PDF Accessibility Checker: Common Mistakes and Fixes
Reviewed by the OnlineFree.app team · Updated
Key points
- The PDF Accessibility Checker returns a PASS/FAIL verdict, a 0–100 score and a fixable issue list for PDF/UA, Section 508 and WCAG 2.1 self-checks.
- A PASS at the basic level says nothing about tags, alt text or reading order; choose the PDF/UA level for those.
- The checker parses structure only: it runs no OCR, rejects encrypted PDFs and caps uploads at 15 MB.
- Fix accessibility problems in the source document and re-export before retro-tagging a finished PDF.
- Treat every verdict as a self-check and confirm important results with a screen reader and a full validator.
What the PDF Accessibility Checker actually reports
The PDF Accessibility Checker is a free online tool that reads an uploaded PDF and returns a PASS or FAIL verdict, a 0–100 score, passed/failed/warning counts, and an itemised issue list with a one-line fix for each — a self-check mapped to PDF/UA, Section 508 and WCAG 2.1 AA.
You upload one unencrypted PDF of 15 MB or less and pick a level. Basic checks language, title, bookmarks and document information. PDF/UA adds tag structure, image alternative text, tables and reading order. Strict adds the Section 508 / WCAG 2.1 AA extensions: form fields, embedded fonts and contrast metadata. The level only decides which rules run; it never changes your file.
Results are ordered error → warning → info, so the red items are the ones that will sink a delivery. Each entry names a category, the detected value or missing state, the WCAG success criterion where one applies (for example 3.1.1 for language), and a page number when the check can locate one.
Mistake 1: exporting a PDF that was never tagged
The most common cause of a FAIL is not a PDF problem at all — it is an export problem. If "tagged PDF" was switched off when you saved from Word, Google Docs or InDesign, the file arrives without a structure tree, and no amount of editing inside Acrobat restores semantics that were never written.
In Word, use File > Save As > PDF, then Options and tick "Document structure tags for accessibility". In InDesign, enable "Create Tagged PDF" in the PDF export dialog. If the PDF already exists, Acrobat's Tools > Accessibility > Autotag Document can add a structure tree as a starting point.
Autotagging guesses. We have watched it merge a two-column newsletter into one interleaved run of text, so re-run the PDF Accessibility Checker after autotagging and read the reading-order warnings, not just the verdict.
Mistake 2: missing language, title and bookmarks
A missing default language — the /Lang entry — is a WCAG 3.1.1 failure and one of the cheapest things to fix. Without it, a screen reader falls back to its own default voice and may mispronounce the entire document, including names and place names.
A missing or generic document title is the close second. The title announced by assistive technology comes from the document metadata, not the filename, so "report_final_v3.pdf" stays invisible to users even though it looks fine on screen. Set Title in the document properties. At the basic level, these two alone will fail many deliverables before anyone examines the tags.
Bookmarks matter on long files. PDF/UA validation guidance treats an outline as expected once a document runs past roughly 21 pages, and the checker reports a missing or empty bookmark tree. Generate bookmarks from headings in Acrobat, or build them from styles in the source before export. The underlying success criteria are published in WCAG 2.1, which is backwards-compatible with the WCAG 2.0 Level AA that the Revised Section 508 Standards reference.
Mistake 3: alt text, tables and reading order
Images without alternative text are the single most frequent PDF/UA error we see. Every meaningful image needs an /Alt value that describes its purpose, not its appearance, and purely decorative rules or flourishes should be marked as artifacts so they are skipped entirely. Empty alt text and missing alt text are different states, and the checker distinguishes them.
Tables create two failures. The first is a data table whose header cells are tagged as ordinary cells instead of TH with a scope, so header-to-cell association is lost. The second is a table used only for layout — a two-column newsletter or a signature block — which should be built from tags or text boxes instead.
Reading order is the quiet one. Tag order must match visual order, and a document can look perfectly correct while a screen reader reads the sidebar before the headline. The checker reports structural warnings here, but the confirmation step is a short manual pass with a screen reader after you fix the tags in Acrobat's Reading Order panel or, preferably, in the source file.
How do you fix a PDF that fails before the check starts?
The PDF Accessibility Checker refuses a file before parsing it when the upload is encrypted, damaged, not a PDF, or larger than 15 MB. That is a guard rail, not a bug — a structure parser cannot read a document it cannot open.
For password-protected or permission-restricted files, open the PDF with the password and save a copy with security removed, provided you have the right to do so, then run the check again. Rights-managed or DRM-wrapped files often cannot be decrypted at all; in that case ask the publisher for an accessible copy rather than fighting the file.
For files over 15 MB, split the document by chapter or section and check each part, or export a lighter version with downsampled images. Scanned pages are the other dead end: the tool does not run OCR, so a PDF that is only pictures of text returns nothing useful, and an empty result is not a pass. Run OCR first — Acrobat's Recognize Text or your scanner's built-in OCR — then re-run the checker.
What a PASS verdict does and does not prove
A PASS means no rule in the level you selected failed, as of the checker's 2026 rule set. It is a self-check, not a certificate, and it is not a legal guarantee of compliance with PDF/UA, Section 508 or WCAG.
Several things sit outside what any structural checker can see: whether alt text is actually meaningful, whether captions match the audio, whether the language is plain enough, how live form fields behave, and how the reading order feels in a real screen reader session. Contrast is assessed from metadata, not from rendered pixels, so a low-contrast design can still slip through.
Treat the output as a to-do list. Fix the source, re-export, re-check, then do a manual pass with NVDA, JAWS or VoiceOver, and if you deliver against PDF/UA, verify with a full validator as well. The U.S. Section 508 programme publishes the official standards text for reference. If you also need to hand off dates or schedules alongside a report, the free tools on OnlineFree.app cover that without an install or an account.
Frequently asked questions
Is the PDF Accessibility Checker free, and do I need an account?
Yes — the PDF Accessibility Checker is free and requires no account or signup. You upload one unencrypted PDF of up to 15 MB, choose a check level, and read the verdict and issue list in the browser. Because parsing happens server-side, avoid uploading documents containing confidential personal data you are not permitted to share.
Does a PASS from the PDF Accessibility Checker mean my PDF is Section 508 compliant?
No. A PASS means no rule failed at the level you selected during that run. Section 508 compliance also requires manual verification, and the Revised 508 Standards incorporate WCAG 2.0 Level AA. Confirm results with a screen reader and, for formal PDF/UA work, a full validator before claiming compliance.
Why was my PDF rejected before the check even started?
Files are rejected before parsing when they are encrypted or password-protected, corrupt, not a PDF, or larger than 15 MB. Remove security from a permitted copy, split a large document into parts, or re-export a smaller version. The tool never partially parses a rejected file.
Which check level should I choose: basic, PDF/UA or strict?
Basic covers language, title, bookmarks and document information. PDF/UA adds tag structure, image alternative text, tables and reading order. Strict extends to Section 508 and WCAG 2.1 AA items such as form fields, embedded fonts and contrast metadata. Pick the level that matches the deliverable you must hand over.
Can the PDF Accessibility Checker fix the problems it finds?
No. It reports issues and gives a one-line fix pointing at Acrobat, Word or your export settings; you make the change yourself. Fixing in the source document and re-exporting is faster than retro-tagging a finished PDF, especially for reading order, table headers and long bookmark trees.