81% of devs trust AI-generated code for accessibility. The data disagrees.

lines of HTML codes

There is a confidence gap forming around AI-generated code and accessibility, and AudioEye just put numbers on it.

The digital accessibility company published new research alongside a business survey covering developer teams that use AI coding tools. The two findings do not line up.

What the Research Found

AudioEye tested leading AI coding tools by having them build websites. The tools tested included products from OpenAI, Anthropic, Google, xAI, and Lovable. Every one of them produced sites with high-severity accessibility issues. Not edge cases. High-severity.

At the same time, the business survey found that 81% of developer teams using these tools are confident the AI-generated code on their sites meets accessibility standards.

Those two data points should not coexist. They do.

The Other Number That Stands Out

The survey also found that 73% of teams report accessibility issues and complaints have increased since adopting AI coding tools. So nearly three-quarters of teams are already seeing real-world fallout, yet four in five still believe their output is compliant. That is the gap AudioEye is pointing at.

Why This Matters for Operators

Accessibility compliance is not just a technical checkbox. Inaccessible sites create legal exposure and lock out users with disabilities. If you are shipping with Cursor, Claude, Copilot, or any vibe-coded stack, AudioEye’s finding is a prompt to add an accessibility audit step before you ship, not after complaints come in.

Kelly Georgevich, CEO of AudioEye, is behind the research release. The company trades on Nasdaq under the ticker AEYE.

Stay on top of AI & Automation with BizStack Newsletter
BizStack  —  Entrepreneur’s Business Stack
Logo