I recently built an accessibility checker called WebMotCheck: https://webmotcheck.com
I wanted to test it on real websites rather than controlled demo pages, so I randomly sampled 140 UK charities from the Charity Commission register, limited to incorporated charities with annual income between £250k and £5M.
Each homepage was rendered in Chromium at desktop width and at 320px, then checked against WCAG 2.2 AA criteria that can reasonably be tested automatically.
The result
138 of 140 homepages had at least one automatically detectable WCAG failure.
The median homepage had 5 distinct issues.
78 sites had 5 or more
20 sites had 10 or more
Most common failures
Issue Sites affected
Insufficient text contrast 79%
Links with no accessible name 69%
Duplicate IDs 64%
Invisible keyboard focus 39%
Form fields without labels 26%
Images missing alt attributes 25%
Reflow / horizontal scrolling at 320px 24%
Buttons without accessible names 11%
Browser zoom disabled 11%
A few numbers stood out.
There were 418 contrast failures, with a median failing contrast ratio of 2.85:1.
I also found 1,025 links with no accessible name, often social icons or search icons implemented as SVGs inside links without a label.
On the 55 sites with keyboard-focus problems, 414 of 1,696 focusable elements showed no visible focus state.
And 34 sites required horizontal scrolling at 320px, usually because of fixed-width elements.
Important limitation
Automated accessibility testing only catches part of the problem.
It cannot reliably decide whether:
alt text is actually useful
reading order makes sense
a custom widget works properly with a screen reader
text over a photograph has sufficient contrast in every part of the image
So these results do not mean 138 sites are completely inaccessible.
They mean 138 had at least one problem that could be detected automatically.
Interestingly, the two pages with no detected failures were not especially sophisticated. They were simply relatively simple sites with dark text, light backgrounds, straightforward layouts, and fewer icon-only controls.
I’d be interested to hear from people working in accessibility or frontend development:
Do these numbers surprise you? And what would you change or add to the testing methodology?
Top comments (0)