AI can generate code in seconds. It writes copy, creates layouts, and suggests components. It’s fast, persuasive, and often… completely inaccessible.Invented ARIA roles. Non-existent color contrast. Incorrect semantic structures. Generic or missing alt text. Forms without labels. AI doesn’t know what it doesn’t know, and it produces outputs that look correct but exclude millions of users.The problem isn’t AI itself. It’s that it’s being used by people who don’t yet have the tools to evaluate its output from an accessibility perspective. And when the reviewer doesn’t know what to look for, errors slip unnoticed into production.In this talk, we’ll cover:The most common accessibility issues generated by AI (with real examples).How to ask the right questions to get more accessible outputs.How to use AI as an ally in testing and reviews.What every digital professional needs to know to avoid blindly trusting generated code.Because the future of the web isn’t about choosing between speed and inclusivity. It’s about learning to demand both.
