We are increasingly asking AI to fix our WCAG errors, but algorithms operate on probability, not empathy. If you are using AI coding assistants to write your front-end components, you might be shipping hallucinated ARIA labels and broken accessibility without knowing it. Join me to explore the promises and perils of automated accessibility, and learn how to use AI as a testing partner rather than a replacement for inclusive design.

Prerequisite

Basic understanding of web accessibility principles (WCAG, ARIA, Semantic HTML).
Familiarity with using AI coding assistants (like Claude, ChatGPT, or GitHub Copilot) in a development workflow.

Target Audience

Developers, UX/UI designers or acceccibility advocates

Outline

We are entering a dangerous new era of web development where teams are asking Generative AI to "fix my WCAG errors." On the surface, it sounds like a utopia: pass your broken DOM to a Large Language Model (LLM) and get a perfectly accessible component returned in seconds.

However, AI operates on statistical probability, not human empathy. When we rely on AI to automatically patch accessibility, we often end up with hallucinated ARIA labels, nonsensical focus trapping, and fundamentally broken screen-reader experiences that technically pass automated audits but fail actual users.

This session explores the intersection of Generative AI and Web Accessibility. We will look at exactly where AI fails spectacularly at inclusive design. More importantly, we will cover how engineering teams can pragmatically use AI as an assistive testing and ideation tool rather than an automated fix. As developers, we cannot outsource human empathy to an algorithm.

Learning Objectives

Understand why LLMs fail at complex inclusive design and how to spot "fake" accessibility fixes in AI-generated code.

Learn how to write specific, context-aware prompts that force AI to act as a rigorous accessibility auditor rather than an automated code generator.

Grasp the ethical implications of outsourcing digital inclusion to algorithms, and establish team guardrails for AI use in front-end development.

Experience level
Intermediate