State, city and university websites, one year from the Title II deadline

Of the 191 state, city, county and public higher-education homepages that returned a real page:

• 77% had at least one serious accessibility failure, a critical or serious WCAG issue.
• 20% had at least one critical issue, the kind that can hard-block a screen-reader user.
• 13% passed every automated check, about one in eight.
• The median site had 9 detectable issues; the worst had 141.

The spread matters more than the median here. 24 sites were clean, 53 had between one and five issues, and 15 had more than fifty. Most of this sector is close to the line rather than far from it, which is a different problem from the one online retail has.

The DOJ rule published in 2024 applies Title II of the ADA to web content and mobile apps, and names WCAG 2.1 Level AA as the technical standard. An interim final rule published in the Federal Register on 20 April 2026 moved both compliance dates back by a year:

• 26 April 2027 for public entities with a total population of 50,000 or more.
• 26 April 2028 for entities under 50,000, and for any special district government.

Every city, county and state in this sample serves well over 50,000 people, so the whole population here is working to the 2027 date. That is roughly 19 months from this scan. The deadline guide covers who is covered, the five exceptions and their conditions, and the separate HHS Section 504 dates for federally funded health programmes.

Splitting the sample three ways shows where the work is:

A serious failure is about as likely anywhere in this sector. The severe end is where they separate: a city or county homepage is more than three times as likely to carry a critical issue as a state portal, and carries a higher median issue count with it.

That ordering is not surprising once you look at what these sites are. A state portal is usually a small, heavily designed front door onto a search box and a list of services. A city or county homepage carries news, events, payment links, alerts and elected-official pages, often across several vendor systems, and every one of those is a place for an unnamed icon button or a low-contrast badge.

Ranked by the share of sites affected:

• Colour contrast, 52% (742 elements). WCAG 1.4.3. Half this sector fails the most basic visual check there is. How to fix it.
• Touch target size, 18% (197 elements). The 24 by 24 pixel minimum added in WCAG 2.2, which is not in the 2.1 standard the rule names, so this one is early warning rather than a compliance gap. What the rule asks for.
• Contrast that breaks on focus, 15% (157 elements). More on this below.
• Links with no discernible text, 12% (102 elements). How to fix it.
• Images missing alt text, 8% (56 elements).

Advisory issues, which are not WCAG failures, were more widespread still: content outside a landmark region on 45% of sites and non-unique landmarks on 34%.

The practical read is that this is mostly cheap work. Contrast, link names and alt text are the three most common failures and none of them requires rebuilding anything. Eight sites failed only on target size, meaning a rule that is not even part of the standard they are being measured against was the single thing standing between them and a clean automated result.

28 sites, 15% of the sample, had a control whose text contrast passes at rest and fails once you tab to it. A button reads fine, you press Tab, and the label all but disappears at exactly the moment it is selected.

Two things make this worth separating out. First, 16 of those 28 sites had no ordinary contrast failure at all, so their palette is fine and only the focus styling is wrong. Second, no standard accessibility checker reports it: axe-core has two contrast rules and both measure the page as it sits, so this is a failure the tools most teams run cannot see. We added the check because of that gap, and the guide explains how it is measured.

For a sector whose users include a lot of people navigating by keyboard, and whose deadline is written into a federal rule, a failure that is invisible to the usual tooling is the one worth knowing about.

We ran the same engine over 94 federal government homepages earlier in 2026. On the measures that compare cleanly, state and local government is behind:

• At least one serious failure: 77% here against 67% federal.
• At least one critical issue: 20% against 18%.
• Median issues per homepage: 9 against 6.

The likely reason is the one the cross-industry comparison landed on: federal agencies build on a shared accessible component library, the US Web Design System, and sectors that build from shared components score better than sectors that hand-roll every page. State and local government has no single equivalent, and this sample runs across dozens of vendor platforms.

One caveat, stated plainly. The contrast figures are not directly comparable across those two scans. This one ran on a corrected engine that resolves text over gradients and background images, where the earlier scan left many of those as "needs review" and counted no failure. Target size is unaffected by that change and compares directly: 18% here against 21% federal.

Building the sample turned up something worth passing on. 20 of these entities refused an ordinary scripted request, returning 403s or refusing the connection outright, and would have been dropped as unreachable by most survey tooling. Scanned properly in a real browser, all 20 returned their homepage, including four state portals.

So a survey that counts a bot wall as an exclusion is measuring its own HTTP client as much as the sector. Of our 202 attempts, 11 were genuinely excluded: seven persistent 403s, a redirect loop, a connection reset, a protocol error and a timeout.

• Sample. 202 homepages attempted, 191 returned a real page, scanned on 15 September 2026: 46 state portals including DC, 61 city and county sites, and 84 public universities and community colleges. Public entities only, since Title II does not reach private institutions. Curated for recognisability and spread across regions and sizes, not random or population-ranked.
• Engine. axe-core against the WCAG 2.x A and AA rules in a real Chromium browser, after the page was scrolled and allowed to settle, plus our own resolution of contrast over gradients, background images and translucent overlays, and our focus-state contrast check.
• Standard. The rule names WCAG 2.1 AA. We report against WCAG 2.x A and AA, which is a slightly wider net: target size in particular is a 2.2 criterion and is reported separately above for that reason.
• Scope. Homepages only, at a desktop viewport. Permit applications, payment portals, course registration and benefits flows sit behind the homepage and are where the harder problems usually live.
• As loaded. Pages were scanned as they rendered, including cookie and consent overlays, which can both add and mask issues.
• Not a compliance audit. Automated testing covers roughly a third of the WCAG success criteria, so these rates are a floor rather than a verdict. An automated failure is not a legal determination, and nothing here says any specific entity will or will not meet its obligations. We report aggregate figures only and name no organisation.