We scanned 3,000 municipal websites. Not one passed.

In August 2026 we ran automated accessibility checks on the public PDF documents of 3,044 U.S. city and town websites, drawn from the federal government's official .gov registry. This is what we found, with the ADA Title II compliance deadlines of April 2027 and April 2028 approaching.

The headline numbers

The forms problem

The finding that matters most for compliance planning: fillable forms fail at almost the same rate as everything else — 95.6% (9,013 of the 9,429 forms we analyzed).

That matters because of how the rule's most-cited safety valve works. The pre-existing content exception spares documents posted before a government's compliance date — unless they are used to apply for or access a service. A permit application, a license form, a records request: those are covered no matter when they were posted. The one category of document that cannot ride out the deadline is failing at 95.6%.

One in five documents is invisible

19.4% of the documents we analyzed — 7,162 files — are scanned images with no text layer at all. To a screen reader these documents are not badly structured; they are silent. A resident using assistive technology opens the town's fee schedule or permit form and hears nothing. These are also the most expensive documents to fix, because text recognition has to run and be verified before accessibility tagging can even begin.

By state

Failure rates are remarkably uniform. This is not a problem of a few laggard regions; it is the default state of municipal documents everywhere we looked.

Colorado deserves its own note: unlike the rest of the country, Colorado municipalities are not waiting on the federal deadline. The state's own accessibility law (HB21-1110) has been in effect since July 2025, carries statutory damages of $3,500 per violation, and gives residents a private right of action. The 300 broken forms we found on Colorado municipal sites are not a future liability; they are a present one.

“If everyone is failing, am I safe in the crowd?”

It is the natural reading of these numbers, and it is wrong, for three concrete reasons.

Enforcement is individual, not collective. ADA lawsuits and complaints are filed against one government at a time, and the plaintiffs who file them by the thousands each year pick their targets from whoever is visibly failing. Being part of a failing majority does not dilute your exposure; it is precisely what makes you a candidate. When the deadline arrives, “everyone else was failing too” is not a defense the rule recognizes.

In Colorado, the crowd is already being thinned. The state law in effect since 2025 carries $3,500 per violation and lets any resident bring the claim. A single municipal website with 40 broken forms is not looking at an abstract future risk.

The crowd is also the opportunity. Because nearly everyone fails, a municipality that fixes its documents doesn’t reach parity — it vaults into roughly the top 4% overnight, with a dated record showing it acted before the deadline. In a compliance regime, moving early is cheap insurance; moving after a complaint is expensive triage.

Why this is fixable

The encouraging part of a 96% failure rate is what the failures actually are. The most common problems we found — missing document titles, undeclared language, unlabeled form fields, absent image descriptions, missing tag structure — are mechanical, well-understood defects with established fixes. Most documents do not need to be rewritten; they need to be repaired. The exception exists too: the rule's pre-existing content provision means a meaningful share of older documents never need to be touched at all, if someone sorts out which ones qualify.

Methodology and honest limitations

Between August 28 and August 30, 2026, we crawled the public websites of municipalities listed as cities and counties in CISA's official .gov registry, reading up to 35 pages per site and analyzing up to 25 publicly linked PDFs each, in memory, storing nothing. Each document was checked for tag structure, headings, image descriptions, declared language, document title, form field labels, and scanned pages with no text layer — the machine-checkable subset of WCAG 2.1 AA.

Limitations, stated plainly: automated checks catch what a machine can see and nothing more; a document that passes every automated check can still be unusable in practice, so if anything these figures understate the problem. Our crawler could not reach documents on roughly two-thirds of sites (heavily scripted sites, bot protection, or genuinely document-free sites), so the sample skews toward municipalities with conventional websites. Per-document failure counts are capped by our 25-document sample per site. We do not name individual municipalities in this report.

The scan is ongoing; figures cover the first 3,044 municipalities of 11,646 in the registry and will be updated as coverage grows.

Citing this study: you are welcome to reference these figures with a link to this page. For questions about the data or methodology, write to reports@pdfaccessibilitycheck.com.