What is a website accessibility score, and how is it calculated?

That definition is deliberately narrow, because the number is easy to over-read. A score is a compression of many individual test results into one figure so you can track direction over time and compare pages at a glance. It is genuinely useful for exactly that, and misleading the moment it is treated as a pass or fail verdict on the law. The rest of this page explains the precise formula Certvo uses, what each band actually looks like on a real site, where the score stops and legal conformance begins, and how the figure is distributed across the 1,564 websites in the Certvo Web Accessibility Report 2026.

How Certvo calculates the score

Certvo does not average pass and fail counts. It weights every detected issue by how much it hurts a real user, sums those weights into a single penalty, and expresses that penalty as a share of everything the engine checked on the page. The weights are fixed by severity: a critical issue that blocks a task costs far more than a minor one that a user can work around.

• Critical issues add 25 penalty points each.
• Serious issues add 15 each.
• Moderate issues add 5 each.
• Minor issues add 2 each.

The total penalty is then divided by the total number of checks that ran, meaning the sum of failed checks (violations) and passed checks, and multiplied by 10. The score is 100 minus that figure, clamped to the range 5 to 99. Dividing by the total number of checks is what keeps the score fair across pages of different sizes: a large page with a hundred passing checks and three failures is not punished the same way as a tiny page where three failures are most of what ran.

A worked example makes the shape of it clear. Suppose a page has 3 critical, 10 serious and 20 moderate issues, across 400 total checks. The penalty is 3 times 25 plus 10 times 15 plus 20 times 5, which is 75 plus 150 plus 100, or 325. Dividing 325 by 400 gives 0.8125, and multiplying by 10 gives 8.1. The score is 100 minus 8.1, which is 91.9, rounding to about 92. Thirty-three separate problems, and the page still scores in the low 90s, because those problems are spread across four hundred checks that mostly pass. This is the single most important thing to understand about any accessibility score: it is a density measure, not a headcount of bugs.

The clamp matters too. A page never scores below 5, so a catastrophic page and a merely bad one are not always distinguishable at the very bottom. And a tested page never scores 100: the ceiling is 99. A raw 100 only occurs when no checks ran at all, which happens when a page is blocked, empty or failed to load, and a page nothing ran against tells you nothing. That is why the research dataset behind this article excludes those scans entirely.

What counts as a good score

Bands are more honest than a single threshold. The shares below come from the 1,564 scanned websites in the research dataset, so they describe how the real web is actually distributed rather than an ideal target.

90 and above (26.1% of sites)

The largest single band. These pages have no critical automated failures and only a light scatter of moderate or minor issues. A site here has usually invested in accessible components: labelled controls, sufficient contrast, alt text on meaningful images. It is a strong automated result, and it is still not a conformance claim, because manual criteria are untested.

80 to 89 (15.3% of sites)

Close. A page here typically has one recurring problem, often contrast on a brand colour or a set of icon buttons missing accessible names, repeated across the template. Because the issue is usually a single shared component, this band is the cheapest to move: one fix propagates everywhere the component appears.

60 to 79 (21.9% of sites)

Real barriers are present. This is the band the median site sits in, at a median score of 72. Expect several distinct problem classes at once: contrast plus unnamed links plus missing landmarks. Users can navigate parts of the site but will hit walls, and the fixes are a short list of concrete tasks rather than a rewrite.

40 to 59 (16.4% of sites)

Serious and critical issues dominate. Assistive technology users will be blocked from common tasks. A page in this band often combines missing form labels, images without alternatives and structure that screen readers cannot parse. Remediation is a project, not an afternoon, but it is a well-defined one.

Below 40 (20.2% of sites, 317 in the sample)

The page is effectively unusable with assistive technology. Critical failures are pervasive and the density is high enough that the penalty overwhelms the passing checks. Sites here usually need an accessibility pass built into the next redesign rather than a patch on the current one. Across the sample, 51.5% of sites carry at least one critical issue, so this bottom band is not an edge case.

Score versus compliance

A score is a health signal. Conformance is a legal state. They are correlated, and they are not the same thing, and conflating them is where organisations get into trouble. Automated tools can only reliably test the criteria that are machine-checkable, which is roughly 30 to 40 percent of the WCAG success criteria. The rest need human judgement: whether alt text is accurate rather than merely present, whether focus order is logical, whether an error message actually explains the error, whether a video caption matches the audio.

This cuts both ways. A page can score 95 and still fail WCAG, because a manual criterion it was never able to test is broken. And a page can score 60 and be one shared component away from 90, because a single unnamed button repeated site-wide is dragging the density down. The score tells you where to look and how urgent it is. It does not, on its own, tell you that you conform. For that you pair the automated scan with a manual review and record the result in a dated accessibility statement.

The distribution across the web

Certvo scanned 1,564 websites and scored each one with the formula above. The distribution is wide and slightly bottom-heavy, and it is not the smooth bell curve people expect.

The average score is 65.9 and the median is 72, so a typical site is passing more checks than it fails but is nowhere near clean. 41.4% of sites score 80 or above, which sounds encouraging until you set it against the other end: 28.0% of sites score below 50. In other words, roughly a quarter of the web scores well and more than a quarter is close to unusable, with the middle strung out between them. The gap between the average of 65.9 and the AA pass rate of 18.8% is the whole point of this page: most sites score like a passable grade and still fail conformance, because the score measures density and conformance measures the manual criteria the score never touches. Only 48.5% of sites even pass the lower WCAG single-A bar.

How to raise your score fast

Because the score is density-weighted, the fastest gains come from the issues that are both high-severity and repeated across a shared component. Work them in this order.

1. Fix colour contrast. The single most common serious issue in the dataset, present on 51.3% of sites. Usually a brand colour used for text or buttons. See how to fix low colour contrast and check pairs with the contrast checker.
2. Give every link an accessible name. Icon-only and empty links are read out as nothing by a screen reader. See how to fix empty links.
3. Name your buttons. The same problem on interactive controls, and a critical severity, so each instance costs 25 penalty points. See how to fix empty buttons.
4. Add alternative text to meaningful images. Another critical, and one manual judgement matters: the text must describe the image, not just exist. See how to fix missing alt text.
5. Add a skip link and proper landmarks. Content outside a landmark region is the most widespread issue overall. A skip link and correct regions fix navigation for keyboard and screen reader users at once. See how to add a skip link.

Then re-scan. The point of fixing shared components first is that each fix removes the issue everywhere it appears, so a single change can move the score by several points. When you are ready to measure the effect, run a fresh accessibility scan.

How other tools score

Certvo is not the only tool that produces a number, and the numbers are not interchangeable. Here is how the common ones work, from their public documentation.