accesspatched

Research

94.8% of the web is broken for someone: inside a scan of a million home pages

Published 14 July 2026

If you want a single website accessibility statistic for 2025, here it is: 94.8% of the top one million home pages had detectable WCAG 2 A and AA failures, with an average of 51 errors on each page (WebAIM Million, 2025). This is not a survey or an estimate. It is an automated scan of an actual million sites, run the same way every year, and the finding is blunt: the broken web is the normal web.

Below is what the numbers say, what they leave out, and why the shape of the data is more useful to a shop owner than the scary headline.

What is the WebAIM Million?

The WebAIM Million is an annual accessibility analysis of the home pages of the top one million websites. WebAIM runs automated testing across all of them and publishes the aggregate results, which makes it one of the few accessibility datasets large enough to talk about "the web" rather than a handful of case studies. Because the method stays consistent, year-on-year comparisons actually mean something.

The 2025 run found detectable errors on 94.8% of home pages (WebAIM, 2025). For context, the 2024 run came in at 95.9%. That is a small improvement, and it is worth being honest about scale: a bit under half a percentage point of the entire web moving is millions of pages, but it still leaves the overwhelming majority failing. Nobody should read this as "problem mostly solved".

How many accessibility errors does an average page have?

The average home page carried 51 distinct errors (WebAIM, 2025). To make that human, WebAIM frames it as roughly one detectable barrier for every 24 page elements. So on a typical page, a person using assistive technology hits something broken about every couple of dozen elements they move through. That is not one unlucky snag on the way to checkout. It is a steady drumbeat of friction across the whole page.

And remember these are only the errors a machine can detect. Automated tools catch the mechanical failures. They cannot tell whether a reading order makes sense or whether a task can actually be completed, which is why the true count of real-world barriers is higher than 51, not lower. We get into that gap in why a clean automated scan proves almost nothing.

Which accessibility errors are most common?

The most useful part of the data is how concentrated it is. Just six error types account for 96% of all detected errors across the million pages (WebAIM, 2025). The long tail is short. Those six are:

Low-contrast text is the runaway leader, present on 79.1% of all home pages (WebAIM, 2025). Nearly four in five sites have text that does not stand out enough from its background. That is a design and CSS problem, not a deep engineering one, which is part of why it is so widespread and so fixable.

The good news buried in the bad news is that concentration. You do not have to boil the ocean. Clearing these six removes the large majority of what any automated scan will ever flag on your site. We break down each one, and the exact fix, in the six mistakes behind 96% of failures.

Is e-commerce any better than the average?

Not meaningfully, and the failures land right on the money-making parts of the site. A peer-reviewed analysis of e-commerce accessibility found that contrast and perception problems affected 54.4% of the sites evaluated, so a majority of shops had contrast issues, and that missing text alternatives on product images were common (PeerJ Computer Science, 2022).

Think about where those two failures sit. Product photos with no alt text mean a screen reader user hears nothing about the thing they are considering buying. Low contrast means prices, specs, and calls to action that some customers simply cannot read. For a retailer, these are not cosmetic. They are the difference between a browsing session and a purchase.

Why these numbers should matter to a business

It is tempting to file "94.8%" under interesting-but-not-my-problem. The reason not to is on the demand side. The World Health Organization estimates that about 1.3 billion people, roughly 16% of the world or one in six, live with a significant disability (WHO, 2023). Put the two facts next to each other and the picture is stark: a mainstream slice of every audience, meeting a web that fails almost universally.

That gap is where quiet revenue loss lives. Shoppers who cannot use a site do not usually complain, they just leave, which we cover in how inaccessible checkouts quietly cost £17 billion. The WebAIM Million tells you the supply side of that story: the barriers are nearly everywhere, and they are concentrated enough that clearing them is a realistic project rather than a moonshot.

So what should a shop owner do with this?

The temptation with a statistic like 94.8% is either panic or fatalism, and neither is useful. Here is the more practical reading. Because failure is close to universal, being genuinely accessible is a real differentiator rather than table stakes, and because the failures are concentrated in six recurring types, getting there is a finite project rather than an open-ended one.

In our experience the sensible order is straightforward. Start with an automated scan to find the machine-detectable errors, since that is where the six common failures show up and where the cheapest wins live. Fix those. Then, and only then, put the site in front of a human using a keyboard and a screen reader to walk the real buying journey, because that is the only way to catch the barriers a scanner structurally cannot see. The average of 51 errors per page is a signal that the first pass usually finds plenty to do, and that the work pays for itself in removed friction long before you reach the harder, manual-only issues.

What the number should not do is make you wait for the industry to fix itself. A move from 95.9% to 94.8% in a year tells you the tide is not going to lift your shop for you. If your competitors are all failing, that is an opportunity, not an excuse.

Common questions

Does a low error count mean my site is accessible?

No. A low automated error count means you have cleared the machine-detectable issues, which is a genuine and worthwhile floor. But automated tools only find a fraction of real barriers, so a clean scan is a starting point for human testing, not a certificate of accessibility.

Has web accessibility improved over time?

Slightly. The failure rate moved from 95.9% in 2024 to 94.8% in 2025 (WebAIM, 2025). The direction is right, the pace is glacial, and the absolute level is still a large majority of the web failing. Improvement at this rate is not something to wait out.

At AccessPatched we treat these statistics as a map, not a verdict. We audit your site against WCAG-EM, confirm the machine-detectable issues and then test manually with a screen reader to find the ones the scanners miss, ship the corrections back to you as pull requests, and issue an EN 301 549 conformance statement when the work is done. You can see the full picture of what we handle on our services page.

See where your own shop stands. Our free scan runs the same automated checks a regulator's tester starts with, then a human goes through your site with a screen reader and emails you the findings, in plain language. If your shop passes, we say so.

Get a free scan

Sources

Back to all research