WebAIM 2026 report on Accessibility errors across the top 1M web pages
4 points by wgx
4 points by wgx
Good resource, but the title (in here, not in the original report) is pretty clickbaity. From their chart, errors increased in 2026, but decreased in 2025, and the 2026 level seems to be on par with the 2024 level.
Some remarks on interpreting this.
False positives exist. “Low contrast text”, the most frequent by far, is particularly bad at this due to overly-simplistic measurement.
Not all errors are equally bad. Some, in fact, are genuinely unimportant. If you use a script that is only used for one language, for example, then “missing language” really doesn’t matter.
Every accessibility tweak made is an opportunity to do it wrong. Show me “missing language” trending down (surprisingly substantially, even), and I’ll show you the far worse “incorrect language” trending up (though not by as much): for when the boilerplate people start from includes <html lang="en">, or the framework starts writing lang attributes and has its config default to English, you end up with a lot of non-English pages claiming to be English. And, as with the first rule of ARIA, no lang is better than wrong lang.
I can't help but think that this sort of accessibility checklist mentality does more to undermine accessibility work than it does to promote it. A completed checklist is a receipt that you went through the mandatory rigmarole, not that human beings can access your website without issues.