PDF accessibility: validation, repair priorities, and human review
A PDF inventory is not a PDF accessibility test. A useful document result identifies what was validated, what failed, what could not be tested, and which human checks remain.
Published by ADAReporting · Sources checked September 26, 2026
Separate four different outcomes
A discovered document may pass automated checks, fail specific checks, be unavailable for testing, or still need manual review. Manual review remains necessary even after an automated pass. Do not turn a failed download or an encrypted file into a claimed accessibility failure.
Several links may point to the same PDF, so the number of document links can differ from the number of unique files. When comparing reports, check which documents were tested, when, and against which validation standard.
Interpret failed rules before prioritizing repairs
PDF/UA identification metadata is a conformance requirement, but missing identification alone does not demonstrate that a reader cannot use the content. Structural, font, annotation, and content-alternative failures require their own interpretation. A rule count is not a measure of the number of unusable pages.
Begin with the actual evidence and source file. Check tags and hierarchy, links, table semantics, figure alternatives, document language, and font behavior as relevant to the finding. Avoid claiming that every failure is a WCAG violation without mapping and validating the specific barrier.
Repair the source, then test the exported document
Use real headings and lists, identify table headers, provide meaningful image alternatives, and create descriptive links in the source. Export a tagged PDF and inspect the result; export settings alone do not establish correct structure or reading order.
For scanned pages, OCR can supply text but does not by itself establish headings, tables, alternatives, reading order, or usable form fields. Decide whether rebuilding the document or providing a suitable accessible format is more reliable than extensive repair of the scan.
- Prioritize current applications, instructions, and documents needed to access a service.
- Keep the editable source and identify the responsible publisher.
- Run validation after changes and retain the profile and failed-rule evidence.
- Manually test reading sequence, meaningful descriptions, tables, links, and form use.
Handle inaccessible or unavailable sources honestly
A document that could not be downloaded or processed has not been shown to pass or fail accessibility checks. It may need a broken link repaired, a later retry, or a separate review because it exceeded the tool’s limits. Keep those categories distinct from confirmed validation failures.
Do not assume that a public document is exempt solely because it predates the rule. DOJ exceptions have conditions, including context about current use. Obtain an appropriate decision for the specific document rather than deleting it from the audit inventory.
Acceptance includes a human reader
Check that a reader can find the document title, navigate headings and links, understand tables and alternatives, and complete any form. Use an appropriate PDF reader and assistive technology.
Automated conformance evidence supports that review, but does not replace it. Report the automated result and human-review status separately so a recipient does not mistake a validator pass for a complete accessibility certification.
Primary sources
Related guidance
Turn findings into a repair plan
Start with a small automated preview or review the scope and pricing for a full public-site report. Automated testing is not legal advice or certification of ADA compliance. Manual testing remains necessary.
