We fetched 1,200 pages of the public site and found 58 PDF documents linked from them. The crawl stopped at our page limit, so documents linked deeper in the site are not counted.
We downloaded a random sample of 58 of them. 41 were still online, and of the 39 we could open, 7 (17.9%) had no tag tree, which is the structure a screen reader needs to read a document in order. 2 were pictures of pages with no text layer at all.
As an independent check, 39 documents from the sample were also validated with veraPDF against the full PDF/UA-1 standard; 100% failed at least one rule. That number is always higher than ours: a document can have a tag tree and still fail on reading order, table structure or missing text alternatives. The grade uses the stricter, simpler fact, no tag tree at all.
Crawl stopped at the page cap after 1200 pages; documents linked deeper in the site are not counted.
Of the documents missing structure, about 71.4% look, from their names and location on the site, like forms, notices and guides that residents use today. Those are the ones the Title II rule reaches first. The rest are largely archives, which the rule exempts unless they are still used to obtain a service.