Accessible PDFs: the hardest part of Title II

The website part of a Title II project has a price, and it's usually reasonable. Then somebody says "what about the PDFs," and the number moves. A small city with a decade of council packets, budget books, ordinances, and forms holds thousands of documents, and remediating one properly takes a trained person anywhere from fifteen minutes to several hours.

That gap — cheap website, expensive documents — isn't a vendor squeezing you. It's a consequence of what a PDF is.

Why a PDF is harder than a web page

An HTML page carries its structure in the markup: an <h2> is a heading because it says so, a <th> is a table header because it says so, and a screen reader gets that for free.

A PDF does not work that way. At its core it is a set of instructions for placing glyphs at coordinates. Visually that's enough. Semantically there is nothing there — no headings, no reading order, no idea which cells in a table are headers. All of that lives in a separate structure called the tag tree, and it exists only if something put it there.

An untagged PDF read aloud is a bag of characters in whatever order the generator wrote them — in a two-column agenda, usually the two columns interleaved line by line.

What a compliant PDF actually needs

Real tags — headings as headings, lists as lists, tables with header cells identified. 1.3.1 Info and Relationships, Level A.
A correct reading order — the tag tree's, not the layout's. 1.3.2 Meaningful Sequence, Level A.
Alt text on every figure, with decoration flagged as artifacts so the letterhead seal isn't announced. 1.1.1 Non-text Content, Level A.
A title set and displayed in the file's properties instead of the file name. 2.4.2 Page Titled, Level A.
A document language. 3.1.1 Language of Page, Level A.

If the document is a scan, none of that applies yet: there is no text at all, only a picture of text, which fails 1.4.5 Images of Text, Level AA, on its own terms. It has to go through OCR and correction before tagging can start. Scanned minutes are the most expensive document type a city owns.

When you're buying this work, listen for PDF/UA — ISO 14289-1, the PDF-specific accessibility standard. A shop that talks in PDF/UA terms knows what a tag tree is. One promising to fix your PDFs automatically is describing Acrobat's autotag button: passable on a one-column memo, garbage on anything with columns.

The export setting that quietly destroys everything

Here's the one that costs cities the most, because it erases work you already did.

If you write in Word and produce the PDF through "Microsoft Print to PDF" — or any print-to-PDF driver — you get an untagged PDF. Every time. The print path places glyphs and has no concept of your headings, so all the structure you applied in Word is thrown away at the door.

You have to export, not print: File → Save As, choose PDF, open Options, and confirm "Document structure tags for accessibility" is checked. Google Docs and InDesign have the same split. Get the clerk's office onto the export path and every new document starts most of the way there.

Read the exceptions before you price the backlog

This is the part that changes the budget, and it usually runs opposite to where the panic is pointed.

The rule's exceptions, at 28 CFR § 35.201, include one for preexisting conventional electronic documents — a term the rule defines narrowly as PDF, word processor, presentation, and spreadsheet files. Documents already posted before your compliance date are excepted, unless they are currently used to apply for, gain access to, or participate in the entity's services, programs, or activities. A separate exception covers archived content, with conditions: created before the compliance date, kept only for reference or recordkeeping, not altered after archiving, and stored in a clearly identified archive area.

Read those together and the shape of the job inverts. Twelve years of council minutes, untouched in a section labeled as an archive, largely fall under an exception. The fillable building permit application from 2019 still linked off the Planning page — the one a resident has to use to get a permit — is squarely in scope, because that is what "used to apply for services" describes. So is the utility assistance form and the job application.

Get your attorney's read rather than mine. But the practical version is: stop pricing the archive and start with the forms a resident fills out to get something from you. There are usually a few dozen, not a few thousand. And note that editing an archived document is what stops it being preexisting.

What to do

Inventory the documents residents use to transact — applications, permits, requests, registrations. That list is short.
Convert what you can into web pages and real HTML forms — cheaper to make accessible than a PDF, and better for the resident.
Fix the export path in every office that publishes, and check the result with Acrobat Pro's accessibility checker or the free PDF Accessibility Checker from the PDF/UA Foundation.
Move dormant material into a clearly labeled archive, leave it alone, and remediate the rest least-used last.

If you'd rather have that inventory done for you — what's in scope, what an exception covers, and what the rest costs — start with a scan.