The PDF/UA standard, in plain language.

PDF/UA — ISO 14289-1, "Universal Accessibility" — is the international standard for accessible PDF. Where WCAG describes what accessible content must achieve, PDF/UA describes how a PDF file must be built so assistive technology can actually read it.

Descriptive transcript →

What a "tagged PDF" actually is

A PDF is, at heart, drawing instructions: put these glyphs at these coordinates. A screen reader can't infer from coordinates what's a heading, what's a table cell, or what order paragraphs should be read in. Tags add that missing semantic layer — a structure tree that says "this is an H1", "this is a list with four items", "read this column first", "skip this decorative line". PDF/UA defines what a complete, correct structure tree looks like:

  • every piece of content is either tagged or explicitly marked as an artifact,
  • headings are properly nested, tables have real row/column semantics,
  • images carry alternative text, the document declares its language and title,
  • reading order matches the logical order, fonts are embedded so text survives extraction.

How PDF/UA, WCAG and EN 301 549 fit together

  • WCAG 2.1 AA is the requirement regulators name — in the EU (via EN 301 549) and the US (ADA Title II, Section 508) alike.
  • EN 301 549 is the EU's harmonized standard; its document chapters point to WCAG.
  • PDF/UA-1 is how a PDF meets those expectations structurally. A PDF/UA-conformant file satisfies the technical foundation of WCAG's document requirements — which is why it's the target a11yfy produces.

What can be machine-checked — and what can't

The Matterhorn Protocol (by the PDF Association) breaks PDF/UA conformance into ~136 failure conditions. Roughly two-thirds are machine-checkable: missing tags, broken heading nesting, absent alt text, missing language or title, untagged annotations. Tools like veraPDF — the open-source validator developed with the PDF Association — test exactly these, and a11yfy validates every output with it.

The rest require human judgment: is the alt text meaningful? Is the reading order logical for this layout? No tool can honestly certify those automatically — which is why our reports flag judgment-call items with their location instead of pretending they're covered.

PDF/UA-1 vs. PDF/UA-2

PDF/UA-2 (ISO 14289-2, 2024) is the successor standard built on PDF 2.0. Regulatory practice today is anchored on PDF/UA-1, which is what a11yfy produces — with structures chosen to be forward-compatible with PDF/UA-2 where the standards overlap.

Want to see what your PDF's structure tree looks like against all this? Upload it — the analysis is free.