How the WCAG success criteria apply to PDFs via chapter 10 of EN 301 549, which versions are authoritative – and what can and cannot be checked automatically.
What WCAG is
The Web Content Accessibility Guidelines (WCAG) of the World Wide Web Consortium (W3C) describe when digital content is accessible to people with disabilities. The current version, WCAG 2.2, has been a W3C Recommendation since 5 October 2023 (revised version of 12 December 2024). It is built on four principles:
- Perceivable: Information must be presented in a way that everyone can perceive – for example images with a text alternative, sufficient contrast.
- Operable: Everything must be operable, including with the keyboard alone – for example links and form fields in a sensible order.
- Understandable: Content and operation must be comprehensible – for example a specified language, labelled form fields.
- Robust: Content must be reliably readable by assistive technologies such as screen readers – for example correct structure and field properties.
Below the principles are guidelines, and below those testable success criteria at three conformance levels: A (basic), AA and AAA. In Europe, the legal requirement is generally level AA, which includes all criteria of levels A and AA. Compared with 2.1, WCAG 2.2 adds nine criteria and removes criterion 4.1.1 (Parsing). Content that meets WCAG 2.2 also meets WCAG 2.1 and 2.0.
How EN 301 549 applies WCAG to documents
WCAG is written for web pages. The European standard EN 301 549 (“Accessibility requirements for ICT products and services”) applies it to other content: chapter 9 covers web pages, chapter 10 non-web documents – including PDFs provided for download – and chapter 11 software.
- The numbers correspond: requirement 10.1.1.1 of the standard is WCAG 1.1.1, requirement 10.1.3.1 is WCAG 1.3.1, and so on.
- Some criteria are marked “Void” (not applicable) for individual documents because they relate to sets of web pages – for example 2.4.1 (Bypass Blocks) and 2.4.5 (Multiple Ways).
- “Page Titled” becomes “Document titled”; interpretation follows WCAG2ICT, the W3C guidance on applying WCAG to non-web content.
- A criterion whose subject does not occur in the document is deemed to be met: a PDF without form fields automatically meets the form criteria.
Which version applies?
- EN 301 549 V3.2.1 (2021-03) is the version listed as a harmonised standard under Directive (EU) 2016/2102, on which the presumption of conformity for public sector bodies is based. It refers to WCAG 2.1.
- EN 301 549 V4.1.1 (2026-09) was published by ETSI in September 2026. It is aligned with WCAG 2.2 and, for the first time, also contains an annex on Directive (EU) 2019/882 (European Accessibility Act). Whether and from when this version will be listed in the Official Journal of the EU as a harmonised standard and replace V3.2.1 had not yet been decided at the time of writing.
Practical consequence: if you test your documents against WCAG 2.2, level AA, you also meet the WCAG 2.1 requirements of the version currently in force and are prepared for the new version.
And where does PDF/UA come in?
WCAG and EN 301 549 say what a document must achieve – independently of the format. PDF/UA (ISO 14289) specifies how this is implemented technically in the PDF file format: tags, roles, metadata, fonts. EN 301 549 V3.2.1 does not mandate PDF/UA; the new version V4.1.1 names PDF/UA-1 and PDF/UA-2 in a note as the technical standards for accessible PDFs that help meet the standard. The two complement each other: a PDF can be PDF/UA-compliant and still incomprehensible – and conversely, WCAG can hardly be achieved in a PDF without sound PDF/UA technology. More in PDF/UA explained.
The W3C itself describes in the WCAG Techniques for PDF (PDF1 to PDF23) how individual success criteria can be met in PDFs – for example PDF1 (alt text via the Alt entry), PDF9 (heading tags) or PDF16 (document language). These techniques are informative: they show a recognised way, but not the only one.
The most important criteria for PDFs
The following overview shows the criteria on which PDFs most often fail in practice, with the matching requirement from chapter 10 of EN 301 549 and the honest answer to the question of whether a machine can assess it.
