How we rate

How a check is judged

Every check ends in one of four results. Only something the scanner actually saw is called a problem.

Four results

passed
Enough evidence that the site does this.
issue found
A clear problem was observed. The report names the page and the evidence.
could not verify
The scanner could not finish the check: a timeout, a page that refused or limited access, a login wall, a request budget that ran out, a robots.txt that keeps the scanner out, or a method it does not have yet (such as running JavaScript). It is not counted as a problem, and it does not mean the site lacks the feature.
not applicable
The check does not apply to this page or site. For example, breadcrumbs on a home page, or a write API on a site its owner declared a content site.

Three dimensions

Results are summarised in three dimensions. Each check belongs to one of them.

Content access

Can AI read it?

AI crawlers and assistants may be unable to read or trust this content.

Content discovery

Can AI find it?

Pages may be missed, indexed under the wrong address, or never surface in AI answers.

Task use

Can AI act on it?

Agents cannot reliably search, sign up or call your API on a user's behalf.

  • Issues found
  • Could not verify
  • Verified
  • Not applicable

A dimension shows “Issues found” if any applicable check has a problem. Otherwise it shows “Could not verify” if anything was left unfinished, and “Verified” only when every applicable check passed. “Not applicable” appears only when every check clearly does not apply, and a site that has not been scanned yet is never shown as not applicable. The counts always sit beside the status, so a problem never hides how much could not be checked.

Which pages are checked

The scanner checks the home page and up to 5 inner pages (product, pricing, docs, help, contact or about), picked from your navigation and sitemap. Checks about a page run on every sampled page. Checks about the whole site (robots.txt, sitemap, APIs, MCP) run once.

A check's result is its worst result across the pages, and the report says which pages. Findings come from this sample, not from every page of the site. Pages your robots.txt does not allow are never picked, and pages the scan's time or request budget did not reach are shown as not checked.

How serious is an issue

Each issue is labelled “Blocks AI” (it stops AI from reading or using the site), “Improvement”, or “Best practice” (not a barrier). The standard sets these labels; the AI report does not.

Site type

When you submit a site you can say what kind it is: content, online product, e-commerce, developer platform, or not sure. It only decides which optional capabilities are expected. A content site is not asked for a write API. “Not sure” assumes nothing, and anything the scanner actually observed is never overridden.