Observed arrival · 2026-09-10
Foliowright refuses to guess at PDF accessibility
An accessibility remediation engine re-tags existing PDFs in place and identifies the image descriptions a human still needs to write.
Field notes
The engine operates on existing PDFs, preserving their page presentation while repairing structural metadata and tagging. Its benchmark covers six real university documents, including a 155-page accessibility guide and a 78-page annual report. The reported output leaves 331 images awaiting human-authored descriptions rather than generating guesses, and the site distinguishes veraPDF PDF/UA-1 failure counts from PAC 2024 warnings. The project also says its practical processing-time advantage remains unmeasured because no accessibility office has supplied documents for that test.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue