AI Doesn't Fix Accessibility. It Amplifies the Problems.

AI learns from existing code. And existing code is, by and large, not accessible. That means AI doesn't just fail to solve the accessibility problems - it amplifies them.

In this talk, I'll use findings from my test series, generating the same app with four different AI tools and testing each for accessibility, as a starting point. From there, we'll look at what this means in a world where vibe coders and AI agents are shipping apps at scale, tackle the "just prompt it better" argument, and cover concrete steps you can take to your own projects.

The barrier to building apps is lower than ever. The need for understanding accessibility has never been higher.
Key takeaways:

- Why AI amplifies existing accessibility problems rather than solves them
- What vibe coding and AI agents mean for accessibility at scale
- Why asking AI to "just make it accessible" can do more harm than good
- Concrete steps to take to your own projects, with or without AI