What Is Redaction? A Plain-English Guide

Redaction is the permanent removal of specific information from a document before it is shared, while leaving the rest of the document readable and useful.

That last clause is what makes it its own discipline. Deleting a whole file is easy. Publishing everything is easy. Redaction is the awkward middle: release the contract but not the pricing, the police report but not the witness's address, the medical record but not the patient's identity.

Where the term comes from

Redaction predates computers by centuries — from the Latin redigere, to bring back or reduce. Editors "redacted" manuscripts, meaning they revised them for publication.

The modern security sense comes from paper: a clerk with a marker pen blacking out lines, then photocopying the marked page and releasing the copy. That second step is the whole trick. The copy has no ink underneath, because the copier only saw black. The original stayed in the file.

Digital redaction went wrong precisely by dropping that step. Drawing a black rectangle in a PDF is the marker pen. There is no photocopier — you send the original with a shape drawn on it, and the words are still underneath.

Redaction vs. the words people confuse it with

Masking hides data while keeping its shape — showing a card as •••• •••• •••• 4242. The full value usually still exists somewhere; masking is a display choice.

Anonymisation removes the ability to identify someone from a dataset, typically across many records. It's statistical, and notoriously hard: seemingly harmless fields can re-identify people in combination.

Pseudonymisation replaces identifiers with a consistent token, so records can still be linked without naming anyone. Reversible if you hold the key — which is why GDPR treats pseudonymised data as still personal data.

Deletion destroys the record entirely.

Redaction removes specific content from a specific document, permanently, and releases the rest. It is closest to deletion, but surgical.

Why it's legally required

Most redaction happens because someone must release a document and must not release part of it.

Freedom of information. FOIA and its equivalents require agencies to release records, with narrow exemptions for personal data, law enforcement and similar. The agency must publish the record and remove the exempt parts — and published records are archived and mirrored, so a failed redaction can't be recalled. This is the government use case in a sentence.

Court filings and discovery. Rules require personal identifiers to be removed from public filings. Failures here are well documented and expensive — the document is on the public docket, and reporters read dockets. See redaction for legal teams.

Health privacy. HIPAA's de-identification standard names specific identifiers that must be removed before health data can be shared without authorisation. Removing them from records is redaction. (Healthcare workflows.)

GDPR and data minimisation. If you're sharing a document containing personal data for a purpose that doesn't need it, removing it is the straightforward answer.

What "permanent" has to mean

A redaction is only meaningful if the removed content cannot be recovered by the person holding the file. Concretely:

• It must not be selectable or copyable.
• It must not appear in text extraction (pdftotext and similar).
• It must not be recoverable by deleting the annotation layer.
• It must not survive inside an embedded image on the page.
• It must not be sitting in the document's metadata.

Any of those is enough to make a redaction cosmetic. The last two are the ones careful people still miss.

Why digital redaction fails so often

Three structural reasons, none of which are about carelessness.

The interface lies. The screen shows a black bar. Every visual signal says done. Nothing indicates that you added a shape rather than removed a word.

Failures are silent. A bad redaction produces no error. It produces a professional-looking document. You find out when someone else finds out.

Scans are a different problem entirely. A scanned page has no text to delete — only pixels. Tools that search-and-remove text find nothing to do, and may report success. Handling scans requires OCR to locate the text and then destruction of the actual pixels; we go through that in how to redact a scanned document.

How to check any redacted document

The test takes ten seconds and works regardless of who produced the file:

1. Open the PDF, select all, copy.
2. Paste into a plain text editor.
3. Search for what should be gone.

If it's there, the document was never redacted — it was decorated. For scanned pages, zoom into the black areas and check nothing is legible, and inspect any embedded images.

The one-line version

Redaction means the information is gone from the file you hand over, not hidden on it. Everything else — which tool, which format, which workflow — is detail in service of that. If you want to see it work, try it on a document and then try to copy the text back out.