Most organizations should remediate fewer PDFs than they think. Every document in your library needs a decision (remediate it, convert it to HTML, or retire it), and making that decision first is far cheaper than tagging two thousand files nobody opens.
PDF libraries grow by accretion: board packets, old newsletters, an application form from years ago that one page still links to. Remediating the whole pile to PDF/UA is slow, and most of the effort lands on documents with no readers. Converting everything to HTML has the opposite problem, because some documents exist precisely because their fixed layout matters. What works is triage, followed by fixes at the source so the next batch doesn't need rescuing.
Inventory before you tag anything
You can't triage what you haven't listed. Crawl the site for PDF links, including subdomains and any document hosting service you use, and pull twelve months of opens from analytics or server logs. Search engines send people straight to PDFs without passing through your pages, so check your search console data too. For a quick first pass on a site with a sitemap:
SITE="${1:?usage: ./pdf-links.sh https://your-website-address}"
curl -s "$SITE/sitemap.xml" | grep -o '<loc>[^<]*' | sed 's/<loc>//' |
while read -r page; do
curl -s "$page" | grep -oiE 'href="[^"]+\.pdf"' | sed -E 's/href="([^"]+)"/\1/'
done | sort | uniq -c | sort -rnThe output counts how many pages link to each PDF, which is a decent first proxy for importance. Put every document in a spreadsheet with these columns: URL, title, page count, last modified, opens in the last year, linked from, owner, needed for a service (yes or no), form (yes or no), and decision. The owner column is often the hardest to fill, and the most revealing.
Four questions decide most documents
- Do people need it to get a service? Applications, enrollment packets, permit requests, claim forms, payment instructions. If a document stands between a person and something they're entitled to, it should be a web page or web form. A well-tagged PDF form is still harder to complete on a phone or with a screen reader than a good HTML form. Keep a printable PDF alongside if people ask for one.
- How much is it used? Low use and no service role make a document a candidate for retirement or the archive.
- How long will it live, and how often does it change? Content that changes every year belongs in HTML, where there's one source and nothing to re-export. Content that is fixed and long-lived, where the page itself is the record, belongs in a remediated PDF.
- Is it a form or a reading document? Forms go to HTML almost every time. Reading documents depend on questions two and three.
A triage table for the usual suspects
| Document type | Example | Decision | Reason |
|---|---|---|---|
| Application or enrollment form | Program application, permit request | Convert to an HTML form | People need it to get a service |
| Policy or fee schedule updated yearly | Handbook, rate sheet | Convert to HTML | One source, no re-export |
| Designed publication | Annual report, research study | Remediate to PDF/UA, add an HTML summary | Layout matters; fix it in the source file |
| Official or signed record | Adopted ordinance, signed agreement | Remediate | The fixed version is the record |
| Superseded or duplicate content | Old newsletter, brochure repeating a web page | Retire and redirect | Nobody needs two versions |
| Historical reference | Minutes from past years | Archive, with accessible copies on request | Low use, clearly labeled, fast turnaround |
Apply the table document by document, but expect clusters. Once you sort by opens and by service role, whole categories tend to decide themselves, and the remediation budget goes to the files that earn it.
Remediate to PDF/UA, then check it the way people read it
The ISO standard for accessible PDF is PDF/UA. PDF/UA-1, published as ISO 14289-1:2014, covers files based on PDF 1.7. PDF/UA-2, published as ISO 14289-2:2024, covers PDF 2.0 and adds to PDF/UA-1 rather than replacing it. Most authoring tools still export PDF 1.7, so PDF/UA-1 is the usual target. WCAG still applies to a PDF published on your website; PDF/UA is the file-format standard that gets you most of the way there.
Remediation means a correct tag tree, not a passing checker score:
- Headings tagged H1 through H6 in a logical hierarchy, plus real lists and paragraphs.
- A reading order that follows the content, not the order objects were drawn on the page.
- Alternative text on meaningful figures, with running headers, footers, and decoration marked as artifacts.
- Data tables with header cells and scope, and no layout tables pretending to be data.
- A document title set to display in the title bar, and the document language set.
- Bookmarks for long documents, and form fields with names, descriptions, and a sensible tab order.
Check in layers. PAC, the free PDF Accessibility Checker, tests the machine-checkable PDF/UA conditions. Acrobat Pro's accessibility checker, Tags panel, and Reading Order tool help you find and fix structure. Then open the file in Acrobat Reader with NVDA running and read it the way a user would: straight through, then by headings, then table by table, then field by field in forms. A PDF can pass every automated check and still read like nonsense. Software can confirm that alt text exists, but not that "image1.png" is unhelpful. It can confirm that a table has header cells, but not that they are the right ones.
Fix it at the source, before export
Remediating an exported PDF is expensive, and it has to be repeated every time the document changes. Fixing the source file is cheaper, and the fixes carry forward.
In Word: use the built-in Heading styles rather than bold text, and real bulleted and numbered lists. Mark the first table row as a header row and set it to repeat across pages. Add alt text, or mark images as decorative. Set the document title and language, and run Check Accessibility on the Review tab. To export, use File > Save As and choose PDF. On Windows, open Options and keep "Document structure tags for accessibility" checked; on a Mac, choose the option for electronic distribution and accessibility. Never use "Print to PDF", which throws the tags away.
In InDesign: map paragraph styles to export tags, set the reading order in the Articles panel, add alt text through Object Export Options, mark decorative objects as artifacts, and export with "Create Tagged PDF" turned on. Set the title in File Info and choose to display it.
For scans: optical character recognition is where remediation starts, not where it ends.
Then fix the templates. A Word template with correct styles and an InDesign template with tag mappings mean every future document starts accessible, and the remediation budget goes to the backlog instead of to new work.
Convert, retire, or archive, and keep exceptions honest
When you convert, build the page in your CMS with real headings and tables, add a download link if people genuinely print the document, and put a 301 redirect on the old PDF URL so bookmarks and search results land on the new page. When you retire, remove the file, redirect to the current equivalent, and fix the internal links your inventory found.
Archiving is for documents kept for reference that nobody needs to complete a current task. Put them in a clearly labeled archive section, say that they are kept for historical reference, and offer an accessible version on request, with a named contact and a quick turnaround. Log the requests. If one document keeps being requested, it belongs back in the remediate pile.
Some regulations, including the Justice Department's ADA Title II rule for state and local governments, contain limited exceptions for certain archived content and preexisting documents. Those exceptions come with conditions, and they don't reach documents people use to apply for or get access to a current program or service. Treat them as a question for counsel, not as a filing strategy.
Where to start with your library
Run the inventory, sort it by opens and by "needed for a service," and decide your top hundred documents this month. That sort usually shows where the money should go. Our document and PDF accessibility team remediates to PDF/UA and checks reading order on every file, from $9 per page, and a digital estate audit finds the documents sitting on subdomains and microsites your crawl missed.