How to remediate hundreds of PDFs: inventory, triage, fix, check

1. Inventory: list what you have

Start with a spreadsheet, one row per PDF. A crawl of your site, a content management system export or a folder listing will give you most of it. For each file record:

• the address (URL) and the page that links to it;
• title, date and page count;
• who owns it (department or person);
• how often it is opened, if your analytics can tell you;
• what kind of document it is: plain text, tables, a form, or a scan.

The last column matters for effort and cost. A scanned page with no text needs OCR before anything else (see Scanned PDFs and screen readers), and forms and data tables take the longest. A free checker can fill in some of this for you; the guide to checking a PDF lists the options.

2. Triage: decide what happens to each file

For every row, pick one of four outcomes:

1. Remove. Duplicates, superseded versions and files nobody has opened in years. Ask the owner before deleting, and keep a copy where your records rules require one.
2. Archive. Old material kept for reference, moved to a clearly labelled archive area. Offer an accessible version on request.
3. Convert to HTML. Short notices, FAQs and anything that changes often usually work better as a web page than as a PDF. A web page is easier to keep accessible and easier to read on a phone.
4. Repair. Documents people need in PDF form: forms that must be printed or signed, formal reports, documents with a fixed layout.

Some rules have exceptions for archived or pre-existing documents. The ADA Title II web rule, for example, has exceptions that are specific and conditional; our ADA Title II and PDFs guide and the deadlines page link to the official text. Decide whether an exception applies with your counsel or accessibility office, not from a summary.

3. Prioritise documents in current use

Within the "repair" pile, order the work by how much it matters to someone trying to get something done. The W3C's guidance on interim repairs suggests starting with key tasks and with high-impact items: problems on frequently used pages, problems that block a process, and the most basic (Level A) issues. For documents, that usually means:

• forms and instructions people need to apply for or use a service;
• documents linked from your home page or most-visited pages;
• current agendas, minutes, notices and fee schedules;
• anything new: fix the template so new documents are made accessible from the start, otherwise the backlog keeps growing.

Then work in batches of similar documents. Twenty council agendas made from the same template share the same problems and the same fixes.

4. Fix

Where you still have the source file, fixing it there and exporting a new tagged PDF is usually best: the repair survives the next edit. Where you only have the PDF, the choice is between doing it yourself in a PDF editor, remediation software, a service, or automated repair. The guide to remediation cost compares them. Whichever you choose, the basics are the same: tags, a title and language, real headings and lists, table headers, labelled form fields, alt text for meaningful images, and a sensible reading order. The untagged PDF guide explains what tags are.

5. Check

Check every repaired file in two ways:

• Machine check. veraPDF is open source and has a command-line version for running large batches. It reports PDF/UA machine checks only. PAC is a free Windows tool that also tests files against PDF/UA.
• Human check. A person confirms what a machine cannot judge: that headings are real headings, alt text describes the image's purpose, table headers are right and the reading order makes sense. Sample at least a few pages of each batch, and every form.

The PDF/UA and veraPDF guide explains why a checker pass is only half the answer.

6. Keep records

Add columns to the inventory for the outcome, the date fixed, the checker result and who did the human check. Keep a short note of known issues you have not yet fixed. Records like these help when someone asks for an accessible version of a document, when you report progress to management, and when the same document comes back next year with new numbers. Publish a contact address so readers can ask for an accessible format of anything you have not reached yet.

Sources

• W3C WAI: Web Accessibility First Aid, Approaches for Interim Repairs
• veraPDF: About
• veraPDF documentation: Getting started (GUI and command line)
• Virginia Tech TLOS: PAC (PDF Accessibility Checker)