The complete PDF remediation guide

PDF remediation is the work of turning a visually fine but structurally broken document into one a screen reader can navigate. This guide walks through every layer of that work — tags, reading order, text alternatives, tables, forms and metadata — and is candid about which parts automation handles well and which still need a human.

What is PDF remediation?

PDF remediation is the process of adding the structure assistive technology needs to an existing PDF: a tag tree that mirrors the logical content, a correct reading order, text alternatives for images, marked-up tables and labeled form fields. The visible page does not change; the machine-readable layer underneath it does.

• What a tag tree is and why untagged PDFs read as noise
• Remediation vs. recreation: when fixing beats rebuilding from source
• The relationship between WCAG 2.1 AA and PDF/UA for remediated files

How do you assess a PDF before remediating it?

Triage each file for four things: is there real text or just a scan, how complex is the layout, does it contain tables or form fields, and is it still worth publishing at all. That assessment sets the remediation level, the cost, and whether OCR has to run first — many "PDF problems" are really scanning problems.

• Text layer check: born-digital vs. scanned
• Complexity tiers: linear text, multi-column, tables/forms, visual-heavy
• Deciding to remediate, recreate, archive or retire
• Estimating effort per tier

What are the core remediation steps?

Every remediation follows the same sequence: establish the tag tree, set the reading order, write text alternatives, mark up tables and lists, label any form fields, then set document metadata — title, language, bookmarks for long files. Done in that order, each step builds on the last instead of undoing it.

• Tagging headings, paragraphs, lists, links and figures
• Reading order and the difference between visual and logical order
• Alt text: informative vs. decorative images, artifacts
• Table structure: header cells, scope, complex table strategies
• Document title, language, bookmarks and initial view

Which parts can automation do reliably?

Automation is strong at the structural layer: detecting headings and lists, building tag trees, setting metadata, flagging every image that needs a description, and validating the result. It is weaker at judgment calls — whether an image is decorative, whether a caption already conveys the chart. Good pipelines automate the structure and route the judgment calls to people.

• What machine remediation handles end-to-end
• Judgment calls: alt text quality, complex tables, ambiguous order
• DocoMatic's model: automated levels plus human review where confidence drops

How do you verify a remediated PDF?

Verification needs two layers: machine validation against PDF/UA and WCAG checks, and functional testing with assistive technology on a sample. A file can pass every automated check and still read badly, so treat checker results as the floor. Keep the verification report with the file as evidence.

• Machine checks: what validators can and cannot prove
• Screen reader spot-checks and what to listen for
• Sampling strategies for large batches
• Evidence: reports, scores and change history

How do you keep new PDFs from needing remediation?

Fix the source. Most inaccessible PDFs are exported from templates that were never set up: styles not used for headings, tables drawn with tabs, images pasted without alt text. Accessible templates plus a check-before-publish step cost minutes and eliminate the majority of future remediation spend.

• Accessible authoring in Word, Google Docs and InDesign
• Export settings that preserve structure
• A publish gate: check every outgoing document automatically