What government and university work taught us about accessibility

Accessibility is often treated as a final checklist: build the site, run an audit at the end, fix what it flags, and ship. We used to work that way too, until government and university requirements ruled it out.

A final audit cannot catch changes made after it, and sites change all the time through forms, new pages, content edits, and dependency updates. Each can introduce an accessibility failure. Automated scanners miss things too. We've had a contrast failure hide below the fold because a scroll animation kept the content invisible while the scanner ran, and the scan came back green.

On government and university work, accessibility is checked on every change. The build pipeline runs several automated and rendered-page checks, because each one catches failures the others miss. We build to WCAG 2.1 AA and Section 508 from the first commit. We also retrofit existing sites, though in our experience that usually costs more than building to the standard from the start. And the person publishing next month's update can undo the developer's work with one image missing alt text, so design, content, development, and QA all have a part in it.

Our build gate supports a WCAG 2.1 AA target. It stops known failures from shipping, and it cannot certify a site by itself.