Automated Accessibility Testing by Anna Maier

Accessibility means building software for the widest variety of users possible, going beyond just accommodating blind users

According to WebAIM’s annual report, 95.9% of the top 1 million web pages have at least one accessibility failure

Automated testing tools can only detect 20-30% of Web Content Accessibility Guidelines (WCAG) violations, as many criteria require human context and judgment

AI currently cannot fully “fix” accessibility issues because:

• AI models are often trained on inaccessible data
• Accessibility is not just a technical problem but requires understanding context and user needs
• AI generates generic solutions without understanding specific use cases

Key areas to test for accessibility include:

• Keyboard navigation
• Semantic HTML roles
• Text resize/layout adjustments
• Image alt text descriptions
• Focus order

Recommended testing approach:

• Use automated tools as baseline (like axe, Playwright, Testing Library)
• Implement tests at both unit and end-to-end levels
• Include manual testing
• Put accessibility tests in CI pipeline

Common testing tools:

• axe-core
• Playwright
• Pa11y
• Google Lighthouse
• Accessibility Insights browser extension

The EU is implementing new accessibility laws affecting private sector websites

Testing should be done continuously and integrated into development pipelines

Context is crucial - only content creators can determine appropriate descriptions and semantic meanings for their content