Find PDFs that fail accessibility checks

If you publish documents for a public body, you have a legal duty to make them accessible — and the hard part is rarely fixing one PDF. It’s knowing which of your two thousand PDFs are the problem.

FileDeck screens the whole library and lists them.

FileDeck AI tier — included in all AI plans. See pricing

Read this first: what this tool is

This is a screening tool. It is not a compliance check, and it cannot tell you a document is accessible.

Automated testing finds only a minority of accessibility problems. FileDeck can tell you a PDF has no tag tree. It cannot tell you the tags are wrong — that the reading order is scrambled, that a heading is tagged as body text, that the alt text says "image1.png", or that a table’s headers mean nothing to a screen reader. Those need a person.

So use this to find the documents that certainly need work, not to certify the ones that don’t. A document that passes every check here may still be unusable with a screen reader.

We’re explicit about this because the alternative is worse than having no tool: a report that let you believe a library was compliant would stop you looking.

What it checks, and why each matters

Go to Documents → Accessibility and choose Scan the library. The scan runs in the background, so you can leave the page.

Finding	What it means	Relates to
Scanned image, no text	The page is a picture. Nothing can be read aloud, searched, or reflowed.	WCAG 1.4.5, 1.1.1
Not tagged for screen readers	No structure tree, so a screen reader gets an undifferentiated wall of text — no headings, no lists, no table structure, no reliable reading order. Usually the most serious finding.	WCAG 1.3.1 · PDF/UA-1 7.1
No document language set	Assistive technology has to guess the language, so English can be read aloud with a French voice.	WCAG 3.1.1
No document title	The window or tab shows a filename instead of a name.	WCAG 2.4.2
Could not be checked	The file is corrupt, or encrypted in a way that blocks reading it.	Not a failure in itself

That last row matters more than it looks. FileDeck reports files it couldn’t read as unknown, never as passes. A report that quietly counted unreadable files as fine would be flattering and wrong.

Fixing what it finds

FileDeck reports; it doesn’t rewrite your PDFs. Remediation happens in whatever produced them:

• Scanned documents need re-creating from the source, or running through OCR and then being tagged. FileDeck’s OCR makes a scan searchable, which is genuinely useful — but it does not make it accessible, and this report will still flag it as a scan.
• Untagged documents are usually fixed at source: export from Word or InDesign with tagging enabled, rather than printing to PDF. Retro-fitting tags in Acrobat is possible but slow.
• Language and title are two fields in the document properties of whatever wrote the file — the cheapest wins on the list, and worth doing in bulk.

Re-run the scan after a batch of fixes to see the numbers move.

FAQ

Does this make my library WCAG compliant?

No, and nothing in FileDeck will claim it does. It finds a specific set of machine-detectable failures. Compliance is a judgement about the whole document made by someone who can assess reading order, alt text quality and structure — none of which a script can see.

What about Word, Excel and PowerPoint files?

Only PDFs are screened. Office files have their own accessibility checkers built into the applications that make them, which see far more than an external scan could. Documents that aren’t PDFs are skipped — and skipped is not the same as passed, so they aren’t counted in the totals either.

How long does a scan take?

It runs in the background in small batches, so it never times out a page. A library of a few hundred PDFs finishes in minutes; several thousand takes longer. You can keep working while it runs.

Does anything leave my server?

No. The screening is structural parsing done locally — it doesn’t call any AI provider, and no document content is sent anywhere. It’s in the AI tier because of who needs it, not because it uses a model.

Why is April 2027 the date everyone mentions?

The US Department of Justice’s ADA Title II rule sets WCAG 2.1 AA for state and local government web content, including PDFs — larger entities from April 2026 and smaller ones from April 2027. UK and EU public bodies have comparable duties already. If you’re procuring now, this is the deadline the questions are about.