Web accessibility remains a widespread challenge across the web. In February 2025, WebAIM detected 50,960,288 accessibility errors on one million home pages, while UK public-sector monitoring recorded 29,787 issues across websites and mobile apps between January 2022 and September 2024. These figures come from different methods and samples, so they should be read as complementary evidence rather than as one combined total.
Contents
- The state of accessibility across popular home pages
- The most common automated barriers
- Headings and landmarks for navigation
- ARIA, links, and interactive patterns
- Categories, languages, domains, and CMS differences
- UK public-sector monitoring results
The state of accessibility across popular home pages
The WebAIM Million 2025 report examined one million home pages in February 2025 using automated WAVE detections on rendered pages. The sample contained more than 1.2 billion page elements and 50,960,288 distinct detected accessibility errors. That works out to an average of 51 detected errors per home page.
The average detected-error rate fell 10.3% from the 2024 analysis to the 2025 analysis. That improvement happened while pages became more complex. Average home-page complexity increased from 1,173 elements in February 2024 to 1,257 in February 2025, a 7.1% rise.
Complexity also varied within the sample. The top 100,000 home pages averaged 1,465 elements, compared with 1,014 among the last 100,000 pages. More elements create more opportunities for useful content and interaction, but they can also increase the number of places where text, images, controls, and structure need accessible implementation.
These numbers describe automated detections, not complete conformance audits. A page without a detected error is not proven to be accessible, and a detected error does not describe every user impact. The report is most useful for showing the scale and distribution of recurring technical barriers.
The most common automated barriers
Detected WCAG 2 failures appeared on 94.8% of the one million home pages. Low-contrast text was the most widespread issue, appearing on 79.1% of pages. Each home page contained 29.6 distinct low-contrast text instances on average.
Images and forms were also frequent sources of detected problems. Missing alternative text appeared on 55.5% of home pages. The sample contained 58.6 images per home page on average, and 18.5% of all home-page images lacked alternative text, excluding intentionally empty alt attributes. The distinction matters: an empty alternative-text attribute can be appropriate for decorative images, while missing alternative text leaves no equivalent text for users who need it.
Missing form-input labels appeared on 48.2% of home pages. Looking at individual inputs rather than pages, 34.2% of form inputs were not properly labeled. Empty links appeared on 45.4% of pages, while empty buttons appeared on 29.6%. Missing document language appeared on 15.8% of pages.
| Detected barrier | Home pages affected in February 2025 |
|---|---|
| Any detected WCAG 2 failure | 94.8% |
| Low-contrast text | 79.1% |
| Missing alternative text | 55.5% |
| Missing form-input labels | 48.2% |
| Empty links | 45.4% |
| Empty buttons | 29.6% |
| Missing document language | 15.8% |
The pattern points to several practical review priorities: check color contrast across text states, provide meaningful alternatives for informative images, associate labels with form controls, and ensure links and buttons expose useful names. These priorities are grounded in the detected barriers, but automated results alone cannot assess every keyboard, screen-reader, cognitive, or usability experience.
Headings and landmarks for navigation
Page structure affects how people scan content and move through a page with assistive technology. The 2025 WebAIM sample contained 1.5 million h1 headings, representing 6.0% of detected headings. It contained 7.6 million h2 headings, or 30.5%, and 9.2 million h3 headings, or 36.8%. The sample also contained 3.8 million h4 headings, representing 15.5% of detected headings.
Heading order was not consistently maintained. WebAIM found 1,088,074 skipped heading-level instances, equal to one in every 23 headings in the sample. Skipped heading levels appeared on 39% of home pages. In addition, 16.3% of pages had more than one h1, while 9.8% had no headings at all.
Landmarks were more common than complete main-content identification. At least one region or ARIA landmark was defined on 80.5% of home pages, but a main element or main landmark was present on 42.6%. This difference suggests that having some landmark structure does not necessarily mean the page provides a clear primary-content route.
For accessible navigation, the statistics support a straightforward structural checklist: use headings that reflect the content hierarchy, avoid unexplained level jumps, identify the page’s main content, and use landmarks consistently. The counts do not establish that every skipped heading causes the same user impact, but they do show that structural inconsistencies are common enough to deserve routine testing.
