Your PDF Backlog Is a Digital Product. Start Treating It Like One.

The question I keep getting is some version of: how many pages can we remediate before the deadline? It is a fair question and it has a depressing answer. Published manual remediation runs $5 to $25 per page. A thousand-document library is not going to be finished by spring, and everyone in the room already knows it.

So people freeze, or they buy a tool, or they pick the newest hundred files and hope nobody notices the rest.

There is a better first move, and WCAG-EM 2.0 (opens in new window) just made it easier to defend.

Documents are now first-class in the methodology

WCAG-EM 1.0 was built around websites. Its unit of evaluation was the web page, defined around a single URI. If you tried to apply it to a folder of PDFs, you were improvising.

2.0 replaced that vocabulary. The methodology now runs on digital product, view, sample, and sample set — and its own glossary lists documents, PDF and Word and EPUB, as examples of digital products. It gives specific handling for them: scope to the whole document or to parts of it depending on complexity, and where no URL is available, identify samples by document title and filename.

Our balanced-lens overview, WCAG-EM 2.0: What a Better Audit Methodology Actually Changes, covers what this does for audit quality generally. What I want to talk about is narrower: what it does for the pile of PDFs you have to answer for in eight months.

It means your document library is a thing the methodology knows how to evaluate. You can sample it. You can defend the sample. You can put the sampling rationale in a report and hand that report to a plaintiff's attorney.

Be clear about what sampling does and does not buy you

I want to say this before the practical part, because it is the place people will overreach.

Sampling a document library gets you an evaluation. It does not get you a conformance claim. WCAG-EM says so directly, in both versions: you cannot claim conformance for a whole product based on a subset. The DOJ rule does not care about your representative sample either. It defines conventional electronic documents — PDF, word processor, presentation, spreadsheet — as web content, and the covered ones have to be accessible. Not sampled. Accessible.

So sampling is not a shortcut around remediation. It is a diagnostic instrument that tells you which remediation to buy, in what order, and whether the money you are about to spend fixes the cause or just the symptom.

That is worth a great deal in a year with no slack in it.

How to actually sample a document library

Map the methodology's exploration steps onto documents and it becomes concrete fast.

Common views. For a website these are the home and login pages. For a library, they are the documents everyone lands on: the current fee schedule, the main application form, this year's budget, the meeting agenda template.

Essential functionality. Anything a resident uses to accomplish something. Fillable forms belong here and they are the hardest category to remediate, which is exactly why they should be found in week one rather than March.

Types of samples. This is the step that pays for the whole exercise. Sort by how the file was made, not by what department owns it. Scanned image with no text layer. Born-digital export from Word. Export from InDesign. Multi-column newsletter. Table-heavy financial report. Signed form. Each of those is a different failure profile and a different remediation cost.

Technologies relied upon. 2.0 names PDF and EPUB explicitly here, and encourages you to record the authoring tools and templates behind the content. For a document backlog, that is the whole ballgame. The CMS, the Word template, the InDesign file, the report generator that spits out a thousand untagged pages a year.

Other relevant samples. Your accessibility statement, your grievance procedure, your request-an-alternate-format instructions. If those are inaccessible PDFs, fix them this week regardless of what the rest of the sample says.

Then add the random set — 10% of the structured set, on top of it, not carved out of it — and see whether it surfaces anything the structured set missed. If it does, your categories were wrong and you go back a step.

Two details that will come up

Set your conformance target to WCAG 2.1 Level AA. WCAG-EM 2.0 is written with WCAG 2.2 references throughout, but Step 1.2 has you select the target yourself. The DOJ Title II rule adopts 2.1 AA. Pick 2.1 AA, write it into the scope document, and note the 2.2 criteria separately if you want the forward view. Do not let a vendor quietly evaluate you against a standard the regulation does not impose, and do not let one claim 2.1 AA is obsolete either.

Product enclosure applies to documents too. The methodology has a named principle: you do not carve pieces out from inside your scope boundary, because excluding them conflicts with the conformance requirements for full pages and complete processes. It has been in the methodology since 2014 under a different name. When a department asks to keep its archive out of the review, that is the language you need. And remember the rule's preexisting-document exception is narrower than people think — it does not cover documents currently used to apply for, gain access to, or participate in your programs.

Transcript with visual descriptions

The PDF backlog problem

On screen: Slide showing the core problem: manual remediation costs $5 to $25 per page, making large backlogs impossible to finish by spring deadlines.

Manual PDF remediation runs $5 to $25 per page. A thousand-document library will not be finished by spring.

There is a better first move, and it starts with treating your document library as a digital product.

WCAG-EM 2.0 now covers documents

On screen: Slide explaining that WCAG-EM 2.0 replaced web-page-only vocabulary with terms like digital product, and explicitly lists PDF, Word, and EPUB as examples.

WCAG-EM 2.0 replaced web-page-only vocabulary with broader terms like digital product.

Its glossary lists PDF, Word, and EPUB as examples. Your document library is now something the methodology knows how to evaluate.

What sampling does and does not do

On screen: Two-column slide. Works column: sampling gives you an evaluation and shows which remediation to buy and in what order. Does not column: sampling does not give you a conformance claim; covered documents must be accessible, not just sampled.

