How to Redact a PDF Properly (Black Boxes Are Not Enough)

Why do black boxes fail to redact a PDF?

Because a box drawn with a markup tool sits on top of the text and does not replace it. The words are still stored in the page, so anyone can select, copy or search them, or delete the box in an editor.

A PDF page is a list of drawing instructions: put this text here, this image there. When you draw a black rectangle, add a black highlight, or place a shape in Word before exporting, you add one more instruction at the end. Nothing earlier in the list changes. Wikipedia's article on sanitization describes several public failures of exactly this kind, including a 2005 US military report published as a PDF with opaque blocks over the sensitive parts.

What does true redaction do?

It removes the content inside the marked area from the page itself, including characters, parts of images and vector graphics, and then draws an opaque box where they were. The reader can see something was removed, and nothing is left to recover.

Redact PDF on this site does this by rebuilding every page that has a box as a flattened image at 200 DPI, with the boxes painted into the pixels. Whatever was under a box no longer exists in the file. The output is a new PDF built from scratch, so bookmarks, attachments and form fields from the original are not carried over, and document metadata is removed by default. It runs in the browser, so the unredacted document is never uploaded, which is the point of redacting in the first place.

The trade-off: text on redacted pages can no longer be selected or searched, and image pages are often larger than text pages. Pages without boxes are left as they are. If you need the rest of a redacted page to be searchable again, run the output through OCR PDF afterwards. OCR only sees the black box, so it cannot bring the removed words back.

• Open Redact PDF and choose the file.
• Draw boxes over names, account numbers, signatures or whole paragraphs, or use Find text to redact to mark every occurrence of a word or number.
• Review every page and the list of boxes, leave Also remove document metadata ticked, then apply redaction and download the new file.

Redacting on each platform

Built-in tools differ a lot here, and most of them only draw on top.

How do you check that a PDF is really redacted?

Open the output file and try to get the hidden text back: search for it, select across the box and paste, and open the file in a second viewer. If nothing appears in any of the three tests, the content is gone.

• Search: press Ctrl+F or Cmd+F and search for the redacted name or number. It must not be found.
• Select and copy: press Ctrl+A or Cmd+A on the page, copy, and paste into a text editor. Nothing from under the box should appear.
• Different viewer: open the file in a second app, such as Chrome and Preview. Redactions that rely on annotations may display differently or not at all.
• Zoom in on scanned pages: the underlying image must be black, not just covered.
• For a final check, run the file through PDF to Text and read what comes out.

Places sensitive information hides

The visible page is only part of a PDF. Before sharing a redacted file, check:

• Metadata: title, author, subject and keywords may contain names. Remove PDF Metadata shows and clears them for files you have not run through the redaction tool.
• Bookmarks and file names: a bookmark titled “Settlement with J. Smith” gives the game away, and so does a file called smith-settlement-redacted.pdf.
• Comments and form fields: reviewer notes and filled field values are separate from page text. Flatten or delete them first.
• Repeated information: a name redacted on page 1 might appear in a header on every page or in a table on page 14. Use the search to find every occurrence, and search for variants such as initials and misspellings.
• Earlier versions: editors that save incrementally can keep previous versions inside the file. A redaction tool should write a new file.
• Context: a redacted name next to an unredacted job title, date of birth or address may still identify the person.

Scanned documents need special care

In a scan, the text is part of a picture. Redaction must blacken the pixels themselves. If the scan was OCR processed, there is also an invisible text layer containing the recognised words; that must be removed too. Rebuilding the page as a single image handles both, and the search test confirms it.

The Find text feature only works where there is a text layer. On a plain scan there is nothing to search, so either draw the boxes by hand or run OCR first and then search. OCR misreads some words, so after a search-based redaction of a scan, still read every page for occurrences the search missed.

Troubleshooting

Only redact what you are entitled to

Redaction is for protecting your own or your organisation's information, or information you are legally required to protect, such as personal data before disclosure. Keep the original unredacted file in a secure location, because redaction is irreversible and you may need to produce the full document later. Name the two files clearly so the wrong one is never attached to an email.