What the PDF Accessibility Checker does
This checker tests a PDF for the accessibility basics that software can decide from the file itself: whether it is tagged, whether the document language and a displayed title are set, whether figures have alternate text and tables have header cells, whether headings skip levels, whether it has bookmarks, whether links and form fields follow a keyboard tab order and have descriptions, and whether every page has real text rather than a scanned image.
Each result says what was found and names the WCAG success criterion, W3C PDF technique or PDF/UA clause it relates to. It then lists the checks a person still has to make - reading order, the quality of alternate text, colour contrast and meaningful link text - as a checklist you can tick and include in the report. The file is read in your browser and never uploaded.
How to use it
- Drop a PDF, or load one of the three examples: a tagged report, an untagged export of the same report, and a scanned document.
- Start with the failures. "Not tagged", "No document language" and scanned pages without text are the problems that most affect screen-reader users.
- Open "Tag structure" to see which tags the document uses and its heading sequence, which shows at a glance whether headings were tagged at all.
- Work through the manual checklist with the PDF open in a screen reader or your PDF editor's tags panel, tick what you have confirmed, and download the report as Markdown or JSON.
Reading the results
Tagged means the PDF has a structure tree and declares itself tagged (MarkInfo Marked true). Tags are how a PDF tells assistive technology which text is a heading, a list, a table cell or a figure, and in what order to read it. An untagged PDF can still be read aloud, but as one undifferentiated stream of text.
Language and title are single entries in the file, but they matter: without a language, a screen reader may pronounce the text with the wrong voice, and without a displayed title, the window shows a file name like draft3-final.pdf. A title that is itself a file name or about:blank is reported as a warning.
Pass results are evidence, not a certificate. A figure can have alternate text that says "image1"; tags can exist in the wrong order. The automated checks here are a subset of what PDF/UA (ISO 14289-1) and WCAG 2.2 require - the Matterhorn Protocol lists 136 failure conditions, and many need human judgement.
Scanned pages are detected when a page's content is mostly an image and it draws no text. Those pages need OCR before anything else, because nothing on them can be read, searched or tagged.
Worked example: the same report tagged and untagged
The tagged example is a two-page annual report printed from HTML with tagging on. It passes all seven automated checks that apply: 44 structure elements with MarkInfo set, language en-GB, the title "Riverside Library Annual Report 2025" shown in the title bar, one table whose 2 header cells are tagged TH, 5 headings in the sequence H1 - H2 - H2 - H2 - H2 with no skipped level, bookmarks, and real text on both pages. Its one decorative image was written as an artifact, so there are no Figure tags - reported as a note, because only a person can say whether an image is decorative.
The untagged example is the same report exported without tags or a language, so the result is 2 failures and 1 warning: not tagged, no document language, and a title of "about:blank" that describes nothing. The scanned example adds a third failure - page 2 is an image with no text layer - and a warning that its title is set but viewers are not told to display it.
Limitations: what the result does not prove
- This is not a full PDF/UA or WCAG conformance check. Reading order, alternate-text quality, colour contrast, meaningful link text, correct table header scope and correct list tagging all need a person, and are listed as manual checks.
- Encrypted PDFs have encrypted text strings, so their title, language and alternate text cannot be read here; the checker says so rather than reporting them as missing.
- Very large structure trees are counted up to 200,000 elements, and page text is examined on the first 400 pages.
- Tag roles are resolved through the RoleMap, but attribute-level rules (such as table header scope or list numbering attributes) are not validated.
Privacy: where your data goes
Everything you paste, type or drop is processed in this browser tab. It is not uploaded, logged, stored or sent to analytics. Session recording and tag-manager scripts are switched off on this page.
Standards and sources
- PDF/UA (ISO 14289) - PDF Association - checked 19 Sep 2026
- W3C WCAG 2.2 PDF Techniques - checked 19 Sep 2026
- ISO 32000-2 (PDF 2.0) - PDF Association resources - checked 19 Sep 2026
- W3C - Web Content Accessibility Guidelines (WCAG) 2.2
- ISO 14289-1 PDF/UA - Document management applications - Electronic document file format enhancement for accessibility
- PDF Association - Matterhorn Protocol (PDF/UA failure conditions)
- ISO 32000-2:2020 PDF 2.0, section 14.8 Tagged PDF
Frequently asked questions
How do I check if a PDF is accessible?
Start with automated checks like these for tags, language, title, alternate text, headings and a text layer, then test by hand: listen to it with a screen reader such as NVDA, check the reading order in your editor's tags panel, and check colour contrast. No tool alone can declare a PDF accessible.
What is a tagged PDF?
A tagged PDF contains a structure tree that labels content as headings, paragraphs, lists, tables, figures and so on, in reading order. Assistive technology uses those tags; without them it has to guess the structure from the page layout.
Is this the same as a PDF/UA validator?
No. PDF/UA (ISO 14289-1) has many more machine-checkable conditions, catalogued in the Matterhorn Protocol, plus conditions that need human judgement. This tool covers the checks that most often fail and explains each one; use a dedicated validator such as veraPDF or PAC for a conformance claim.
Why does it say my title is not displayed?
PDF viewers show the file name in the window title unless the file's ViewerPreferences set DisplayDocTitle to true. WCAG technique PDF18 asks for the document title to be shown, so a set but hidden title is a warning.
How do I fix an untagged PDF?
The best fix is to go back to the source document, use real headings, lists and table headers, add alternate text to images, and export or print to PDF with tagging (accessibility) switched on. Adding tags to a finished PDF by hand is possible in professional editors but slow.
Why are scanned pages a failure?
A scanned page is a picture of text. Screen readers cannot read it, search cannot find anything on it, and it cannot be tagged. Running OCR adds a text layer, after which the document can be tagged and checked again.
Is my PDF uploaded to check it?
No. The PDF is parsed by JavaScript in this page, in your browser. The file, its text and the report stay on your device unless you download or copy the report yourself.
Last reviewed by the A2Z.Tools team against the sources listed above.
