Website accessibility compliance statistics show a persistent gap between accessibility goals and the experience delivered by many websites. In February 2025, the WebAIM Million scan detected 50,960,288 accessibility errors across one million homepages, with detected WCAG failures appearing on 94.8% of them. UK public-sector monitoring from January 2022 through September 2024 also shows that remediation can produce measurable progress, but that fixing issues, publishing useful statements, and reaching a clean retest are separate outcomes.
Table of contents
- Global website accessibility error levels
- WCAG failure prevalence across homepages
- Images, contrast, and forms
- Headings, landmarks, and navigation
- ARIA and interactive controls
- Public-sector compliance and remediation
Global website accessibility error levels
The WebAIM Million is a large automated scan of the top one million homepages. Its February 2025 results provide a broad measurement of detectable accessibility problems, not a claim that every possible accessibility barrier on every site was identified. The scan found 50,960,288 distinct accessibility errors, equal to an average of 51 detected errors per homepage. The analyzed homepages contained more than 1.2 billion page elements, and detected errors affected 4.1% of all homepage elements.
The average error level moved in a better direction than the overall complexity level. Average detected errors fell 10.3% from the 2024 WebAIM analysis to the 2025 analysis. At the same time, average homepage complexity rose from 1,173 elements in February 2024 to 1,257 in February 2025, an increase of 7.1%. Across the six years covered by the WebAIM 2019–2025 series, average homepage element counts increased 61%.
That combination matters when interpreting compliance statistics. A homepage with more content, components, images, and interactive elements has more opportunities for a detectable problem. The 2025 sample also showed a difference by site rank: the top 100,000 homepages averaged 1,465 elements, compared with 1,014 for the last 100,000 sites in the sample. The WebAIM Million therefore describes both the scale of errors and the changing complexity of the pages being measured.
Source: The WebAIM Million: The 2025 report on the accessibility of the top 1,000,000 home pages, February 2025 unless otherwise stated.
WCAG failure prevalence across homepages
Detected WCAG 2 failures appeared on 94.8% of the one million homepages tested in February 2025. The comparable share was 95.9% in 2024, so the measured prevalence declined by 1.1 percentage points between those analyses. Across the longer series, the share of homepages with detectable WCAG failures declined from 97.8% in 2019 to 94.8% in 2025.
The results also show how widely error counts vary. Only 23.4% of homepages had five or fewer detected errors in the 2025 sample, while 33.6% had ten or fewer. Those figures describe detected automated errors, not a complete accessibility score or a legal determination of compliance. They are useful for showing that a relatively small group of homepages had low detected error counts, while the large majority had at least one measurable issue.
The most common detected categories were concentrated in familiar WCAG-related implementation areas. Low-contrast text appeared on 79.1% of homepages, missing alternative text for images occurred on 55.5%, and missing form-input labels occurred on 48.2%. Empty links appeared on 45.4% of homepages, empty buttons on 29.6%, and missing document-language markup on 15.8%.
| Detected issue | Share of homepages | Measurement period |
|---|---|---|
| Low-contrast text | 79.1% | February 2025 |
| Missing image alternative text | 55.5% | February 2025 |
| Missing form-input labels | 48.2% | February 2025 |
| Empty links | 45.4% | February 2025 |
| Empty buttons | 29.6% | February 2025 |
| Missing document-language markup | 15.8% | February 2025 |
These categories should be read as prevalence measurements in the scanned sample. They do not establish that every instance has the same user impact, and they do not replace manual testing with assistive technology or user research.
Images, contrast, and forms
Images were a substantial part of the 2025 homepage sample. The one million homepages contained more than 58.6 million images, averaging 58.6 images per homepage. The average number of images per homepage increased 5.4% from 2024 to 2025. As visual content became more common in the scanned pages, alternative text remained one of the most frequent detected accessibility categories.
Missing alternative text affected 18.5% of all images in the February 2025 WebAIM sample, excluding intentional empty alternative text. Linked images represented 44% of the images missing alternative text, and one in five linked images lacked alternative text. A missing description on a linked image can be especially consequential because the image may be carrying the link’s usable name. The supplied WebAIM measurement does not assign an individual user impact to each image, but it quantifies how often this implementation gap appeared.
Alternative text quality also matters beyond whether an attribute exists. Questionable or repetitive alternative text appeared on 11% of images that had alternative text in the February 2025 sample. That statistic distinguishes presence from usefulness: a page can contain an alternative-text value while still offering repetitive or insufficient information.
Contrast was another dominant category. Each homepage had 29.6 distinct low-contrast-text instances on average in February 2025, and low-contrast text was detected on 79.1% of homepages. The homepage-level prevalence and the average number of instances measure different things: the first indicates how widespread the issue was across sites, while the second indicates the average amount detected on a homepage.
