State of Web Accessibility 2026

Overview

Introduction

Most of the web still does not work for people with disabilities. To measure how far off it is, and where the gaps actually are, we analyzed our own scan database: 1,079 websites, scanned with our current engine between 21 July and 18 August 2026, across nine industries.

Unlike vendor indexes that publish a single headline and no method, this report shows the full picture: how we measured, what we found, where we sit against other public benchmarks, and the limits of our sample. The data is aggregate and anonymized. No site names, URLs, or personal data appear anywhere in it.

Methodology

How we measured this

We are publishing the method before the findings on purpose. A benchmark is only as trustworthy as its method, and most accessibility indexes do not show theirs.

The Headline Number

Almost every website fails, and by a wide margin

99.0% of the websites we scanned have at least one WCAG failure. That is not an artifact of strict testing; it is the norm. For comparison, the WebAIM Million 2026 study found detectable errors on 95.9% of the top one million homepages. Two very different samples, the same conclusion.

Depth is the harder problem. The average site carries 62.5 distinct issues, dozens of barriers per site across contrast, structure, forms, and navigation. On our 0 to 100 compliance score, the average site scores 56.[1] (footnote)

“99.0% of websites fail at least one WCAG check. The average site has 62.5 accessibility issues.”

WebAbility, 1,079 sites, 2026.

Common Failures

The most common failures

Color contrast is the defining accessibility problem of 2026. It appears on 69.3% of sites and accounts for 21% of every issue we found, the single largest share. It is also among the most fixable, which makes its prevalence a story about attention, not difficulty.

A broad structural criterion, 1.3.1 Info and Relationships, registers even higher in our raw data (78.6% of sites). Because automated engines map a wide range of structural issues to 1.3.1, we treat it as a category rather than a single defect and lead this table with contrast, which is discrete and directly comparable to other public studies.

Severity

Severity matters, and most indexes ignore it

Counting issues without weighting them is misleading. A missing decorative-image alt is not the same as a checkout form a screen reader cannot complete. So we grade every issue.

• 36.1% of sites have at least one critical-severity issue. The other 63.9% are critical-free but almost all still carry lower-severity issues.
• Critical and serious issues together make up 46% of all issues found.[2] (footnote)

Two sites with the same issue count can be worlds apart if one has critical barriers on its core journeys. Severity is the difference between a cosmetic backlog and a legal exposure.

Industry Breakdown

Accessibility by industry

Across the 483 sites in our sample with a labeled vertical, E-commerce is the clear laggard.

Sites per industry and critical-issue rate. WebAbility, 2026. Industry labeled for 483 of 1,423 sites.[3] (footnote)

The E-commerce finding matters most commercially: these are transactional sites where an inaccessible checkout is both a lost sale and a legal exposure, and they are the most common target of accessibility demand letters. See our ADA demand letter guide.

What Helps

What actually moves the number

No single tool makes a site compliant, but the gap is not split evenly. On the same sites, applying our accessibility widget raised the average compliance score from 56 to 88 out of 100, a measured before-and-after on identical pages, not a projection.[4] (footnote)

The remaining gap to 100 is the part automation cannot honestly claim. Automated testing reliably catches roughly 30 to 40 percent of WCAG issues; the rest, whether alt text is meaningful, whether a custom component works with a screen reader, whether the keyboard path through a checkout completes, needs a person. The path that works is: automate what is safe, put humans on the rest, and keep a report you can show. See our widget vs audit vs automated decision guide.

Benchmarks

How our data compares

Our numbers sit inside the ranges published by other credible studies. That is the point: a consistent read, not an outlier built for a headline.

Unit differences apply; our “per site” is not identical to WebAIM's “per homepage.” Shown for orientation, not as a like-for-like claim. What no other study publishes: a severity-weighted score, cross-industry cuts, and a measured before-and-after widget effect.

Transparency

Method, limitations, and transparency

• Sample bias

  These sites ran through our scanner. Not a random sample of the web, and likely more accessibility-aware than average. The open web is probably worse.

• Automated detection has a ceiling

  We report what an engine can detect; it cannot judge whether alt text is meaningful or a flow is usable end to end.

• No CMS or platform breakdown

  Platform data is not captured in our pipeline, so we make no WordPress-vs-Shopify claims.

• Scoring window

  Only current-engine scans (from 21 July 2026) so the score is consistent; we do not publish a year-over-year trend, because an engine change would make it misleading.

• Aggregate only

  No site names, URLs, customer identifiers, or personal data appear in this report.

• Language of Page excluded

  3.1.1 Language of Page is intentionally excluded: our detector undercounts it (3.8% vs WebAIM's 13.5%), so publishing it would misrepresent the data.