How we measured 615 public documents

We measured 615 public documents from 126 organisations. Not one met PDF/UA-1: every one of the 615 failed at least one of its rules. This page says how they were checked, how they were chosen, what was left out, and how to read the figure.

What was checked

Each document went through the same automated check as the free check on this site, without OCR: veraPDF 1.30.2, the reference open-source validator, against its PDF/UA-1 profile. A document passes only if it passes every rule in that profile. The checks ran between 27 September and 4 October 2026.

These are the rules a machine can check. Passing all of them would not on its own make a document accessible, since some requirements need a person's judgement. Failing any of them means the document does not meet PDF/UA-1.

How the organisations were chosen

We read published accessibility statements, and took an organisation only when its own statement says some of its documents are not accessible. The sample is therefore not random, and the result should not be read as a rate for organisations in general.

How the documents were chosen

• One document per organisation, chosen by our crawler from its public site: 117 documents.
• A sample from the document library of 16 organisations, taken where the statement admitted a problem or the first document failed: 498 documents.

What was left out

223 documents could not be checked and are not among the 615:

• 93 were scans with no text layer. This study does not run OCR, so there was no text to check.
• 80 could not be downloaded.
• 50 were refused by the site's robots.txt, and we did not fetch them.

How to read it

The documents are concentrated: 16 organisations hold 505 of the 615 (82%), and 110 organisations contributed one document each. A rate pooled over documents would mostly describe those 16 organisations. So when we report a rate, we cap each organisation at three documents.

The headline is not a rate. It holds for every one of the 615 documents, whichever way they are counted.

What we do not publish

No organisation is named, on this page or anywhere else. The purpose is to show the state of published documents, not to single anyone out.