How to Make a PDF Accessible: Tags, Reading Order, PDF/UA

What a screen reader needs from a PDF

A PDF describes where to paint glyphs and images on a page. That is enough for the eye, but a screen reader needs more: which line is a heading, where a list starts, which cells are table headers, what an image shows, and in what order to read the columns. In a PDF that information lives in tags, a logical structure stored alongside the visible page.

The W3C's PDF technique on reading order puts it simply: the reading order is determined primarily by the tag order, and the technique applies to tagged PDF documents. A file without tags leaves the screen reader guessing from the drawing order, which for two-column layouts, sidebars and tables is often wrong.

Before fixing a PDF, ask whether it needs to be a PDF. Section508.gov notes that PDFs are often not the most accessible or mobile-friendly option and that federal policy asks agencies to prioritize HTML. A web page with real headings is accessible with far less effort; the WCAG basics guide covers what that takes.

What an accessible PDF needs

The W3C's Techniques for WCAG 2.1 include 23 PDF techniques. The main ones, grouped by what they fix:

Need	W3C technique	What it means
Tags and logical structure	PDF3, PDF9, PDF21	Headings, paragraphs and lists are tagged, in reading order
Text alternatives	PDF1	Every informative image has alt text
Decorative content	PDF4	Borders, background art and running headers marked as artifacts
Tables	PDF6	Table, row, header and data cell tags
Links	PDF11, PDF13	Link annotations with meaningful text
Real text, not images	PDF7	Scanned pages run through OCR and corrected
Document title	PDF18	Title entry set and shown in the title bar
Language	PDF16, PDF19	Default language in the catalog, changes marked
Navigation	PDF2, PDF17	Bookmarks for long files, page labels that match
Forms	PDF10, PDF12, PDF5	Labeled fields with name, role and value; required fields marked

Alt text in a PDF follows the same rules as on the web: describe what the image tells the reader, and mark purely decorative images as artifacts so they are skipped. The alt text guide has examples for charts, photos and logos.

Start from the source document

Accessibility is cheapest to build in where the content is written. In a word processor, use real heading styles instead of bold text, built-in list and table tools, alt text on every image, and a document title in the file properties. Then export with the option for a tagged PDF switched on. The export turns the styles into tags; a document built from manual formatting gives the export nothing to work with.

The Word to PDF tool on this site turns a .docx or .txt file into a PDF with selectable text in your browser. It does not write tags, so for a PDF that has to meet an accessibility requirement, export from a word processor with tagging on, or add tags afterward in a PDF editor that supports them.

Scanned PDFs

A scanned page is an image. The W3C technique on OCR states the problem: the content of the document is images, not searchable text, so a screen reader has nothing to read and users cannot select or resize the text. The fix is optical character recognition, which the technique describes as the fallback when the original source file is not available.

OCR is a first step, not the finish. Software flags words it could not recognize with confidence, and the technique says each one should be reviewed and corrected. After that the document still needs tags. The OCR PDF tool adds an invisible, searchable text layer to a scan in your browser; it does not add tags. The guide to fixing a scanned PDF covers straightening and cleaning the scan first, which also improves recognition.

Title and language

Two small settings are easy to miss and easy to fix. The W3C technique PDF18 asks for a document title in the Title entry of the document information dictionary, plus the DisplayDocTitle flag set to true, so viewers show the title in the title bar instead of a file name like scan_0042.pdf. Assistive technology can then announce it. The PDF metadata editor lets you see and set the Title entry in your browser; whether a viewer shows it instead of the file name depends on the DisplayDocTitle flag, which you set in a full PDF editor's initial view settings.

The default language goes in the Lang entry of the document catalog (PDF16), so a screen reader uses the right pronunciation. A passage in another language gets its own Lang entry (PDF19).

What PDF/UA adds

PDF/UA is ISO 14289. The Library of Congress describes PDF/UA-1 as a constrained form of PDF 1.7, approved and published in 2012 with a minor update in December 2014, and intended to be consistent with WCAG 2.0. Its requirements, in the Library's summary, include:

All real content is tagged; artifacts are marked as artifacts and kept out of the structure tree.
Text characters map to Unicode, so they can be read and searched.
Content is tagged in a logical reading order.
The natural languages used in the document are declared.
Metadata is stored as XMP, and a pdfuaid:part entry with the value 1 identifies a PDF/UA-1 file.

A later part, PDF/UA-2, is ISO 14289-2:2024. WCAG says what the reader must be able to do; PDF/UA says how a PDF file has to be built to make that possible. PDF/UA is not the same as PDF/A, the archiving standard; the guide on PDF vs PDF/A explains the archive side.

How to check a PDF

The W3C techniques each end with a test. The quickest ones need nothing beyond a screen reader or a read-aloud feature:

Read it aloud. Use a screen reader or read-aloud tool and listen for the order. Columns, sidebars and footnotes are where it usually breaks.
Tab through it. Press Tab to move through links and form fields. The focus order must follow the logical order of the document.
Look at the title bar. It should show the document title, not the file name.
Try to select text. If a page selects as one image, it needs OCR.
Inspect the tags. A PDF editor with a tags panel shows whether headings, lists and tables are tagged as such.

Section508.gov publishes a PDF testing checklist and a testing and remediation guide for documents that must meet the Section 508 Standards, and an automated checker can flag a missing tag or title. It cannot judge whether alt text makes sense or whether the reading order matches the meaning, so a human read-through stays part of the job.

Frequently asked questions

What makes a PDF accessible?

Mainly tags: a hidden structure that tells assistive technology what each piece of content is and in what order to read it. On top of that it needs real text instead of images of text, alt text for images, a document title, a declared language, and tagged tables, lists and form fields. The W3C publishes a technique for each of these.

Is a PDF with selectable text accessible?

Not by itself. Selectable text means a screen reader can find the words, but without tags it has to guess headings, lists, tables and reading order. The W3C reading order technique applies to tagged PDFs, because the tag order is what sets the reading order.

What is PDF/UA?

PDF/UA is ISO 14289, the international standard for accessible PDF. The Library of Congress describes PDF/UA-1 as a constrained form of PDF 1.7, published in 2012 with a minor update in December 2014, intended to be consistent with WCAG 2.0. PDF/UA-2 is ISO 14289-2:2024.

How do I make a scanned PDF accessible?

A scanned page is a picture, so there is no text to read. Run optical character recognition to get real text, review the words the software was unsure about, and then add or check tags. If you still have the original document, exporting it again as a tagged PDF is usually less work.

Should a federal agency publish PDFs at all?

Section508.gov says PDFs are still used across government but are often not the most accessible or mobile-friendly option, and that federal policy asks agencies to prioritize HTML and use PDFs only when necessary. When a PDF is needed, it has to conform to the Section 508 Standards.

Sources

Written by Tobias. Updated October 30, 2026.