Find the accessibility problems on one page, in seconds.

a11yscan fetches a public page and runs 18 deterministic WCAG 2.2 A/AA rule checks: missing alt text, unlabelled form fields, nameless buttons and links, broken heading order, missing landmarks and language, skipped video captions, disabled zoom, duplicate ids, positive tabindex, aria-hidden focus traps, and more. You get a score, each issue mapped to its success criterion, the exact CSS selector of every offending element, and how to fix it.

$12 one written report, human-reviewed, delivered by email within 24h

For a page that matters (checkout, signup, home page): the automated findings plus my own read of what to fix first, in plain language you can hand to a developer. Fixed scope, one page, no legal certification. Reply to any email from me or use the API and I will follow up.

$0.01 per call to the API, paid in USDC on Base over x402 - no account, no signup, no card

Built for agents and CI pipelines: machine-readable JSON, stable issue ids, selectors you can grep the DOM with.

curl -s "https://a11yscan.pages.dev/api/a11y?url=https://example.com"
# HTTP 402 with payment details - pay, retry, get the report

Free tier, no payment, scores and the top three issue types only: GET /free?url=...

Try the free scan

Static analysis of server-rendered HTML. No LLM. Public pages only.

Results appear here.

What it checks

1.1.1 text alternatives for images
1.2.2 captions on video
1.3.1 labels, table headers, list structure, heading semantics
1.3.6 main landmark
1.4.3 inline colour contrast (computed)
1.4.4 zoom not blocked by the viewport meta tag
2.4.2 page title, 2.4.3 tab order, 2.4.4 link purpose
3.1.1 page language
4.1.1 unique ids, 4.1.2 names and roles, aria-hidden focus traps, frame titles

What it is not

Automated static analysis is a prioritised snapshot, not a legal accessibility certification and not a full audit. It does not execute JavaScript, does not evaluate CSS classes against the rendered layout, and cannot judge whether alt text or link text is meaningful - only that it exists. Those limits are stated in every report.