Clean PDF to text for screen readers (WCAG 2.2 compliant)

Tagging, reading order, alt text, header cells, and a two-tool validation pass that proves the file is actually accessible.

How to convert a PDF to clean text for screen readers (WCAG compliant)

By ScoutMyTool Editorial Team · Last updated: 2026-05-27

Introduction

I have sat next to a blind colleague trying to read a 40-page report in JAWS and watched the reader announce sidebar text in the middle of every paragraph because the tag tree was wrong. Visually the document was fine. To a screen reader it was unusable. The fix is not stylistic — it is structural. You add tags, you set the reading order, you label images, you mark decorative content as artifact, and you run two independent validators before publication. None of that is automatic. Here is the working pipeline I use to take a visually-fine PDF and turn it into one that NVDA, JAWS, and VoiceOver can navigate cleanly.

Vocabulary, quickly

TermMeaning
Tagged PDFPDF with a logical structure tree that screen readers can traverse
Reading orderSequence in which an assistive technology announces page content
Artifact tagMarks decorative content so the screen reader skips it
Alt textImage description spoken by the screen reader
WCAG 2.2 AAWeb content accessibility guidelines — the practical compliance bar for PDFs
PDF/UA-1ISO accessibility profile for PDFs; stricter than WCAG
Header cellsTable cells tagged TH so screen readers announce row and column context

Step by step

  1. Start from a structured source whenever possible. A Word document with proper heading styles tags itself on export. A flattened scan does not. Avoid scan-then-OCR-then-tag if you can re-export from the source.
  2. OCR scanned material first. If the source is genuinely paper, run a high-quality OCR pass and verify the recognized text reads correctly before tagging — bad text under correct tags reads as gibberish.
  3. Auto-tag, then audit. Use the editor’s auto-tag function as a starting point. Open the Tags panel and verify each heading, list, table, and figure is tagged correctly. Auto-tag is usually 60–80 percent right; the audit closes the gap.
  4. Set reading order explicitly. Use the Reading Order tool, drag tags into the order a sighted reader would follow, and confirm with the Read Out Loud preview before publishing.
  5. Add alt text to every figure. Decorative images get empty alt and an artifact tag. Informative images get a 1–3 sentence description. Charts that the body text already describes get an artifact tag.
  6. Tag tables with TH and scope. Header rows and columns need TH tags with scope attributes; data cells need TD. Without scope, screen readers cannot announce "column header X" when the user lands on a cell.
  7. Validate twice. Run Acrobat Pro Full Check for WCAG issues; run PAC for PDF/UA conformance. Fix everything that comes back as a failure, not just warnings — the warnings often hide real problems.
  8. Smoke-test with a real screen reader. NVDA on Windows is free and matches the most common user setup. Listen to the first three pages. If anything sounds wrong, fix the tag, do not change the layout.

Accessibility checklist before publishing

  • Document language is set in the PDF properties — screen readers use this to pick the correct voice and pronunciation rules.
  • Every heading uses the correct level (H1 once per page, H2/H3 nested in order). Skipping levels confuses navigation shortcuts.
  • Lists are tagged as L with LI children, not as visually-indented paragraphs. Otherwise the reader announces every item as a new paragraph.
  • Form fields have labels in the tag tree, not just visible captions next to the field on the page.
  • Decorative borders, page numbers, and running headers are artifact-tagged so they do not interrupt content reading.
  • Color contrast for text passes WCAG 2.2 AA (4.5:1 for body, 3:1 for large text). Hex-darker is cheaper than re-export.

Related reading and tools

FAQ

Why does my PDF read in a strange order?
Untagged PDFs have no logical structure — the screen reader walks page objects in the order they were drawn, which is rarely the order they should be read. Tagging adds a structure tree (headings, paragraphs, lists, tables) and the reader follows the tree instead of the draw order. If your PDF reads as nonsense, it has no tags or the tags are wrong; visually fixing the layout does nothing.
Does an OCR pass make a scanned PDF accessible?
Partially. OCR adds a text layer, which is necessary but not sufficient. The text layer has no structure — no headings, no list semantics, no table headers. Screen reader users can hear the words but cannot navigate. After OCR you still need a tagging pass that classifies the recognized text into structural roles.
What is the right alt-text length for a diagram?
Long enough to convey the information the diagram provides — usually 1–3 sentences for a chart, 1 sentence for a decorative graphic. If the surrounding text already describes the diagram, mark it as an artifact so the reader does not hear the description twice. Empty alt and missing alt are different: empty means "decorative", missing means "this is an unlabeled image" which screen readers announce as a problem.
How do I handle a multi-column page?
Tag the columns in reading order, not draw order. Most PDF accessibility tools have a Reading Order panel that visualizes the tag sequence; drag the column tags into the order a sighted reader would follow. Verify with NVDA or VoiceOver — the column order should match the visual flow, not jump from top of one column to top of the next.
Tables that span pages — how do I keep header context?
Mark the header row with TH and set the scope attribute, then mark it as a header that repeats across page breaks. Most editors have a checkbox for "repeat header row on page break"; turning it on at layout time also produces the right tagging if the underlying file was generated from a structured source.
Which validator should I trust?
Run two. Adobe Acrobat Pro accessibility checker for the common WCAG issues, and PAC (PDF Accessibility Checker) for PDF/UA. They catch different categories of problem. Pass both before publishing — Acrobat will miss UA-specific structural rules, and PAC will miss color-contrast checks.

Citations

  1. W3C — “Web Content Accessibility Guidelines (WCAG) 2.2.” w3.org/TR/WCAG22
  2. Wikipedia — “Tagged PDF — structure and accessibility.” en.wikipedia.org/wiki/Tagged_PDF
  3. Wikipedia — “Screen reader — PDF navigation patterns.” en.wikipedia.org/wiki/Screen_reader

Audit and tag PDFs in your browser

Run the accessibility checks locally — ScoutMyTool flags missing tags, missing alt text, and reading-order issues without uploading the file anywhere.

Open the PDF toolkit →