The PDF accessibility compliance guide for local governments
Most guides like this one sit behind a form: give up your email, wait for a download, get a PDF full of general advice. This one doesn't. We ran automated accessibility testing on 481 real documents already published by Florida cities and counties: agendas, minutes, budgets, and annual financial reports. Here's exactly what we found, and what it means for your documents.
By Wesley Black, founder of PairedUp · Published July 20, 2026 · Updated October 22, 2026
What "compliant" means, and by when.
The DOJ's rule under Title II of the ADA sets WCAG 2.1 Level AA as the standard for state and local government web content, including PDFs published on your website. The compliance dates depend on population:
Cities, counties, and other public entities of 50,000 or more people.
Smaller public entities and special district governments.
There are exceptions, including one for certain archived content, but they're narrow. A resident can still ask for an accessible version of an "archived" document, and you still owe one. The DOJ moved both deadlines back a year in April 2026 without changing the standard. There's runway, but no reason to wait for it.
We didn't write this guide from theory. We tested 481 documents.
Across cities and counties already publishing agendas, minutes, budgets, and annual comprehensive financial reports (ACFRs), here's what automated accessibility testing found.
Fail outright
477 documents had at least one failed PDF/UA machine check, and 367 had no tag structure at all. To a screen reader, an untagged file is a stream of text with no headings, tables, or order. This is the default output of most publishing tools.
"Tagged" still fail
Someone, or some tool, tried. A tag tree that exists is not a tag tree that works. Of the 114 tagged documents in the study, 111 still failed automated testing.
Files named "ADA" fail too
Files like Budget-Book-ADA-Compliant.pdf, posted in good faith as the accessible version, still fail. Failure counts ran from 5 to 1,293. Without verification, no one on staff can tell which version they actually published.
Your desktop checker disagrees
Showed five or fewer issues under Acrobat's accessibility rules while failing PDF/UA machine checks by the hundreds or thousands. The widest gap: 5 issues in one checker, 566,482 in the other. The two tools answer different questions.
Your flagship documents fail worst
All 249 ACFRs in the study failed, with a median of 52,548 machine-check failures each. The most scrutinized document a government publishes fails harder than routine agendas and minutes. The documents that get the most outside attention get the least accessibility treatment.
Progress doesn't carry forward
Of 95 governments publishing two or more years of their ACFR, 54 have a newer edition that fails worse than an earlier one, and 12 of those by ten times or more. Last year's fix didn't become the standard. Each year's file starts over from whatever the toolchain produces.
99% of the 28.17 million machine-check failures in the study trace to two PDF/UA requirements: content that was never tagged into the document's structure (clause 7.1), and text requirements such as a declared language (clause 7.2). These are two systemic problems repeated document after document, which makes them fixable at scale with a repeatable process.
Only 2 of 481 documents passed, and both were meeting agendas produced by agenda management systems. None of the 353 ACFRs and budget books passed PDF/UA machine checks. Read what that means for agenda packet accessibility, or see a county ACFR before and after remediation.
Results by document type
Financial documents fail hardest because they come out of finance and audit workflows that produce print-ready PDFs with no structure. Agendas fail less, because some agenda systems tag their output.
| Document type | Tested | Failed | Median failures |
|---|---|---|---|
| Annual reports (ACFR) | 249 | 249 | 52,548 |
| Budget books | 104 | 104 | 7,278 |
| Agendas and minutes | 101 | 97 | 1,779 |
| All documents | 481 | 477 | 36,336 |
The remaining 27 documents are other public records such as ordinances and budget supplements. Median failures are for failing documents only.
What we collected
481 live, publicly posted PDFs from the official websites of 101 Florida local governments (61 cities and 40 counties), collected June to July 2026. We focused on the documents the rule weighs most: annual comprehensive financial reports, budget books, agendas, and minutes. This was a targeted sample of Florida jurisdictions, not a random national one.
How we tested
Every file was run through veraPDF 1.30.2 with the PDF/UA-1 profile, and through PairedUp's implementation of Adobe Acrobat's published accessibility rules (not Adobe software). A document "fails" if it has at least one failed PDF/UA machine check. No file was altered. 2 files could not be parsed and are counted as neither passing nor failing.
What the numbers do and don't mean
A failure count is a count of failed machine checks, not of separate legal violations, and one missing tag can trigger thousands of them. A failed check is a test result, not a finding that a government is out of compliance with the ADA rule. Passing every machine check doesn't by itself establish WCAG 2.1 AA conformance, which also needs human review. No individual government is named for a failing result.
Download the data
All 481 documents, one row each: florida-government-pdf-study-2026.csv. Government names, titles, and links are removed, and rows are sorted by document type so they don't group by government.
- doc_id: row id for this study
- entity_type: city or county
- doc_type: acfr, budget, agenda (agendas and minutes), or other
- fiscal_year: year in the document's title, when it has one
- pages: page count
- tagged: whether the file has a tag structure at all
- pdfua_failed_checks: failed veraPDF 1.30.2 PDF/UA-1 checks; blank means the file could not be parsed
- acrobat_rule_failures: failures under our implementation of Acrobat's published rules
- result: fail (one or more failed PDF/UA checks), pass, or unscorable
- ada_in_filename: whether the published filename says "ADA"
- collected_date: date we downloaded the file
Cite this research
Black, W. (2026). Florida local government PDF accessibility study: 481 documents from 101 cities and counties (methodology v1.1, updated October 22, 2026). PairedUp. https://pairedup.ai/accessibility/pdf-compliance-guide/#data
v1.0 (July 20, 2026): 275 documents from 52 governments. v1.1 (October 22, 2026): expanded to 481 documents from 101 governments, same tests.
