Unredaction Techniques: How Failed Redactions Get Exposed

TL;DR

The problem: Most document redactions are superficial: black boxes over text that's still extractable.
The techniques: Copy-paste, metadata extraction, OCR layer analysis, layer toggling, image adjustment.
When it's legitimate: Journalists, researchers, and the public have found crucial information through failed redactions in court documents, FOIA releases, and government files.
The ethics: Faulty redactions are disclosure failures by the redacting party. Exposing improper redactions serves accountability, but consider victim privacy.

The black boxes on government documents aren't always as opaque as they look. From the Manafort plea deal to TikTok's internal research to the 2025 Epstein files disaster, redaction failures have exposed information that was meant to stay hidden. Here's how it works, and how to check any document for yourself.

Why Redactions Fail

Most digital redaction failures come from a fundamental misunderstanding: covering text is not the same as deleting it.

When someone draws a black rectangle over sensitive text in Microsoft Word, Adobe Reader, or most PDF editors, they're adding a visual layer on top of the document. The underlying text remains in the file's data structure. [1]

It's like putting a sticky note over text on a printed page: the text is still there underneath.

Common Redaction Mistakes

Highlight Tools

Using black highlighting instead of actual redaction. The text is fully preserved underneath.

Drawing Rectangles

Using shape tools to draw black boxes. Creates a layer over text without removing it.

Font Color Changes

Changing text to white or to match the background. Text remains selectable and searchable.

Cropping Images

Cropping sensitive areas from scanned documents. Original content often preserved in file structure.

The Techniques

1. The Copy-Paste Test

The simplest technique. It works on a shocking percentage of "redacted" documents.

How it works:

Open the PDF in any reader (Adobe, Preview, Chrome, etc.)
Click and drag to select the "redacted" black area
Copy (Ctrl+C / Cmd+C)
Paste into a plain text editor (Notepad, TextEdit)
If text appears, the redaction is fake

Why it works: Highlighting a black box selects the text underneath. Copy captures that text. Pasting into a plain text editor strips out formatting and reveals the content. [2]

Real example: In January 2019, journalists discovered Paul Manafort's lawyers had filed court documents with easily copy-paste-able "redactions." The exposed text revealed Manafort had shared campaign polling data with a Russian intelligence-connected associate: exactly what investigators wanted to know. [3]

2. Select All and Paste

A variant that catches more content:

Open the document
Ctrl+A / Cmd+A (Select All)
Copy the entire document
Paste into a text editor
All text, including "hidden" content, should appear

This catches redactions scattered throughout long documents that you might miss with targeted selection.

3. PDF Text Extraction

Command-line tools extract all text from PDFs, ignoring visual layers:

Using pdftotext (poppler-utils):

pdftotext document.pdf output.txt

Using Python (PyPDF2):

import PyPDF2
with open('document.pdf', 'rb') as f:
    reader = PyPDF2.PdfReader(f)
    for page in reader.pages:
        print(page.extract_text())

These tools extract all text objects from the PDF structure, regardless of visual overlays. If the text wasn't actually deleted, it will appear. [4]

4. Metadata Inspection

PDFs contain metadata that often reveals more than intended:

Author name: Who created or edited the document
Creation/modification dates: Timeline of edits
Revision history: Previous versions embedded in file
Comments: Internal notes that weren't removed
Bookmarks: Navigation elements containing unredacted headings

Example: In 2021, the European Commission published an AstraZeneca vaccine contract with pricing "redacted." But they forgot about PDF bookmarks, which contained the original unredacted section headings, revealing the secret pricing. [5]

Tools for metadata extraction:

exiftool (command line)
pdfinfo (poppler-utils)
Adobe Acrobat's Document Properties
mat2 (Metadata Anonymization Toolkit)

5. Layer Analysis

PDFs can contain multiple layers that can be toggled on/off or contain hidden content:

Using Adobe Acrobat:

View → Show/Hide → Navigation Panes → Layers
Toggle layer visibility
Hidden or occluded content may appear

Using QPDF (command line):

qpdf --qdf --object-streams=disable document.pdf unpacked.pdf
# Now inspect unpacked.pdf structure

6. OCR Layer Extraction

Scanned documents often have invisible OCR (Optical Character Recognition) text layers for searchability. The visible image might be redacted while the searchable text layer remains intact.

