Ideas and guides about making the web work for everyone.
The free accessibility tools that run inside AI coding assistants have quietly become good. We spent time with the two most popular ones. They find problems and they fix them, and they share an architecture worth understanding before you rely on one.
WebAIM counted 133 million ARIA attributes in 2026, a 27% jump in one year, alongside more errors per page than the year before. More accessibility code, less accessible web. Here is what that says about AI-generated markup and the criteria nobody ever wrote down.
WCAG-EM 2.0, the W3C's new official method for evaluating conformance, says in plain text what the 'automate it all' vendors never will: no tool alone can prove a site is accessible. Here's why that vindicates the way we built Jeikin.
AI made producing plausible-looking code nearly free. Accessibility is the property where plausible and correct diverge the most, which is why it regressed the year everyone adopted tooling to fix it.
Security and quality tools are moving their checks inside the AI writing loop, enforced by the agent as it generates code. Accessibility is arriving there too. Here's the pattern, and what it takes to count.
The whole accessibility industry now agrees the fix belongs inside the AI authoring loop. The only question left is what "in the loop" actually has to do to count.
One year after the EAA deadline, the picture is clearer than the headlines suggest. France has convicted a retailer, most states have issued notices, and not a single confirmed fine has yet been collected. Here's the verified, per-country reality.
Automated scanners set a capability floor for accessibility, not a ceiling. Here's what they catch, what they can't, and how to cover the rest.
Small founders on Shopify and Woo lean on Cursor, Copilot, and Claude Code for the custom bits of their store. Here's the map of what AI tools can actually fix, what they can't, and which parts of accessibility are yours to own in the first place.
For two decades, accessibility specialists have spent a meaningful slice of their week on the boring 30–40% — missing alt text, wrong heading levels, ARIA in invalid combinations. In the last 60 days, that layer started handling itself. Here's what that frees them to do.
GitHub shipped its own accessibility scanner. A community pack shipped 79 specialised agents. The lane is no longer empty. Here's what those tools do, what they miss, and where evidence still has to come from.
A practical comparison of accessibility MCP servers and tools that work with Claude Code, Cursor, and Windsurf — what each one does, what it doesn't, and when to pick which.
HHS's May 2026 healthcare accessibility deadline still stands, DOJ extended Title II by a year, and the EAA is enforcing in Europe. What changed, what didn't, and what evidence regulators want.
The Jeikin Accessibility app is on the GitHub Marketplace. Install it on a repo and every pull request gets a WCAG review before it merges.
AI agents don't see your website. They read the accessibility tree. The same structure built for screen readers is now the interface for every AI agent on the web.
Forbes just put a number on it. Settlements run $25k to $100k per case, before legal fees. One person can't merge every PR. Compliance isn't a role, it's a pipeline.
Every AI coding tool can find accessibility issues. Almost none can prove they were fixed. That gap is where compliance falls apart.
AI coding assistants can follow instructions. But instructions without enforcement are just suggestions. Here's what changes when accessibility has a system behind it.
Accessibility overlay widgets promise instant compliance, but they create more problems than they solve. Here's what actually works.
Why we built an accessibility tool that works through your AI coding assistant — and what makes it different from everything else.