You can't fix everything in week one. Fix these first.
Prioritize by exposure, not by convenience:
1. Your ACFR and budget book
Highest scrutiny, worst failure rate, and the documents an auditor, reporter, or advocacy group downloads first. Fixing your flagship financial documents is also where the "tagged ≠ passing" gap is most visible if someone checks.
2. Anything already labeled "ADA" or "accessible"
A mislabeled document is worse than an unlabeled one, because it tells a resident, and a reviewer, that the work is already done. Verify those first; they're your highest legal exposure per document.
3. The weekly stream, not just the backlog
Agenda systems and department uploads publish new PDFs every week. A one-time cleanup without a way to catch new documents is a backlog that starts refilling the day you finish it.
4. Whatever your traffic tells you
If you have web analytics on document downloads, use them. The guide above is a starting order for governments with no data yet — your own usage numbers should always win.
A guide tells you what's wrong. This is how we fix it.
If you'd rather not build this in-house, this is the process PairedUp runs for local governments end to end.
Audit
We inventory the documents already published on your website and test them with automated accessibility tooling. You get a prioritized map of what fails and what residents use most.
Remediate
We rebuild the document's structure so assistive technology can read it: real headings, real tables with working headers, described images, and correct reading order. The pages render exactly as they did before.
Verify twice
Every document is checked two ways: PDF/UA machine checks with veraPDF, plus our implementation of Adobe Acrobat's published accessibility rules, the same checker many government offices use. You can re-run either with tools you trust.
Publish back
Each fixed document comes back ready to replace the original at its existing web address, so links, bookmarks, and search results keep working. One click per document for your team, or give us a limited login and we handle the uploads too.
Confirm live
After the swap we re-test the copy actually being served from your website. A document is marked complete only when the public file is the passing file.
Stay fixed
We keep watching your site. New documents are flagged and queued automatically, including the weekly stream from agenda and meeting systems.
Automated testing is necessary. It isn't sufficient.
Reading order, image descriptions, color contrast, and a handful of other checks still require human judgment, ideally from someone who knows the document. We flag which pages need a look, and our free review checklist shows your staff how to do it with free tools. When we say a document passes, we mean it passes the tests we name, and you can re-run those tests yourself.
Every document ships with its proof
Before and after results from automated accessibility testing, plus a private status portal your team can hand to a council member, an auditor, or anyone who asks.
Who this guide is for
City and county clerks, communications teams, ADA coordinators, and IT directors: the people who own the website, the records, and the deadline.
Zero homework to start
No inventory to compile, no files to send. Ask us for a private demonstration built from your own government's published documents. We prepare it before we ever meet, at no cost, from public records.
PDF accessibility questions, answered.
How can I check whether a PDF is accessible?
Start with two free automated checks: veraPDF for PDF/UA machine checks, and the full accessibility check in Adobe Acrobat Pro. Then open the file in a free screen reader such as NVDA and listen to a page with a table. Automated checks find structural failures. Listening finds the ones a person would hit. Our review checklist walks through it step by step.
Why does Acrobat say my PDF passes when other checkers fail it?
They test different things. Acrobat's checker runs a fixed set of about 30 rules, many of them at the document level. PDF/UA machine checks test whether each piece of content is tagged and described. In our study, 138 documents looked nearly clean under Acrobat's rules while failing PDF/UA machine checks by the hundreds or thousands.
What is PDF/UA, and how does it relate to WCAG 2.1 AA?
PDF/UA (ISO 14289) is the international standard for accessible PDF files. WCAG 2.1 AA is the standard the ADA Title II rule actually requires. They overlap heavily, and PDF/UA's machine checks are the most rigorous automated way to test a PDF's structure, but meeting WCAG 2.1 AA also takes human review of things like reading order and alt text.
What are the most common PDF accessibility failures?
Content that was never tagged, and text with no declared language. Those two account for 99% of the machine-check failures in our study. After that come tables without header cells, images without alt text, and a reading order that jumps around the page.
Why do annual financial reports fail worst?
ACFRs are long, table-heavy, and usually produced by finance and audit software that exports a print-ready PDF with no structure. Every financial table needs header cells tied to every number, which is the hardest part of remediation to do correctly. See a county ACFR before and after.
Should we convert our PDFs to HTML instead?
For simple content, often yes. Federal guidance prefers HTML where it works. But audited financial statements, budget books, and official agenda packets usually have to stay PDFs, and any PDF you post has to be accessible. A separate accessible version is allowed only when the original can't be made accessible for technical or legal reasons.
Still no form. Just a straight answer.
30 minutes: how big the job really is for your government specifically, what it takes to fix and stay fixed, and one clear next step. Or tell us your city or county and we'll run the same testing shown above against your own published documents first — before you talk to anyone.
Book 30 MinutesPick a time. That's it.