WCAG success criteria relevant to PDFs, the corresponding EN 301 549 requirement and automatic testability| WCAG criterion (level) | EN 301 549 | What it means in a PDF | Machine-testable? |
|---|
| 1.1.1 Non-text Content (A) | 10.1.1.1 | Images tagged as Figure with alt text; purely decorative elements as artifacts (techniques PDF1, PDF4). | Whether alt text exists: yes. Whether it describes the image accurately: no. |
|---|
| 1.3.1 Info and Relationships (A) | 10.1.3.1 | Headings, lists and tables are tagged as such; tables have header cells (PDF6, PDF9, PDF21). | Formal structure rules: yes. Whether the right tag was chosen: only a human can tell. |
|---|
| 1.3.2 Meaningful Sequence (A) | 10.1.3.2 | The order in the tag tree matches the logical reading order (PDF3). | No – tools can only flag anomalies. |
|---|
| 1.4.3 Contrast (Minimum) (AA) | 10.1.4.3 | Text at least 4.5:1, large text at least 3:1 against the background. | Partly: measurable for text on a plain background, unreliable for text on images and gradients. |
|---|
| 2.4.2 Page Titled (A) | 10.2.4.2 | Meaningful document title in the metadata, displayed in the window instead of the file name (PDF18). | Present and displayed: yes. Meaningful: no. |
|---|
| 2.4.3 Focus Order (A) | 10.2.4.3 | Links and form fields are reached with the Tab key in a sensible order (PDF3). | Partly: the coupling to the structure yes, whether the order makes sense no. |
|---|
| 3.1.1 Language of Page (A) | 10.3.1.1 | The main language of the document is specified, e.g. en-GB (PDF16). | Yes. |
|---|
| 3.1.2 Language of Parts (AA) | 10.3.1.2 | Passages in other languages are marked up with their language (PDF19). | Only to a limited extent – language detection provides hints, not certainty. |
|---|
| 3.3.2 Labels or Instructions (A) | 10.3.3.2 | Form fields have a visible label; required fields are recognisable (PDF5, PDF10). | Whether a label exists: yes. Whether it is understandable: no. |
|---|
| 4.1.2 Name, Role, Value (A) | 10.4.1.2 | Form fields have an accessible name (tooltip), a role and a readable state (PDF10, PDF12). | Largely yes. |
|---|
What can be checked automatically – and what cannot
For the technical side, the PDF Association's Matterhorn Protocol provides a precise list: of 136 failure conditions for PDF/UA-1, 87 can be detected by software alone, while 47 generally require human judgement. The same pattern applies to the WCAG criteria:
- Reliably by machine: Are tags present? Do images have alt text? Are language and title set, fonts embedded, tables formally correct, form fields labelled?
- By machine, with limits: colour contrast (reliable for text on a plain background, uncertain on images), indications of conspicuous reading orders or passages in other languages.
- Only by humans: Does the alt text describe the image accurately? Is the reading order logical? Do the heading levels match the content? Are link texts meaningful? Is information conveyed by colour alone (WCAG 1.4.1)?
How DokAudit reflects this
- Standards check with veraPDF: DokAudit checks the machine-testable PDF/UA rules with veraPDF, the PDF Association's open-source validator – for PDF/UA-1 and PDF/UA-2. Every finding states its reference to PDF/UA and to the WCAG criterion.
- Own contrast measurement: Colour contrast is not among the rules veraPDF checks. DokAudit therefore measures it itself and reports probably insufficient contrast as a warning – with the advice to re-measure text on coloured areas if in doubt.
- The honest remainder: Where no machine can decide, the report says so explicitly: a complete assessment additionally requires human review. AI suggestions for alt text are never adopted unchecked.
How to carry out the human checks yourself is explained in Test it yourself: the 5 quick checks.
Sources:
As of 09/2026. This article gives a general overview and is not legal advice. The statutory and standards texts in force are authoritative; state law may differ in individual cases.