The evidence
We scanned 676 randomly selected governments. 97.8% failed.
We publish our method and our numbers because we ask you to trust the report we send you. Everything below comes from scans we ran in September 2026, not from a vendor survey.
By type of government
| Type | Scanned | Failing | Mean issues |
|---|---|---|---|
| School districts | 259 | 99.6% | 17.0 |
| Special districts | 192 | 97.4% | 13.4 |
| Cities and towns | 138 | 94.2% | 13.3 |
| Townships | 54 | 98.1% | 14.4 |
| Counties | 33 | 100.0% | 14.0 |
Size barely predicts anything. Large counties fail at about the same rate as small townships, and only 15 of 676 sites were clean. This is an ordinary problem, not an exceptional one.
The documents are worse than the websites
We pulled 213 published documents — agendas, minutes, policies and notices — totalling 1,968 pages, and checked each for extractable text, structure, language, a document title, image descriptions and table headers.
94% failed at least one check. The largest single group is not the dramatic one — it is ordinary agendas exported from Word that are technically structured and still have no title and nothing to navigate by.
What this sample is, and is not
The website sample is a seeded random draw across all 50 states and every type of local government. The document sample is smaller and drawn from one state, so we treat those rates as provisional and say so. Passing our checks does not mean a document is conformant — reading order, whether a description is useful, and meaning carried by colour can only be judged by a person.