Guide
How to check if a PDF is accessible
An accessible PDF is one a screen reader can read in a sensible order, with headings, tables, pictures and form fields that make sense when heard. Checking one takes a tool that tests the structure and a person who judges whether it is right.
What to look for
The W3C's PDF techniques for WCAG describe what assistive technology relies on. In plain terms:
- Tags. A tagged PDF carries a structure tree saying which text is a heading, paragraph, list, table or figure. Reading order follows the tags; in a file that is not properly tagged, a screen reader may read a two-column page straight across both columns. If a file has no tags at all, start with what an untagged PDF means.
- Headings. Heading tags (H1 to H6) let a screen reader user jump between sections instead of listening from the top.
- A title, shown as the title. Set in the document properties, with the viewer setting that shows it instead of the file name.
- A document language. With the language set, a screen reader can load the right pronunciation rules.
- Real text. A scanned page is a picture of words; until it goes through optical character recognition (OCR), a screen reader has nothing to read. See scanned PDFs and OCR.
- Alternative text and artifacts. Meaningful images need a text description; purely decorative ones can be marked as artifacts so assistive technology skips them.
- Tables with header cells. Header cells (TH) and data cells (TD), so a header can be heard with its data.
- Labelled form fields. A screen reader announces a field by its name, which in Acrobat is the field's tooltip, not its internal field name.
- Bookmarks in long documents. An outline lets readers move through the structure without paging.
Tools that check the structure
Adobe Acrobat Pro: Accessibility Checker
Adobe says Acrobat Pro includes the Accessibility Checker "to assist authors in evaluating and fixing issues", and that Reader and Standard lack the complete set of tools. Acrobat Pro is a paid product; if your office already has it, its checker is a natural first stop.
PAC (PDF Accessibility Checker)
PAC is free, runs on Windows 10 and 11, and checks many PDF/UA and WCAG requirements at the press of a button. Its screen reader and structure preview lets a sighted person see what a screen reader will read, and in what order. It has been hosted by axes4 since 2023; there is no Mac version.
veraPDF
veraPDF is an open-source validator for PDF/A and PDF/UA. It validates against PDF/UA-1 and PDF/UA-2, runs on Windows, macOS or Linux with Java 8 or later, with a desktop app and a command line for batches. More in PDF/UA and veraPDF, explained.
A screen reader pass
NVDA is a free, open-source screen reader for Windows. Open the PDF and listen: is the title announced, can you move by headings, is the order right, are table headers read with their cells, do images and form fields make sense? The W3C techniques use this kind of listening as their test procedure.
What no automated check can judge
veraPDF's documentation is explicit that for PDF/UA it performs only the machine-verifiable checks; the checkpoints that need human judgement are not validated. PAC describes its preview as support for "the required visual check". A program can confirm that a structure exists, not that it is right:
- whether text tagged as a heading really is a heading, at the right level;
- whether the reading order follows the meaning of the page;
- whether alternative text says what the picture is for in this document;
- whether the right cells are marked as table headers;
- whether a form field's label matches the question printed next to it.
A clean checker result is a good sign, not a verdict.
A practical order of work
- Run a quick structural check to sort files into "basically sound" and "needs work".
- Run PAC or veraPDF on the files you will keep posting, and listen to the most-used ones.
- Where you have the source file, fix it there and re-export with tagging turned on.
With hundreds of files, triage first: see how to remediate many PDFs.
The quick check on our home page
The free quick check reads your PDF in your browser (with pdf.js); the file is never uploaded. It accepts files up to 25 MB and examines up to 150 pages (it counts the rest and says so). It reports whether the file is tagged (with structure and headings on its pages), whether a language and a displayed title are set, which pages have no text (likely scans), how many form fields lack a label, whether images are tagged as figures with alternative text (best effort), and, from 20 pages, whether there are bookmarks.
It does not run the PDF/UA rule set, and it cannot judge whether headings, reading order, table headers or alternative text are right. It cannot tell a decorative image from an untagged one. Files that need a password to open are refused with a plain message.
Sources
- W3C: PDF3, correct tab and reading order in PDF documents
- W3C: PDF9, headings in tagged PDF
- W3C: PDF18, document title
- W3C: PDF16, default language
- W3C: PDF7, OCR for scanned text
- W3C: PDF4, decorative images as artifacts
- W3C: PDF6, table elements
- W3C: PDF12, form field names
- W3C: PDF2, bookmarks in long documents
- Adobe: PDF accessibility overview
- PAC: PDF Accessibility Checker and system requirements
- PAC: history
- veraPDF documentation: getting started
- veraPDF documentation: validation profiles
- NV Access: About NVDA
Check a file now
The free quick check gives you the structural picture without uploading anything. If you then want files repaired, Pagorin's Starter plan repairs the structure of 100 pages a month for US$79, runs veraPDF on each result, and lists what still needs a person.