a11y.sense

Landing page
perceivable-critical-panel
robust-critical-panel
Dev-Handoff-MD-FileGen

Inspiration

1.3 billion people live with a disability. a11ysense is an AI-powered sandbox dev QA tool that verifies accessibility ensuring smooth disabled user experiences on any live URL.

"Something like 99% of all websites have accessibility issues. Assume there's something wrong and you're probably right. No tool required" - Comment from disabled reddit user

Our mission is to help and empower developers ensure their website is up to WCAG Accessibility standards.

What it does

Paste any URL and a11y.sense audits it through the Web Content Accessibility Guidelines (WCAG) core principles. It returns WCAG-mapped issues with fixes, and displays everything on an interactive 3D human skeleton where principles map to body systems and common linked disabilities.

How we built it

Next.js + React 19 frontend, Three.js for the 3D skeleton, Google Gemini Flash for parallel AI audits across all 5 profiles simultaneously, Gemini Vision for screenshot transformations, and Playwright for browser automation and element coordinate resolution.

Challenges we ran into

Getting Gemini to return consistent structured JSON with CSS selectors and WCAG codes across 5 parallel profiles was the hardest part, As well as being able to consistently find issues that go against the WCAG standards and parsing the code from Playwright to find fixes for the devs. Ensuring 3d Model vectors aligned with found issues on the website was also a very nuanced and complicated issue.

Accomplishments that we're proud of

The Playwright parsing into Gemini pipeline to find WCAG issues is extremely consistent and amazing at finding issues devs could've missed. The 3D skeleton makes WCAG's four abstract principles immediately intuitive: click the brain for cognitive issues, click the hands for motor issues.

What we learned

Building a11y.sense gave us a deep appreciation for the craft of AI prompt engineering. Getting Gemini to return consistent, structured JSON across five parallel audit profiles taught us that the difference between a useful AI output and an unusable one often comes down to how precisely you constrain the prompt. We also came face-to-face with just how pervasive accessibility gaps are in the real world, auditing live URLs made it viscerally clear that accessibility is still treated as an afterthought by the vast majority of developers, reinforcing why tools like a11y.sense genuinely matter.

What's next for a11y.sense

We're expanding a11y.sense beyond the web. Mobile accessibility is a largely underserved space, and we want to bring the same WCAG-mapped auditing experience to iOS and Android apps. On the developer workflow side, we're building an IDE plugin and CI/CD integration so accessibility checks happen automatically at every commit, helping devs catch issues before they ever reach production, not after.