Scanned and born-digital PDFs: inspect the actual file

Origin helps plan inspection—not predict cost

A Portable Document Format (PDF) file may contain page images, encoded text, or a mixture. How it was created is useful context, but does not establish accessibility, complexity, or remediation cost. Inspect representative pages and the underlying structure before estimating work.

Image-only scans and recognized text

A scan can begin as an image of a page. Without an accessible text representation, people who cannot read that image face a barrier. Optical character recognition (OCR) can add machine-readable text, but recognition errors need review. A scanner or later process may already have added OCR, so not every scanned PDF is still image-only. W3C PDF7: OCR for scanned documents.

Electronically created files still need review

A file exported from an authoring application can preserve text and structure, but may also contain outlined text, image pages, incorrect reading order, missing tags, or inaccessible forms. “Born digital” is not a guarantee of a usable text layer or correct semantics. Check the exported file, not only its source.

Choose the work from the findings

• If accessible text is missing, recover or recreate it and verify accuracy.
• Inspect headings, lists, tables, figures, language, links, and logical reading order.
• Check form controls and keyboard order where present.
• Evaluate the final file using automated checks and informed manual/assistive-technology review.

Section508.gov’s PDF guidance provides a practical starting point. The federal agency guidance is a technical resource here, not a statement that Section 508 directly governs every public entity.

Quick triage checks are not conformance tests

Try selecting and copying a visible sentence, searching for a visible word, and inspecting the tags and reading order. A successful search proves only that searchable text exists for that test. File size and appearance do not reliably identify origin, tagging, or accessibility.

Record actual findings and representative task time rather than using unsupported per-page price or time multipliers. Consider HTML publication or source repair where it serves the content and records requirements.