PDF/UA: the standard for accessible PDFs
PDF/UA (ISO 14289) describes how a PDF must be built so that screen readers and other assistive technologies can use it reliably. Here you can read what the standard requires and what can be checked automatically.
General overview. This page summarises the basics in simplified form. It is a draft and is currently under expert review. The text of the standard itself is authoritative.
What PDF/UA is
PDF/UA stands for “PDF/Universal Accessibility”. The international standard ISO 14289 builds on the PDF specification and defines the properties a PDF must have so that it can be used accessibly.
It sets requirements for documents, but also for software that displays PDFs and for assistive technologies. For organisations that publish PDFs, the part about the document itself matters most.
At a glance
- PDF/UA-1
- ISO 14289-1, first published in 2012, based on PDF 1.7
- PDF/UA-2
- ISO 14289-2, published in 2024, based on PDF 2.0
- Identification
- A PDF/UA document declares the standard in its metadata.
- Checking aid
- Matterhorn Protocol of the PDF Association (for PDF/UA-1)
Why structure matters
A PDF can look perfect to sighted readers and still reveal nothing about what is a heading, a table or a footnote. That meaning is carried by an invisible structure layer: the tags. PDF/UA requires all relevant content to be tagged.
-
Screen readers
Content is read aloud in a meaningful order; headings, lists, tables and links are announced as such.
-
Navigation
Headings and bookmarks let readers jump straight to a section instead of reading page by page.
-
Reflow and zoom
Tagged content can be reflowed in suitable viewers, for example at high magnification or on small screens.
Typical requirements at a glance
The following selection shows what usually matters in practice. The full rules are in the standard itself.
-
Tags for structure
Headings, paragraphs, lists, tables, links and figures are tagged appropriately.
-
Logical reading order
The structure reflects the order in which the content makes sense, including columns and boxes.
-
Alternative text
Figures that convey information have alternative text.
-
Tables
Header cells are marked as such and associated with their data cells.
-
Language and title
The document language is set, language changes are marked, and the document title is displayed.
-
Artifacts
Purely decorative elements such as lines, backgrounds or repeated headers and footers are marked as artifacts and are not read aloud.
-
Bookmarks
In longer documents, bookmarks that follow the outline help with orientation.
-
Forms
Form fields are tagged, clearly labelled and reachable in the reading order.
-
Fonts and Unicode
Fonts are embedded and characters map to Unicode values, so text can be read aloud, searched and copied correctly.
What machines can check, and what they cannot
Many requirements can be checked automatically. The PDF Association’s Matterhorn Protocol breaks PDF/UA-1 down into individual failure conditions and distinguishes those that are machine-checkable from those that need human judgement. Validators such as the open-source tool veraPDF check the machine-checkable part.
Machine-checkable (examples)
- Is content tagged or marked as an artifact?
- Is the document language set?
- Does every figure have alternative text?
- Are fonts embedded and characters mapped to Unicode?
- Is a document title present and displayed?
- Is the PDF/UA identifier in the metadata?
Human review needed (examples)
- Does the alternative text describe the image accurately?
- Does the reading order make sense?
- Are headings chosen and nested correctly?
- Is content marked as an artifact really without meaning?
- Are complex tables tagged in an understandable way?
Passing validation shows that the form is right, not necessarily the meaning. That is why EqualDoc validates its results with veraPDF and lists in its report the points that need human judgement.
What this means for organisations
- Start at the source: templates built cleanly with headings, lists and tables usually export to much better PDFs from word processors or layout software.
- Remediate the backlog: for existing PDFs whose source files are missing or that were scanned, remediation is often the practical route.
- Check the results: automatically first, then targeted human review, especially for tables, forms and figures.
- Set priorities: frequently used documents and those needed for key processes first.
PDF/UA and the law
The BFSG and BITV 2.0 do not refer to PDF/UA directly. But PDF/UA describes how the requirements for documents can be implemented technically in a PDF, which makes it a widely used benchmark.
Frequently asked questions
Is a PDF/UA-conformant document automatically accessible?
Not automatically. PDF/UA provides the technical basis. Whether a document is genuinely usable also depends on content and language, such as meaningful alternative text and clearly structured tables.
PDF/UA-1 or PDF/UA-2: which version is right?
PDF/UA-1 is the most widely used so far and is supported by most tools. PDF/UA-2 builds on PDF 2.0 and extends the tagging options; software support is growing. The right choice depends on your tools and what your recipients need.
What is the Matterhorn Protocol?
A freely available document from the PDF Association that turns the PDF/UA-1 requirements into testable failure conditions. It also shows which checks cannot be automated.
Can a scanned document meet PDF/UA?
Yes, if the text is recognised (OCR) and the document is then structured. A plain image of a page contains no text a screen reader could read.
How well structured are your PDFs?
The free PDF check shows what is missing in your document. EqualDoc adds tags, reading order, alternative text and metadata, and checks the result against PDF/UA.