Before the practical steps, one boundary is important to set clearly.

Sampling your library gives you an evaluation. It tells you which remediation to buy and in what order.

It does not give you a conformance claim. Covered documents must be accessible, not just sampled. The DOJ rule is clear on this.

How to sample a document library — four steps

On screen: Numbered list of four sampling steps: common views (high-traffic documents), essential functionality (fillable forms), types by how files were made, and technologies behind content such as authoring tools and templates.

Map the methodology's steps onto documents and it becomes concrete quickly.

Start with common views: the documents everyone lands on, like fee schedules, application forms, and meeting agendas.

Next, find essential functionality. Fillable forms are the hardest category to remediate, so surface them in week one, not March.

Then sort by how files were made: scanned image, Word export, InDesign, newsletter, financial report. Each is a different failure profile and a different cost.

Technologies and other relevant samples

On screen: Slide covering two more sampling steps: record authoring tools and templates behind content, and fix accessibility statements and grievance procedures first if they are inaccessible PDFs.

Record the technologies behind your content: the CMS, the Word template, the InDesign file, the report generator that produces untagged pages.

Fix your accessibility statement, grievance procedure, and alternate-format instructions this week, regardless of what the rest of the sample shows.

Add a random 10 percent set on top of the structured set. If it surfaces something new, your categories were wrong and you go back a step.

Set your conformance target correctly

On screen: Slide with two key details: set the conformance target to WCAG 2.1 Level AA as required by the DOJ Title II rule, and remember that product enclosure means you cannot exclude parts of your library from scope.

Two details will come up when you write the scope document.

Set your conformance target to WCAG 2.1 Level AA. The DOJ Title II rule adopts 2.1 AA. Write it into the scope and do not let a vendor evaluate you against a different standard.

Product enclosure means you cannot exclude pieces from inside your scope boundary. A department cannot opt its archive out of the review.

The preexisting-document exception is narrower than most people expect. It does not cover documents used to apply for or access your programs.

Government PDFs: 97% inaccessible

On screen: Stat slide showing that government PDFs test at roughly 97 percent inaccessible across large samples, pointing to a few broken pipelines rather than thousands of unique problems.

Government PDFs test at roughly 97 percent inaccessible across large samples.

That number is not describing thousands of unique problems. It is describing a few broken pipelines running at scale.

Fix the pipeline, then work the backlog

On screen: Slide explaining the strategy: use the sample to find broken templates and export workflows, fix those first, then work the backlog by program access — documents residents use to apply, pay, appeal, or attend.

A representative sample finds those pipelines.

Fixing a Word template or an export setting stops new inaccessible documents at a cost of a few staff days. No per-page vendor can sell you that.

Then work the backlog in order of program access: the documents residents use to apply, pay, appeal, or attend.

Automated remediation is not conformance. Complex tables, forms, and anything with legal consequence for the reader still need human hands.

Takeaway: start with the sample

On screen: Closing slide with the single takeaway: use WCAG-EM 2.0 to sample your document library, find the broken pipelines, and prioritize remediation by program access before the deadline.

Find the broken pipeline. Fix the source. Then work the backlog by program access. Start with the sample.

Sources: WCAG-EM 2.0 W3C Group Note, WCAG-EM 1.0, 28 CFR Part 35 Subpart H, and the Allyant PDF Accessibility Index 2025 to 2026.

Sources

• WCAG-EM 2.0, W3C Group Note, 23 July 2026
• WCAG-EM 2.0 background reading list
• WCAG-EM 1.0, 2014
• 28 CFR Part 35, Subpart H
• Allyant PDF Accessibility Index 2025–2026

What the sample is actually for

Here is the payoff. Nearly every document library I have evaluated fails in a small number of systemic ways, repeated thousands of times, traceable to a handful of templates and one or two export workflows.

Government PDFs test at roughly 97% inaccessible across large samples. That number is not describing thousands of unique problems. It is describing a few broken pipelines running at scale.

A representative sample finds the pipelines. Fixing a Word template and an export setting stops the bleeding for every document you publish from now on, at a cost of a few staff days. No remediation vendor can sell you that, because it is not a per-page service.

Then, and only then, work the backlog by program access — the documents residents use to apply, pay, appeal, or attend. That is where cost has been the binding constraint, which is why we built pdf.idioma.chat (opens in new window) at $0.50 per page: to make backlog work an operating decision instead of a capital project. And the same limit I always name applies. Automated remediation is not conformance. Complex tables, forms, and anything with legal consequence for the reader still need human hands.

One last detail, and it is my favorite thing in the new version. If you publish an evaluation statement, 2.0 says the statement and its accompanying report have to be published in an accessible format themselves. An inaccessible PDF describing your accessibility conformance is a real thing I have seen more than once.

Do not be that. Start with the sample.

More on the deadline: WCAG 2027: The Clock You Got Back Is Almost Gone

Sources: WCAG-EM 2.0, W3C Group Note, 23 July 2026 (opens in new window) — see also its background reading list (opens in new window) · WCAG-EM 1.0, 2014 (opens in new window) · 28 CFR Part 35, Subpart H (opens in new window) · Allyant PDF Accessibility Index 2025–2026 (opens in new window)