PDF/UA

Short definition

PDF/UA (Universal Accessibility) is the standard for accessible PDF documents, codified as ISO 14289. It ensures that a PDF is readable and navigable by assistive technologies — above all screen readers. PDF/UA defines not only what appears in a PDF, but also how it must be structured so that blind and visually impaired users can understand it.

Current versions:

• PDF/UA-1 (ISO 14289-1, 2014) — based on PDF 1.7, widely deployed
• PDF/UA-2 (ISO 14289-2, 2024) — built on PDF 2.0, extended structure elements

Core requirements

PDF/UA requires that a document:

1. Is a Tagged PDF — every content element has a semantic role (Heading, Paragraph, List, Table, Figure…)
2. Defines a logical reading order — independent of the visual arrangement
3. Provides Unicode mapping for every text glyph — so screen readers can actually read the text
4. Contains alternative texts for images, graphics and formulas
5. Sets language tags (/Lang) — for correct pronunciation
6. Does not rely solely on visual cues (colour-only differentiation forbidden)
7. Uses real table tags for tables, not just visual alignment

Tagged PDF — the heart of it

A Tagged PDF carries a structure tree parallel to the visual content. Every element in the tree has a role name from a fixed list:

/Document
 ├─ /H1 "Invoice 2026-0042"
 ├─ /P "Dear Sir or Madam …"
 ├─ /Table
 │   ├─ /TR (header row)
 │   │   ├─ /TH "Line"
 │   │   ├─ /TH "Quantity"
 │   │   └─ /TH "Price"
 │   └─ /TR
 │       ├─ /TD "Consulting"
 │       ├─ /TD "8 h"
 │       └─ /TD "€960.00"
 └─ /Figure (alt: "Company logo")

Assistive technologies read the structure tree instead of the visual page — reading order, heading hierarchy and table relationships become unambiguous.

PDF/UA, PDF/A and WCAG

Three frequently-confused standards:

	Purpose	Requires
PDF/A	long-term archiving	visual fidelity, self-contained
PDF/UA	accessibility	Tagged PDF, Unicode, alt texts
WCAG	web accessibility (general)	projected onto PDF via WCAG techniques

They are not mutually exclusive: a document can be PDF/A-2a (or PDF/A-3a) and PDF/UA-compliant at once. The -a conformance level of PDF/A already requires Tagged PDF — the combination is seamless.

In Germany, the Behindertengleichstellungsgesetz (BGG) combined with BITV 2.0 references EN 301 549, which effectively mandates PDF/UA conformance for publicly accessible PDFs.

Creating PDF/UA with the Dokmatiq API

When content areas in a DocGen request are semantically tagged, the API automatically produces a Tagged PDF — and, on request, a PDF/UA-1-compliant document:

curl -X POST https://api.dokmatiq.com/v1/docgen/render \
  -H "Authorization: Bearer $DOKMATIQ_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "outputProfile": "PDF/UA-1",
    "language": "en-GB",
    "stationery": { "firstPage": "…" },
    "contentAreas": [
      { "role": "Heading1", "x": 20, "y": 40, "html": "<h1>Invoice 2026-0042</h1>" },
      { "role": "Paragraph", "x": 20, "y": 60, "html": "<p>Dear Sir or Madam …</p>" },
      { "role": "Table", "x": 20, "y": 100, "html": "<table>…</table>" },
      { "role": "Figure", "x": 150, "y": 30, "image": "logo.png", "altText": "Company logo" }
    ]
  }'

The API maps each role to the appropriate PDF structure tag, sets /Lang to en-GB and embeds Unicode mappings for all fonts used.

Validation

Three established validation routes exist for PDF/UA:

• PAC 2024 / PAC 3 — free validator by the Access for All foundation
• veraPDF — open source, supports PDF/UA-1 in addition to PDF/A checks
• Matterhorn Protocol — test catalogue with 31 failure conditions and 136 techniques

The Dokmatiq API uses veraPDF internally and rejects documents that violate the target profile before returning them.

Common pitfalls

1. “Tagged PDF” from Word often isn’t enough — Word produces tags, but tables and forms frequently land without correct structure
2. Alt text missing or generic — “image001.jpg” is not an alternative text; it must convey the image’s meaning
3. Colour as the only meaning carrier — “red amounts = overdue” without a symbol or text violates PDF/UA
4. Missing language tag — without /Lang, the screen reader pronounces every text in its default language