You Don’t Need to Worry About Accessibility with AI (Or Do You?)
summary
A demo generates code fixes, accessibility tests and CSS tweaks using Playwright MCP and AI copilots and pushes the page to a Lighthouse score 100, but keyboard focus traps and real-world interaction bugs remain undetected. AI excels at pattern‑based rules, test generation and translating WCAG prose into code suggestions, yet context decisions (aria-hidden vs aria-label), unexpected UX flows, and PR review scale require human ownership, CI checks, TDD, Storybook and automation; a reported 356 → 50 bugs reduction shows measurable but partial gains.
AI-written summary · about
chapters (9)
- 0:00AI and accessibility intro
- 3:46Why accessibility feels hard
- 7:46AI for accessibility rules
- 11:57AI fixing accessibility live
- 14:49AI limits in accessibility
- 19:08Risks of AI-only accessibility
- 21:17Human-AI accessibility workflow
- 25:58Engineering workflow with AI
- 27:25Summary
related talks
43mHow to Use AI to Build Accessible Products
32mAccessibility powered by AI
35mAutomated Accessibility Testing
33mAutomated Accessibility Testing
53mAutomating Accessibility Assurance
25mAI-Powered Debugging & Browser Automation with Playwright MCP
44mTesting Web Accessibility
59mFixing Accessibility After the Fact Is Already Too Late