Skip to content
Blog7 min read

Continuous accessibility: why annual WCAG audits keep failing

Annual accessibility audits are stale before they're read. Here's what CI-native WCAG 2.2 monitoring looks like, and why regressions — not initial findings — are the real compliance risk.

accessibilityWCAGcomplianceCI/CD

Digital accessibility lawsuits have risen every year for a decade, and most defendants have something in common: they had an audit. A PDF from a reputable firm, findings dutifully fixed, box checked. Then fifty deploys happened, and by the time the demand letter arrived, the audit described a website that no longer existed.

Audits measure a moment; websites are a stream

The failure isn't the audit's quality — manual expert review catches things automation never will. The failure is cadence. A quarterly or annual snapshot cannot govern a codebase that changes daily. Accessibility regressions behave exactly like functional regressions: they're introduced by ordinary pull requests, they compound quietly, and they're cheapest to fix at the moment they're introduced.

What CI-native monitoring changes

  • Every deploy is scanned against WCAG 2.2 AA — the delta since the last scan is the signal
  • Regressions block the pull request that introduces them, with the selector and criterion attached
  • Legacy issues become a tracked baseline instead of an undifferentiated wall of findings
  • The scan history forms a dated evidence trail — proof of an active conformance process, which matters legally

Automated coverage maps to roughly 30–50% of WCAG criteria, so continuous scanning doesn't replace expert review — it makes the expensive human pass durable. Pay for the audit once, then defend its findings automatically on every merge, instead of re-purchasing the same discoveries every year.

This is the architecture behind Clearpath: axe-core scanning wired into CI, regression diffs per deploy, and an append-only evidence record — accessibility as an engineering practice rather than a procurement event.

Have the problem this post describes? The AI Architect will match it to a product or practice — with a price.

Run the AI Architect