Testing:

Try to search the document (Ctrl+F)
Search for words that should be in redacted sections
If they're found, the OCR layer wasn't redacted

Real example: The 2025 JFK assassination records release contained documents where the visible page was redacted but the invisible OCR layer retained full Social Security numbers and other PII. [6]

7. Image Adjustment for Translucent Redactions

Some redactions use semi-transparent overlays rather than opaque black. Image processing can reveal the text underneath.

Simple method (phone or basic image editor):

Screenshot the redacted area
Open in any image editor
Adjust exposure, brightness, contrast
Increase shadows, decrease highlights
Text may become visible

Advanced method (GIMP/Photoshop):

Open the image
Duplicate the layer
Apply "Difference" blend mode
Adjust levels and curves aggressively
Use color channel isolation

Real example: The 2025 Epstein files had redactions that could be defeated with basic phone image editing: adjusting exposure revealed text under translucent black boxes. [7]

8. Document Structure Analysis

PDFs are structured files. Tools can reveal the internal structure:

Using pdf-parser (Python):

pdf-parser.py document.pdf

This reveals all objects in the PDF, including streams that may contain unredacted content.

Using peepdf:

peepdf -i document.pdf

Interactive analysis tool that can extract embedded objects, streams, and reveals document structure.

Legal and Ethical Considerations

When Unredaction Is Appropriate

The legal landscape varies, but generally:

Public documents: If a document is publicly filed (court cases, FOIA releases), examining it for redaction failures is standard journalism and research practice.
Faulty redaction = disclosure: Courts have held that improperly redacted information in public documents was effectively disclosed. The failure is on the redacting party.
Accountability purposes: Journalists have legitimate interest in government and corporate transparency.

When Caution Is Warranted

Victim information: Just because you can extract victim names, addresses, or other sensitive personal information doesn't mean you should publish it.
National security: Information that could endanger lives or ongoing operations raises different considerations than corporate embarrassment.
Private documents: Unredacting documents you obtained without authorization is different from analyzing public court filings.

Precedent Cases

Manafort (2019): Multiple news organizations published information from failed redactions in public court filings. No legal consequences for the journalists: the responsibility was on the lawyers who botched the redaction. [3]
TikTok (2024): NPR published internal TikTok research exposed through faulty redactions in a Kentucky lawsuit. TikTok complained about the leak being "highly irresponsible," but didn't dispute the content. [8]
Epstein files (2025): Public analysis of redaction failures in DOJ releases became widespread, with no prosecutions for those who extracted "hidden" content from public documents.

How to Verify Your Analysis

If you find potentially unredacted content:

Cross-reference: Verify the information matches other known facts
Consider context: Does the extracted text make sense in the document's context?
Check for artifacts: OCR errors, formatting issues, or garbled text might indicate false positives
Document your methodology: If you'll publish findings, document exactly how you extracted the information
Get a second opinion: Have others verify your extraction before publishing

Protecting Yourself When Redacting

If you need to redact documents yourself, here's how to avoid being the next cautionary tale:

Use Actual Redaction Tools

Adobe Acrobat Pro DC: Tools → Redact → Mark for Redaction → Apply Redactions
PDF-XChange Editor: Protect → Redaction
Redactable: Cloud-based, legal-grade redaction

The key word is "Apply": marking isn't redacting. You must apply redactions to permanently remove content.

Sanitize Metadata

After redacting, remove document metadata:

Adobe: File → Properties → Remove All
Command line: exiftool -all= document.pdf
mat2: mat2 document.pdf

Test Your Work

Before publishing:

Try to copy-paste from redacted areas
Search for redacted terms
Extract all text with pdftotext
Check metadata with exiftool
Have someone else verify

The Nuclear Option

For maximum security:

Replace sensitive text with "[REDACTED]" in original document
Copy all text to plain text editor (strips formatting and hidden data)
Create new document from plain text
Export to PDF

This destroys all hidden layers, metadata, and embedded content.

The Bottom Line

Black Boxes Are Often Transparent

The vast majority of redacted documents can be checked in seconds with copy-paste. More sophisticated techniques can defeat nearly any improper redaction.

This matters because:

Government transparency depends on public accountability
Court documents shape legal precedent
Corporate secrets hidden behind faulty redactions deserve scrutiny
Victims deserve proper protection, not theatrical black boxes that fail

If you're researching public documents, check the redactions. If you're redacting documents, do it properly. The same techniques work both ways.

Unredaction Techniques: When Black Boxes Aren't Really Black