Forms were smaller in count but important in interaction. Homepages had an average of 6.3 form inputs in February 2025. Of those homepage form inputs, 34.2% were not properly labeled in the sample. At the homepage level, missing form-input labels occurred on 48.2% of homepages. Together, these figures show both the average number of inputs and the share of inputs or pages affected by labeling problems.
Headings, landmarks, and navigation
Structural markup was another major part of the WebAIM measurement. The February 2025 scan detected nearly 25 million headings, averaging nearly 25 headings per homepage. H1 elements accounted for 6.0% of detected headings, H2 elements for 30.5%, and H3 elements for 36.8%.
The total number of headings increased 4.6%, from 23.7 million in 2024 to nearly 25 million in 2025. That change is consistent with the broader increase in homepage elements, but the statistic does not show whether individual sites added useful structure or simply added more heading elements.
Several heading-pattern measurements identify common structural irregularities. Multiple H1 elements appeared on 16.3% of homepages in February 2025. The sample contained 1,088,074 skipped-heading-level instances, and skipped heading levels appeared on 39% of homepages. No headings were present on 9.8% of homepages.
Landmarks were more common than complete heading structures. At least one ARIA region or landmark was defined on 80.5% of homepages in February 2025. This is a presence measurement: it does not indicate that every region was correctly named, nested, or useful for navigation. The contrast between 80.5% with at least one region or landmark and 9.8% with no headings also illustrates why accessibility review needs multiple structural checks rather than a single pass-fail number.
The practical pattern in these statistics is clear: page structure is widely implemented, but it is not consistently ordered or complete. A homepage may contain headings and landmarks while still presenting skipped levels, multiple H1 elements, or no headings at all. Those are separate measurements within the same one-million-homepage scan.
ARIA and interactive controls
ARIA usage was widespread in February 2025. The sample contained 105,509,338 ARIA attributes, averaging nearly 106 per homepage. ARIA code usage increased 18.5% from 2024 to 2025, and homepages averaged 25 aria-label, aria-labelledby, or aria-describedby attributes.
ARIA was present on 79.4% of the one million homepages, excluding ARIA landmark roles. The WebAIM comparison found an average of 57 detected errors on homepages with ARIA, versus 27 on homepages without ARIA, in the 2025 sample. This comparison describes the pages in each group; it does not establish that ARIA caused the higher error count. Pages using ARIA may also be more complex or contain more interactive components.
Menus provide a more specific example. An ARIA menu was present on 4.5% of homepages. Of homepages with ARIA menus, 35% had barriers associated with missing required menu markup or interactions. The statistic concerns the subset that used ARIA menus, not all homepages.
Other interactive-control patterns also increased. Homepages averaged 18 aria-hidden=true attributes in February 2025, up 16% from 15.5 in 2024. They averaged 26 tabindex=0 or tabindex=-1 instances, up 13% from 23.1 in 2024. These numbers quantify code patterns, but they do not by themselves show whether focus order, keyboard operation, or hidden-content behavior worked correctly for every user.
Public-sector compliance and remediation
The UK Government monitored 1,203 public-sector websites and 21 mobile apps between January 2022 and September 2024. This was a public-sector monitoring program, so its findings should not be treated as a direct estimate for all websites worldwide. It offers a separate view of compliance work: what organizations fixed, what remained, and how documentation changed after monitoring.
Across the monitored websites and apps, the program found 29,787 accessibility issues. Public-sector organisations fixed 16,482 of them, equal to 55.3% of the issues found. A total of 3,693 issues, or 12.4% of those found, remained unfixed at retest during the monitoring period. These percentages describe the reported issue outcomes and should not be combined with the WebAIM homepage percentages, which use a different population, method, and measurement period.
The UK results also measured organizational follow-through. After monitoring and an opportunity to fix issues, 68% of the websites and apps had fixed the issues or had short-term plans for remaining fixes. Initially, 85% had published an accessibility statement. After the monitoring process, 97% had published one, and 65% had statements containing all information required by the model template.
Simplified testing covered 1,151 websites and found no accessibility issues on 10 of them. Among the websites tested, half had 23 or fewer issues, while the maximum was 73 issues on one website. This distribution shows why a single average can hide the range of remediation work across organizations.
The UK Government’s monitoring statistics therefore separate three related but different dimensions of compliance: the number of technical issues found, the number fixed or still open, and the completeness of the accessibility statement. A published statement increased from 85% to 97% after monitoring, but the share with all information required by the model template was 65%. Documentation coverage and technical remediation moved together in the program, but they were not identical outcomes.
Source: Accessibility monitoring of public sector websites and mobile apps from 2022 to 2024, January 2022–September 2024.