How to Redact a Scanned PDF (OCR Required)

Three types of “scanned” PDF (identify yours first)

Not every scan behaves the same. Open the file and try Select All (Ctrl+A). If text highlights, you have a searchable layer—maybe from the scanner’s OCR or a prior export. If nothing highlights, you likely have pure image pages. If highlights appear but do not match what you see on screen, you have bad OCR offset—a dangerous middle case where auto-search redacts the wrong coordinates.

Why scanned PDF redaction fails

• Redacting only the OCR text layer while the scanned image still shows SSN digits to the human eye.
• Drawing black shapes in Preview or Word—overlay on image PDFs; underlying image data often remains extractable via image extraction tools.
• Low-DPI fax headers: OCR never sees account numbers in the header band; zero auto-matches feels like success.
• Phone photos embedded in PDFs: skew, glare, and dual layers (photo + auto-OCR) confuse detection.
• Assuming Ctrl+A silence means the file is safe—image PDFs leak visually and via pdfimages extraction even when paste is empty.

myFICO forum users recommend print → marker → rescan for personal statements. That works when the final file is a flat image with no hidden text re-added—but it is tedious for multi-page productions. Desktop tools that burn redaction into page images give repeatable results with verification.

Workflow A: OCR first, then pattern redaction

When pages are clean scans at 300 DPI or better, run OCR (correct language, searchable PDF output). Validate on three sample pages: click a word in the OCR layer and confirm it aligns with the visible scan. Then run SSN, account, phone, and email detection across the document.

1. Save original scan as scan-original.pdf (never email this copy).
2. Run OCR on all pages; export searchable PDF if tool requires separate step.
3. Spot-check OCR alignment on page 1, a middle page, and last page.
4. Run auto-detection for financial and identity patterns.
5. Manually box regions OCR missed (signatures, handwritten amounts, fax headers).
6. Apply redaction that removes both OCR text and image pixels in marked regions.
7. Verify: visual read + Find search + extract images from PDF to inspect thumbnails.

Workflow B: Area redaction without reliable OCR

Faxes, low-DPI scans, and degraded copies may never OCR cleanly. Switch to region-based redaction: draw boxes over visible sensitive bands (header, signature block, account table). Apply burn-in that replaces image pixels—not annotation overlays.

Print-marker-rescan remains valid: print, cover with opaque ink or tape, scan at 300 DPI, confirm no selectable text in output. For law firms and SMBs processing volume, offline burn-in redaction is faster and easier to audit than physical rescanning.

Scanned IDs, checks, and exhibit photos

ID scans and check images need region redaction on the photo itself—numbers appear in multiple zones (front number, barcode area, MICR line on checks). OCR may read digits incorrectly; manual boxes plus visual verification are mandatory. Phone photos saved as PDF should be treated as image-only unless you confirm a clean OCR layer.

Verification for scanned PDFs (paste test is not enough)

• Visual inspection of every page at 100–150% zoom on redacted regions.
• Find/Search for known strings if OCR layer exists post-redaction.
• Ctrl+A paste test when text layer remains (should return nothing sensitive).
• Export page images or use pdfimages to confirm blacked regions have no readable digits in thumbnails.
• Open in a second viewer (Edge, Chrome, Acrobat Reader) for consistency.

Step-by-step workflow

1. Determine PDF type: image-only, OCR scan, bad OCR, or mixed.
2. Save unredacted original locally; never edit the only copy.
3. For clean scans: run OCR; validate alignment on sample pages.
4. Run auto-detection for SSN, account, routing, phone, email patterns.
5. Manually box fax headers, check images, handwriting, and barcode zones.
6. Apply burn-in redaction (image pixels + OCR text removed in regions).
7. Verify visually on every page; run Find/paste tests where text layer exists.
8. Scrub metadata; export as scan-redacted-descriptive-name.pdf.
9. Send redacted copy via secure channel.

Common mistakes

• Redacting only the OCR layer

  Digits remain visible in the scan image. Humans and image extraction tools still see them.

• Trusting zero detection matches on a fax

  Low DPI and header fonts break OCR. Zero matches often means broken detection, not a clean file.

• Using Word highlight on a scanned PDF

  Forums document black-on-black text that still copies out. Highlight does not destroy image pixels.

• Skipping visual review because paste test was empty

  Image-only PDFs pass paste tests while still displaying account numbers on screen.

Verification before you share

• ✓Every page visually inspected at zoom on redacted zones.
• ✓Find search returns no SSN/account hits (if OCR layer present).
• ✓Ctrl+A paste test clean (if text selectable).
• ✓Sample page images extracted—no readable digits in redacted bands.
• ✓Fax header and footer bands manually confirmed.

FAQ