The accessibility audit

The audit loads your pages in a real browser on Cloudflare's infrastructure, runs axe-core against each, and records every finding as evidence. It is the version to use for a remediation project or an insurance questionnaire.

Where

Web: SEO, the domain, Accessibility, Run audit; or the Website module.
iOS: Website, Accessibility audit: the run/quote card, the score ring, impact counts, issues grouped by rule with the WCAG criterion, a rule detail with every affected element, and manual review decisions. The evidence export is viewed in the app.
API/MCP: GET /api/accessibility-audit/quote?domainId=, POST /api/accessibility-audit, GET /api/accessibility-audit?domainId=, GET /api/accessibility-audit/:id, POST …/:id/manual-review, POST …/:id/complete, GET …/:id/export; tools quote_accessibility_audit, start_accessibility_audit (domain_id, request_key, optional max_pages, viewport), list_accessibility_audits, get_accessibility_audit.

Cost and limits

	Value
Price	5 credits per audit, reserved before the job is queued and charged when it completes
Pages	Up to 10 same-origin pages, discovered from the homepage's links; the discovery looks at up to three times that many URLs to find real, canonical pages
Viewport	1280x900 by default; another can be requested
Rate	3 audit starts per hour per account
Refund	If the job cannot be queued or the scanner's retries run out, the reserved credits are returned exactly once

The audit is on by default in the add-domain flow, and its cost is stated there because it is the one add-time step that spends.

What comes back

For each page: the axe findings with rule, impact, WCAG criterion, the selector and HTML of every affected element, and the fix guidance. Rolled up: counts by impact, by WCAG criterion, and the regression against the previous audit of the same domain.

Manual review

Sixteen WCAG criteria cannot be judged by a machine (captions and audio description, keyboard operability, timing, focus order and visibility, on-focus changes, error identification and labels, name/role/value for custom widgets, among others). The audit lists them; POST …/:id/manual-review records one decision at a time with the reviewer's note; POST …/:id/complete closes the audit once every manual criterion has a recorded decision. A completed audit is a full record: automated evidence plus a person's judgements.

Export

GET /api/accessibility-audit/:id/export returns the whole audit as JSON: pages, findings, manual decisions and the evidence event log, with the axe and browser versions used. It is the document to attach to a remediation ticket or a compliance file.

The honest limit

Automated findings are evidence only and never establish WCAG conformance. A clean axe run means axe found nothing, and the manual criteria still stand. The audit says this on its own page.

Common questions

Can I audit more than 10 pages? Not in one audit. Run several with different starting points, or use the crawl's static accessibility checks for coverage and the audit for depth.

The audit says 0 pages scanned. The homepage did not load in the browser (a bot challenge, a redirect loop, a timeout). Credits are returned. Check the site allows a headless browser from Cloudflare.