How to Redact a Scanned PDF

To redact a scanned PDF, the tool must first run OCR to turn the picture of the page into searchable text, then delete both the recognised text and the image pixels under each redaction. The free PDF Redaction tool does this automatically in more than 100 languages: upload the scan, review the AI highlights, add boxes for handwriting or stamps, and download a file where the redacted regions are removed from the image itself.

This guide explains why scans behave differently, walks through the steps, and shows how to check the result.

Why scanned PDFs are different

A scanned document, a fax or a phone photo saved as PDF contains one image per page and no text. Without a text layer:

• Search finds nothing, so Acrobat's Find Text & Redact and similar tools return no results.
• Automatic detection has nothing to read, so names and numbers are not flagged.
• A black box does not remove anything: the pixels of the original page remain in the image under the annotation.

OCR fixes the first two problems by recognising the characters in the image. A proper redaction tool then removes the pixels, not just the recognised text, because the image is the real content.

Method 1: redact a scan with built-in OCR (recommended)

Steps

1. 1

   Upload the scanned PDF

   Drag the scan into the free PDF Redaction tool. Scans, faxes and photos exported as PDF are all accepted. The file is processed in your browser.

   
2. 2

   Let OCR read the pages

   Pages without a text layer are recognised automatically with OCR in more than 100 languages, including Latin, Cyrillic, Arabic, CJK and Devanagari scripts. If a scan already has a poor text layer, enable "Always run OCR" to re-read it.

3. 3

   Review the AI detections

   The AI highlights names, addresses, dates, ID numbers and other personal data found in the recognised text. Turn on face detection to find photos and signatures.

   
4. 4

   Fix what OCR missed

   Handwriting, stamps and low-quality regions may not be recognised. Use the rectangle tool to draw boxes over anything the AI did not catch, and the search box to redact repeated terms.

   
5. 5

   Download and verify

   On download, the image pixels under each box are removed and re-encoded, the recognised text is deleted, and metadata is cleared. Open the result in the redaction checker to confirm.

   

Method 2: Adobe Acrobat Pro

1. Open the scan and run Scan & OCR → Recognize Text for the whole document.
2. Open Tools → Redact and use Find Text & Redact or mark areas by hand.
3. Apply the redactions, confirm removal of hidden information, then run Sanitize Document.
4. Save as a new file.

Acrobat's OCR handles common languages and clean scans well; it has no automatic detection of names or addresses, so every item must be found by reading.

Method 3: desktop and offline

For scans that cannot leave the building, PDF Redaction Studio Desktop runs OCR, AI detection and face detection entirely on the machine, with no internet connection. It is available for Windows and Linux.

What does not work on scans

• macOS Preview: its Redact tool needs a text layer and Preview has no OCR, so it only redacts the visual area you drag over. On a scan that is acceptable for one page, but nothing is detected for you.
• Annotation rectangles in any viewer: the image underneath is untouched.
• Blurring or pixelating in an image editor: blur can be partially reversed and the original file still exists.

Checking a redacted scan

Because a scan has no original text, the usual "try to select the text" test proves less. Instead:

1. Zoom into a redacted region; the box should be solid with no ghosting.
2. Extract the page image (or open the file in the redaction checker) and confirm the region is flat colour.
3. Check the document properties for leftover metadata.

Quick summary

1. Upload the scan to PDF Redaction; OCR runs automatically.
2. Review the AI highlights; enable face detection for photos.
3. Draw boxes over handwriting, stamps and anything OCR missed.
4. Download: pixels and text are removed, metadata cleared.
5. Verify the result.