Glossary · Glossary · Data protection
Document redaction: definition and how to get it right
Anonimatum Team
In short
Redaction is the permanent removal of selected content (a name, an account number, a signature) from a document before it is shared, so that it cannot be recovered from the text, the images or the hidden data of the file.
Redaction is routine work for anyone handling subject access requests, Freedom of Information responses, court bundles or published reports. The goal is always the same: the recipient gets the document, but not the information they are not entitled to see.
In a digital file, though, what you see is not all there is. A black rectangle drawn over a name hides it from view while leaving the name in the file. That is why it matters to separate cosmetic masking from true redaction, which deletes the content from every layer of the document.
True redaction vs. drawing black boxes
A PDF or Word file contains a text layer, images, annotations, stacked objects and document properties. Masking text with a black shape, black highlighting or a font coloured like the background only adds something on top: the original text is still underneath, and copying it into Notepad reveals it.
The US National Security Agency made the point in its 2005 paper *Redacting with Confidence*: information hidden or covered in a computer document can almost always be recovered, so sensitive content must be deleted, not covered. The UK Information Commissioner's Office (ICO) gives the same warning in its guidance on disclosing documents to the public securely, which lists ineffective redaction among the common causes of accidental breaches.
| Technique | What stays in the file | Safe? |
|---|---|---|
| Black rectangle drawn on top | All of the original text under the shape | No |
| Black highlight or font colour | The text itself, only recoloured | No |
| Print, black out with a marker, rescan | An image; marker ink may show through | Partly, and slow |
| True redaction | Content deleted from text and image layers; area painted over | Yes, provided metadata is checked too |
Every layer you need to clean
- Text layer: anything you can select, search or copy. This is where most failed redactions are exposed.
- Image layer: in a scanned PDF the data lives in the page pixels, so the area itself must be painted, not covered by an object.
- OCR layer: many scans carry invisible OCR text so they can be searched. Mask the image but leave that layer, and Ctrl+F will still find the name.
- Metadata and hidden content: author, title, comments, tracked changes, hidden spreadsheet rows, embedded thumbnails. See our entry on document metadata.
- Repetitions: a name removed on page 1 that reappears in a header, a footnote or the file name.
Redaction, anonymisation and pseudonymisation
Redaction is a technique applied to a specific document. Anonymisation is an outcome: nobody can identify the person by means reasonably likely to be used (recital 26 of the GDPR, mirrored in the UK GDPR). You can redact an employee's name and still not have anonymised the document if it keeps their job title, site and the date they left.
Pseudonymisation replaces identifiers with a consistent code or alias (“Person 1”) that can be reversed with separately held information. Pseudonymised data remains personal data. It is useful when readers need to follow who said what without knowing who they are.
When redaction is required
In the UK, redaction is part of answering subject access requests under the UK GDPR and the Data Protection Act 2018 without disclosing third-party data, and of responding to requests under the Freedom of Information Act 2000, where exempt information is removed and the rest released. In the US, agencies redact exempt material under FOIA before releasing records, and HIPAA-covered entities must strip identifiers before sharing health information outside permitted purposes.
How Anonimatum handles it
Anonimatum detects personal data with its own AI running on servers in the European Union, plus patterns for identifiers such as emails, phone numbers or IBANs. You review everything in a preview, then the redaction is applied irreversibly in the same format: native or scanned PDF (with OCR), Word, Excel, PowerPoint or images. Areas without text, such as handwritten signatures or stamps, can be covered with zones and shapes. For a quick job, start with our online PDF redaction tool.
Frequently asked questions
Is converting a Word file with black boxes to PDF enough?
No. The conversion copies both the box and the text beneath it. Delete the text in the source file or use a true redaction tool on the PDF.
Is a redacted document anonymised?
Not necessarily. It is only anonymous if what remains cannot identify the person by reasonable means. Check indirect details such as roles, dates and places.
How do I check a redaction worked?
Open the output, search for the redacted term, select and copy the text around the redacted area, and inspect the file properties. If anything comes back, the redaction failed.
What about scanned documents?
Paint over the image and remove any hidden OCR text layer as well. OCR-based tools also help find every occurrence across pages.
Redact documents you can safely release
Compare the plans and start redacting PDFs, Office files and images irreversibly, with a human review step before anything is applied.