4.4Building an Accessibility Testing Strategy

Level 1: Automated Scanning (Every PR)

Run axe-core or similar tools on every page during every PR. This catches the 30-40% of issues that are detectable by rule-based scanning.

# Minimum viable accessibility CI
npx playwright test tests/accessibility/ --reporter=json > a11y-results.json

Level 2: AI-Assisted Auditing (Weekly)

Use AI agents to evaluate qualitative aspects -- alt text quality, reading flow, error message clarity. This catches an additional 20-30% of issues.

Level 3: Manual Expert Audit (Quarterly)

A human accessibility expert uses screen readers, keyboard-only navigation, and magnification tools to find the remaining issues. This is irreplaceable for complex interactions, cognitive accessibility, and WCAG AAA compliance.

Level 4: User Testing (Semi-Annually)

Test with real users who have disabilities. They find issues that no tool or expert anticipated.