How to redact a PDF properly

Including the verification step most guides leave out.

Redacting a PDF has two halves, and most instructions only cover the first. Covering the text is easy. Making sure it is actually gone is the part that matters, because a black box drawn over text does not remove the text.

In the browser, with no upload

Open the tool and drop your PDF onto the page. It is read by your browser and not sent anywhere.
Click Auto-detect personal data to find emails, phone numbers, card and bank numbers, NHS and National Insurance numbers, passport numbers, postcodes and dates of birth. Boxes appear over each hit.
Open Redact people by name and paste the names in the document, one per line. Names have no pattern to match on, so no tool can find them for you — but once you list a name, every form of it is covered, including “Mr Alvarez”, “J. Alvarez” and ALL-CAPS.
Drag over anything else that needs covering, and click any box you want removed.
Review every page. Detection is an assistant, not a guarantee — anything without a matching pattern or a term you typed will be missed.
Click Download redacted PDF. Every page is flattened to an image on export, so no text layer survives.

In Adobe Acrobat Pro

Use Tools → Redact. This is the important part: the redaction tool is not the same as the drawing, markup, or highlight tools, and only the redaction tool removes content.
Mark the text you want removed.
Click Apply. Until you apply, nothing has been deleted — the marks are still just marks.
Use Sanitize document afterwards to strip metadata and any hidden content.
Save as a new file, so you keep an unredacted original.

Acrobat Pro does this correctly. The common failure is reaching for the rectangle tool out of habit, or marking redactions and never clicking Apply.

In macOS Preview

Recent versions of Preview include a redaction tool that removes content properly. Older versions offered only shape and annotation tools, which do not. If you are unsure which you have, treat the output as unverified and run the check below — this is exactly the situation where a file looks finished and is not.

The check that actually proves it

Run all three of these on the finished file, not the draft:

Copy and paste. Select across the blacked-out area, copy, and paste into a text editor. Nothing hidden should appear.
Search. Use the reader's find function to search for a name or number you removed. It should find nothing.
Check the properties. Open document properties and look at author, title and any file path. These routinely name a client, a matter, or an employee.

If all three come back clean, the file is safe to send. If any of them return the text you thought you removed, the redaction was decorative and the document should not leave your machine.

Common mistakes

Black highlighting instead of redaction. Looks identical; removes nothing.
White boxes over text. Same problem, and less obvious that anything was hidden at all.
Changing text colour to match the background. The characters are still there and still searchable.
Flattening but forgetting metadata. Clean pages, and the author field still names the person who made the file.
Redacting a scan without checking for a text layer. Many scanned PDFs have had OCR run on them at some point and carry an invisible, fully searchable text layer.
Trusting automatic detection on its own. Patterns find structured data. They do not find names, addresses in free text, or a detail that only identifies someone in context.

If your document is a scan

A scanned page has no text to select, which sounds safer and is not: the information is still legible to anyone who can read, and to OCR. You still need boxes over it, and you still need those boxes burned in. BLACKOUT runs OCR in your browser to find personal data on scanned pages, then flattens the result the same way.

OCR is less reliable than a real text layer, so scanned pages need closer review than typed ones — the tool says so on screen when it uses OCR, and it is worth taking seriously.