We scanned 2,400 real websites. 96.6% of the ones we could read fail an automated accessibility check.
Original research, 2026-08-10. Method, limitations and reproduction instructions included, because a statistic you cannot check is marketing.
Method, first
A seeded random sample of 2,400 domains was drawn from the Tranco research list, restricted to ranks 10,000–300,000. Below 10,000 the list is dominated by CDNs, infrastructure and platform domains that no business really "owns"; above 300,000 the parked and dead rate climbs steeply.
Each domain got one request to its home page, preceded by a robots.txt check that
was honoured, identifying itself as SteplessScanner/1.0. Where the apex did not resolve we
retried once with www.. The served HTML was then run through the same open-source check
engine that powers the free scanner on this site.
| Domains attempted | 2,400 | |
|---|---|---|
| Successfully scanned | 1,397 | 58.2% |
| Blocked, unreachable or errored | 959 | 40.0% |
| Excluded by robots.txt | 44 | 1.8% |
What this cannot tell you
- It undercounts. Colour contrast — the most common failure on the web — is not checked here at all, because deciding it needs the page rendered. Neither is keyboard operation. Every number below is a floor, not a ceiling.
- Only home pages. Checkout, which is where complaints usually point, is not in this sample.
- Only served HTML. Sites that build their interface in the browser look better here than they are.
- Non-response is not random. 40.0% of domains could not be read, many because they sit behind bot protection. Larger and better-resourced sites are likelier to be in that group, so the scanned set probably skews smaller.
The headline
Across 1,397 sites we could read, 96.6% had at least one detectable WCAG failure and 70.3% had at least one we classify as critical — meaning a control that assistive technology cannot name, or an image with nothing to announce. The median site had 9 failing elements; the mean was 22.8, pulled up by a long tail of sites with hundreds.
This independently replicates the reference study
WebAIM's Million report tests the top million home pages every year with a different codebase, a different sample and no connection to us. On the categories both studies measure, the numbers land in the same place — 96.6% of sites failing here against WebAIM's 95.9%, and near-identical prevalence for missing document language and empty links. Two independent measurements agreeing is the main reason to believe either of them.
What actually fails
Share of scanned sites with at least one instance of each problem.
| Failure | Sites affected | Count |
|---|---|---|
| No main landmark | 52.1% | 728 |
| Links with no accessible name | 48.7% | 680 |
| No bypass mechanism | 47.7% | 667 |
| Form fields with no label | 38.5% | 538 |
| Duplicate ids | 34.8% | 486 |
| Skipped heading levels | 30.8% | 430 |
| Images with no alt text | 26.6% | 372 |
| Buttons with no accessible name | 25.8% | 360 |
| Non-descriptive link text | 21.7% | 303 |
| No h1 | 21.3% | 297 |
| Pinch-zoom disabled | 16.3% | 228 |
| No document language | 13.4% | 187 |
| aria-hidden hiding focusable controls | 12.0% | 168 |
| Uninformative alt text | 11.2% | 157 |
| Broken aria-labelledby/describedby | 8.4% | 117 |
| Empty headings | 7.9% | 111 |
| Untitled frames | 7.5% | 105 |
| Broken aria-controls/owns | 7.2% | 100 |
The striking thing is how ordinary the list is. These are not exotic edge cases; they are alt attributes, form labels and button names. Most of them are a few hours of work for someone who knows where to look.
Accessibility overlays
We detected 18 sites running an accessibility overlay widget — 1.3% of those scanned.
| Vendor | Sites |
|---|---|
| userway | 8 |
| accessibe | 4 |
| equalweb | 4 |
| audioeye | 1 |
| reciteme | 1 |
| aioa | 1 |
Of those sites, 100.0% still had at least one detectable failure in the served HTML and 77.8% still had a critical one. Their median count of failing elements was 26, against 9 for sites with no overlay.
Read that comparison carefully
Overlay adopters are self-selected. They are, by definition, sites that already knew they had an accessibility problem, and they skew towards businesses large enough to be worth suing. So this is not evidence that overlays make sites worse. It is evidence of something narrower and still important: installing one does not remove the failures from the markup, which is what a tester, a regulator or an expert witness reads.
For the wider evidence on overlays — the FTC's $1,000,000 action, and the 983 US lawsuits filed in 2025 against sites that already had a widget installed — see our sourced summary.
By platform
| Platform detected | Sites in sample |
|---|---|
| wordpress | 125 |
| nextjs | 122 |
| ghost | 107 |
| drupal | 42 |
| hubspot | 34 |
| webflow | 28 |
| shopify | 26 |
| woocommerce | 16 |
| joomla | 15 |
| magento | 8 |
Platform is fingerprinted from markup signatures and is a floor: a site whose platform we could not identify is counted as unidentified, not as custom-built.
Only 9.4% link an accessibility statement
An accessibility statement is explicitly required under the European Accessibility Act and is the standard way to document good-faith effort everywhere else. Almost nobody has one. It is also the cheapest thing on this entire page to fix — there is a free generator here, and it will not let you claim conformance you have not verified.
Reproduce it
The sampling is seeded, the crawler and the check engine are open source, and the analysis script prints everything above. You can regenerate the identical sample and check our arithmetic: github.com/bol-ship-it/stepless-scanner.
We do not publish per-domain results, and will not. A list of named small businesses with their accessibility failures attached is a target list for the plaintiff firms driving the current wave of litigation. Publishing it would be the opposite of the point.
If you want to know about your own site
The free scan runs the same checks on any page in a few seconds, with no signup. If you would rather have it done properly, the Fix Pack is $249.