What PDF/UA is
PDF/UA, short for "PDF / Universal Accessibility", is ISO 14289. Part 1 (PDF/UA-1, ISO 14289-1) was first published in 2012 and is the version almost everyone means. It defines how the inside of a PDF has to be put together so that assistive technology can use it reliably: what must be tagged, how, and which metadata must be present.
It is a standard for files and for the software that makes and reads them. It doesn't tell you how to write accessible content. It tells you how a PDF must represent the content you have written.
PDF/UA versus WCAG and Section 508
| WCAG 2.x | PDF/UA-1 | Section 508 | |
|---|---|---|---|
| What it is | Content guidelines from the W3C | ISO file-format standard | U.S. federal law and standards |
| Covers | Any digital content: web pages, documents, apps | PDF files only | Federal ICT, including documents |
| Focus | Outcomes a user experiences: alt text, contrast, structure, meaning | Technical correctness of tags, metadata and fonts | Adopts WCAG 2.0 AA for documents |
| Colour contrast | Yes | No | Yes (through WCAG) |
The two standards complement each other. A file can pass PDF/UA and still fail WCAG, for example with light grey text or alt text that says "image". A file can also look fine to a WCAG reviewer while its internal structure breaks in some screen readers. Section 508 requires WCAG conformance; it doesn't name PDF/UA, but PDF/UA is the most rigorous way to verify the structural half, and many contracts now ask for both.
What PDF/UA-1 requires
The main requirements, in plain terms:
- A tagged PDF, with the document marked as tagged and a complete structure tree.
- All real content tagged with the right semantic type: headings, paragraphs, lists, tables, figures, links, notes. All decorative content (page numbers, running headers, rules, background art) marked as artifacts. Nothing can be neither.
- A logical reading order in the tag tree.
- Alternative text on every figure, and on other content that can't be read as text, such as formulas.
- Headings nested properly, without skipped levels.
- Tables with header cells (
TH) and a regular structure. - Lists tagged as
L/LI/Lbl/LBody. - A document title in the metadata, displayed in the window title bar instead of the file name.
- The natural language specified for the document.
- Embedded fonts with Unicode mappings, so every character can be read, searched and copied correctly.
- Annotations (links, form fields) tagged and given descriptions.
- No meaning carried by colour, layout or flicker alone.
- No security settings that block assistive technology from reading the content.
- A PDF/UA identifier in the file's XMP metadata, declaring that it claims conformance.
The Matterhorn Protocol
The standard itself is written for software developers. The
Matterhorn Protocol, published by the PDF Association,
translates it into a checklist of 31 checkpoints, each broken into specific
failure conditions. Some can be tested by software, for example "a
Figure tag has no alternative text". Others need a person, for
example "the alternative text does not describe the figure". Every serious
PDF/UA checker is built on it, and it is the clearest public description of
what "PDF/UA compliant" actually means.
How to check a file
- PAC (PDF Accessibility Checker): free, Windows, the most widely used PDF/UA checker. It also offers a screen-reader preview and a separate WCAG-oriented report.
- veraPDF: open-source validator for PDF/UA (and PDF/A), useful for batch and automated checks.
- Acrobat Pro: its accessibility checker overlaps with PDF/UA but is less strict. Its tag and order panels are the usual tools for fixing what a checker finds.
Passing a machine check covers only the machine-testable conditions. The human ones, such as accurate alt text, sensible heading levels and the right reading order, still need a reviewer. See the accessible PDF checklist for the full manual review.
PDF/UA-2
PDF/UA-2 (ISO 14289-2) was published in 2024. It is built on PDF 2.0 rather than the older PDF 1.7, uses the newer PDF 2.0 tag set, and was developed together with a "well-tagged PDF" specification. It doesn't replace PDF/UA-1 overnight. Most authoring tools, checkers and contracts still target PDF/UA-1, and a PDF 1.7 file can only conform to part 1. Unless a requirement names PDF/UA-2, PDF/UA-1 is the target.