How to Redact a PDF (And Actually Remove the Text)

Redacting a PDF sounds like it should be simple: cover the sensitive bits, save, send. The problem is that the obvious way to do it — draw a black rectangle over the text — doesn't remove anything. The words stay in the file, and anyone who receives it can get them back by selecting the page and pasting it into a text editor.

This guide covers how to do it properly: the methods that genuinely remove content, the ones that only appear to, and — most importantly — how to check which kind you just used.

First, understand what you're removing

A PDF stores text as instructions, not as a picture: place these characters, in this font, at these coordinates. Everything you can select and copy exists as one of those instructions.

Drawing a filled rectangle adds a new instruction to the end of the list. It does not modify or delete the one underneath. Visually the text disappears; structurally nothing has changed. That's the entire trap.

Real redaction means deleting the text instruction from the file. On a scanned page — which contains no text, only an image — it means destroying the pixels and discarding the original image.

Method 1: A dedicated redaction tool (browser)

The fastest route for most documents.

1. Open the free tool and upload your PDF.
2. Type the names or exact phrases you want removed, or select a PII type — Social Security numbers, credit card numbers, emails, phone numbers.
3. Preview the matches, then redact and download.

What happens underneath: on pages with a text layer the matching text objects are removed from the content stream; on scanned pages the page is OCR-read to find where the text physically sits, and those pixels are painted out before the page is rebuilt. Metadata — author, title, creation software — is stripped too, since it often carries more identifying information than the visible page.

Good for: most documents, scanned or digital, when you know what to remove. Limits: handwriting isn't detected, and one page at a time on the free tier.

Method 2: Adobe Acrobat Pro (desktop)

Acrobat's Redact tool is genuine redaction — this is worth saying plainly, because plenty of comparison pages imply otherwise. Mark for Redaction followed by Apply really does remove the underlying content.

1. Tools → Redact.
2. Mark the text or regions to remove (or use Find Text to mark every instance).
3. Click Apply, then save.

The critical step is Apply. Marking alone changes nothing — the marks are just annotations until applied. Saving at the marked stage produces a file that looks redacted and isn't.

Good for: large documents, precise region control, offline work on sensitive material. Limits: around $240/year, a desktop install, and a workflow that's easy to half-complete.

Method 3: What not to do

These look like redaction and are not:

• Drawing a black box in Preview, Word, or any annotation tool. The text remains underneath, fully recoverable.
• Highlighting in black. Same problem — it's a visual layer.
• Printing to PDF after covering text. This sometimes flattens content, but it depends entirely on the software and the layer. Never rely on it.
• Converting to an image and back. This can work, but it destroys the text layer of the whole document, bloats the file, and still leaves the sensitive pixels visible unless you painted them out first.

If your process is "cover it, then save," assume the data is still there until you prove otherwise.

Method 4: Command line (qpdf, pdftotext)

There's no reliable one-liner that redacts a PDF, and you should be sceptical of any guide claiming otherwise. Command-line tools are excellent for checking a redaction, though — see below.

How to verify — the ten-second test

Whatever method you used, verify it. This is the part most guides skip, and it's the only part that actually tells you whether you're safe.

For digital PDFs:

1. Open the redacted file.
2. Select all (Ctrl/Cmd-A), copy, paste into a plain text editor.
3. Search the pasted text for what you removed.

If it appears, it was never removed. Alternatively, run pdftotext redacted.pdf - and read the output — it prints every text object in the file, black boxes and all.

For scanned pages:

1. Zoom in hard on the redacted areas. Faint characters showing through mean the pixels are intact.
2. Export or screenshot the page and read it.
3. Extract the embedded images if your tool allows it. What you get should already be blacked out — if you get the original image, the box was drawn on top.

Also check the metadata. Document properties often carry the author's name and the original filename, which can be as revealing as anything on the page.

The failure that catches careful people

Here's the one worth knowing about, because it defeats a sensible process.

Documents are often both things at once. A statement has selectable text and a scanned remittance slip. A contract has a typed body and a photographed ID attached. An invoice has text and a stamp.

A tool that asks "is this page text or a scan?" and picks one route will handle these pages as text — reasonably, since they have text — redact the text layer perfectly, and never examine the image. The output looks immaculate, the text-layer data really is gone, and the account number inside the embedded image is still sitting there in plain sight.

So when you verify, check the images too, not just the text. We cover this in more depth in how to redact a scanned document.

A short checklist

Before you send anything:

• Copy-paste test returns nothing sensitive
• Scanned regions are opaque under magnification
• Embedded images checked, not just the text layer
• Document metadata stripped
• You kept an unredacted original somewhere safe — redaction is permanent
• A human read the output, not just the tool's summary

That last one matters more than any tool choice. Automatic detection finds well-formed patterns and the terms you supply; it can't know that the third paragraph names a confidential supplier. For high-stakes documents — court filings, records requests, patient records — read the output before it leaves your hands.

The short version

Covering isn't removing. Use a tool that deletes content from the file, apply the redaction properly, and then spend ten seconds proving it worked. The test is faster than the apology.