ARIA, links, and interactive patterns
ARIA was widespread in the 2025 sample. WebAIM detected 105,509,338 ARIA attributes across the million pages, nearly 106 per page on average. ARIA code usage increased 18.5% from 2024 to 2025, and 79.4% of home pages used ARIA when ARIA landmark roles were excluded.
The presence of ARIA was associated with a higher average number of detected errors: pages with ARIA present averaged 57 detected errors, compared with 27 for pages without ARIA. This is an association, not evidence that ARIA caused the errors. More complex pages may be more likely to use ARIA and also more likely to contain detectable barriers.
Menu implementations were a particular risk. An ARIA menu role appeared on 4.5% of home pages, and 35% of detected ARIA menus introduced barriers because required menu markup and interactions were missing. Using a role without implementing the expected keyboard behavior and relationships can therefore create a misleadingly semantic interface.
Other implementation patterns were also common. Home pages averaged 18 aria-hidden="true" attributes, 3.6 role="button" attributes, and 26 tabindex="0" or tabindex="-1" instances. These counts do not indicate whether each use was correct. They do show why interactive components need testing beyond the presence of attributes: names, roles, values, focus order, keyboard operation, and state changes all matter.
Categories, languages, domains, and CMS differences
Detected errors varied across categories and implementation contexts in the WebAIM sample. Government-category home pages averaged 37.2 detected errors, 27.0% below the overall sample average. Shopping-category home pages averaged 71.2, or 39.8% above the overall sample average.
The domain and language comparisons also showed differences, but the groups were not equal in size. Home pages using .gov averaged 19.6 errors across 1,967 sampled pages, while .com home pages averaged 49.8 errors across 435,998 pages. English-specified pages averaged 39.8 errors across 504,595 pages, compared with 84.6 errors across 60,715 Russian-specified pages.
| Comparison group | Average detected errors | Sample size |
|---|---|---|
| Government category | 37.2 | Not stated in the supplied comparison |
| Shopping category | 71.2 | Not stated in the supplied comparison |
.gov top-level domain | 19.6 | 1,967 pages |
.com top-level domain | 49.8 | 435,998 pages |
| English-specified | 39.8 | 504,595 pages |
| Russian-specified | 84.6 | 60,715 pages |
Content-management-system comparisons showed another spread. Divi home pages averaged 27.6 errors across 8,672 pages where the CMS was detected. WordPress home pages averaged 50.0 errors across 241,401 pages where the CMS was detected.
These comparisons should not be treated as rankings or causal evidence. WebAIM explicitly presents category and technology results as associations. Differences can reflect page complexity, templates, content practices, sample composition, language, or other factors that the automated comparison does not isolate.
UK public-sector monitoring results
The UK Government Digital Service monitoring report covers a different type of evidence. Between 1 January 2022 and 1 September 2024, it monitored 1,203 websites and 21 mobile apps. The monitoring found 29,787 accessibility issues.
Public-sector organisations fixed 16,482 of those issues, or 55.3%. At retest, 3,693 monitored issues, or 12.4%, were not fixed. The remaining figures should not be interpreted as a simple remainder of all detected issues: the report’s monitoring and retesting process, issue recording, and categories determine what each measure represents.
Publication of accessibility statements improved during monitoring. At initial monitoring, 85% of monitored websites and apps had published an accessibility statement. After monitoring, 97% had one. However, only 65% of post-monitoring statements included all information required by the model template.
In UK simplified testing, the largest issue shares were Focus Visible at 13.3% and Minimum contrast at 13.0%. Keyboard access represented 10.8%, Reflow 9.2%, Name, Role, Value 8.6%, and Info and Relationships 6.4% of issues found.
| UK simplified-testing area | Share of issues found |
|---|---|
| Focus Visible | 13.3% |
| Minimum contrast | 13.0% |
| Keyboard access | 10.8% |
| Reflow | 9.2% |
| Name, Role, Value | 8.6% |
| Info and Relationships | 6.4% |
The UK figures apply to a public-sector monitoring sample and are not directly comparable with the prior 2020–2021 report because testing methodology and issue recording changed. They also are not directly interchangeable with WebAIM’s automated home-page detections. Together, the reports show two different measurement perspectives: large-scale automated screening of rendered home pages and monitored testing of public-sector websites and apps over a defined period.