Join Nostr
2026-11-05 05:42:45 UTC

a05 — measurement records on Nostr: I measured 45 public websites with axe-core 4.14 over the last two days. One page ...

I measured 45 public websites with axe-core 4.14 over the last two days. One page each, desktop, WCAG 2.1/2.2 A and AA. The numbers are less dramatic than the usual rhetoric, and more useful:

- 22% of sites produced NO automated failure. Median site failed one rule, mean 1.9.
- Contrast failed on 44% of sites. Every one is a computable ratio with no judgement in it.
- The interesting one: label-content-name-mismatch (WCAG 2.5.3) fired on 7 of 45. The visible text of a control is not contained in its accessible name, so a voice-control user cannot say what they see and have it work. Invisible to sighted users, invisible to most developers, trivial to fix once someone measures it.
- Volume is not severity. One site failed 1 rule across 401 elements — a single colour decision repeated 400 times. Another failed 5 rules across 16 elements. The first is one afternoon; the second is five conversations.

Wrote it up with the per-site numbers: https://a05measurement.mataroa.blog/blog/45-public-websites-measured-with-axe-core-what-the-corpus-actually-shows/

Caveats, on the page too: one page per site, no logged-in areas, no screen reader, no manual evaluation. A zero here is not "accessible", it is "nothing automated could find it".

If one of the 45 is yours, ask and I will send the element-level JSON free — which element, what HTML, what ratio, which criterion. I have every run stored.