.
Three pillars, 100 points.
The Machine-Readability Score AccessKnight shows on every scan breaks down into three pillars. Each row below is what it checks and how many rule guides exist for it today.
| Pillar | Points | What it checks | Rules |
|---|---|---|---|
| AI Crawler Access | 30 pts | Whether the crawlers that feed ChatGPT search, Bing, and other AI answer engines can reach the page at all. Gating — a critical failure here caps the whole score. | 3 of 8 |
| Shared Core | 40 pts | The accessibility ⇄ AI-readability overlap: the same semantic structure that helps a screen reader also helps a retrieval agent parse the page. | 4 of 5 |
| Extractability | 30 pts | Whether an answer engine can lift a clean, self-contained answer from the page to cite, from question-phrased headings to comparable data in tables. | 3 of 5 |
Every check, in plain English.
AI Crawler Access
3 rulesOAI-SearchBot is blocked from crawling
OAI-SearchBot must be allowed in robots.txt — blocking it removes the page from ChatGPT search, even if GPTBot already crawled it for training.
Bingbot is blocked from crawling
Bingbot must be allowed in robots.txt — OpenAI names Bing as one of the third-party search providers ChatGPT search sends queries to, so a bingbot block can close that path as well as Bing Search and Microsoft Copilot.
Page carries a noindex directive
The page must not carry a noindex directive in the meta robots tag or the X-Robots-Tag header — a noindex drops the page from every index, so no AI surface can retrieve or cite it regardless of robots.txt.
Extractability
3 rulesHeadings are not phrased as questions
Headings phrased as literal questions match the way people and machines phrase queries — credit is awarded only for genuine interrogatives, not headings that could merely be reworded as one.
No direct answer in the first sentences under a heading
Each section should answer its heading in the first one or two sentences, before the supporting detail, so an engine can lift a self-contained answer.
No explicit definitional sentences
An explicit "X is a …" sentence is what an engine extracts when someone asks what something is.
Guides backed by the score.
Every guide documents a check AccessKnight actually runs — the same checks, point values, and plain-English fixes you get in a scored Machine-Readability report.
Is your site
AI readable?
Scan any page against all 18 AI-readability checks and get a scored report with fixes — free, no credit card.