Repair PDF accessibility in Acrobat

Save a working copy and establish whether the source file exists. Inspect the tag tree, language, title, tables and image descriptions. An untagged file needs different work from a mostly well-structured file with a few local problems.

Adobe provides tools for checking, tagging and manual repair. Some checks explicitly need manual assessment. Marking one as passed without examining it is not an assessment.

Sources: [1]

Start with three questions. Is the text actual text or a scanned image? Is there a tag tree? Does its structure broadly match the document? The answers distinguish OCR work, structural reconstruction and local repairs.

If the source exists and contains systematic faults, re-exporting is often more sustainable. Twenty incorrect headings may stem from one style. Editing them individually in the PDF can fix this edition while leaving the next export broken. Where only the PDF remains, document repairs so they can be checked and, when necessary, repeated.

Acrobat’s current interface groups these tools under Prepare for accessibility. Check for accessibility runs the configured checks. Its report starts an investigation: an object may need manual assessment even without an automatic failure. Menu names vary by language and version.

• Identify meaningful content and decoration.
• Repair overall structure and reading order.
• Repair headings, lists and table relationships.
• Review descriptions, links, language and title.
• Validate, perform manual checks and try real reading tasks.

Suppose “Results” has become a P, an ordinary paragraph. Locate the tag and verify that it identifies the heading’s text. Change its role to the level required by the outline. Then inspect its tree position and try finding the section using heading navigation. Only the first of these steps is a name change.

For a table, changing P to TD is insufficient. You need coherent table structure, correct rows and cells, and header relationships. If automatic tagging split each number into separate objects, substantial reconstruction may be necessary. Begin with a small representative part and check that the method preserves structure and content.

Do not move large tag groups without knowing what they contain. Highlight associated page content and follow each tag’s references. After major changes, check for missing text, duplication or content placed beneath the wrong heading. Save intelligibly named intermediate versions so regressions can be traced.

Automatic tagging assigns P to headings and table cells alike. Do not merely change a few names to reduce the error count. Rebuild the outline and cell relationships first. Check that tags still point to the correct page content after moving them.

Moving tags and repairing structure can affect more than the selected object. Compare visible text, figures and links afterwards. If the file is digitally signed, also address the effect on that signature; do not assume the original signature remains valid.

Finish with another machine check and the reading tasks affected by your edits. If you moved a table between sections, inspect the surrounding transitions as well as the table. The repair is complete only when the intended information works in context.