Redacting a PDF means permanently removing information from the file — not hiding it under a black rectangle. Anyone who receives a badly redacted PDF can often select the "hidden" text, copy it, and paste it out in seconds.
This guide is a working process: pick a tool that actually removes content, run the redaction, clean what still leaks, and verify the output. The slow part of the job is finding every occurrence, including on scans. Automated detection is built for that. Desktop editors still work if you are marking boxes you already know about.
This is practical guidance, not legal advice. If the document is leaving the organisation under a legal duty — a subject access request, a freedom of information release, a court filing — the judgement about what to remove is a separate job. See DSAR redaction and FOI redaction.
What "redacted" actually has to mean
A PDF is a set of drawing instructions, not a picture of a page. Text, images, annotations and document properties all live as separate objects. Drawing a filled rectangle on top of a name adds one more instruction. It does not delete the name.
If you can still select the text under the mark, or copy the page and paste the original wording into a text editor, the file is not redacted. That failure mode is common enough to have caused real disclosure incidents; the mechanics are in why drawing a black box isn't redaction.
Proper redaction removes the underlying objects (or destroys them by rasterising the page) and then sanitises metadata.
A process that works on any tool
Do the finding and the applying as two passes. Mixing them is how names survive on page forty.
- List what has to come out before you open the file. Names and variants, email addresses, phone numbers, account numbers, reference numbers, dates of birth. Working from a list is what makes consistency possible.
- Search the whole document for each item on the list, including headers, footers and signature blocks. Apply a real redaction mark to every hit — not a highlight, not a shape.
- Read for indirect identification. "The finance director" identifies someone in an organisation that has one. A job title plus a location often does the same. This pass catches things search will not.
- Apply / commit the redactions. Until you do this step, most tools have only marked the content. The underlying text is still in the file. This is the step people skip when they are in a hurry.
- Sanitise metadata and hidden content. Author, title, original filename, comments, attachments. See what still leaks after you redact a PDF.
- Verify the output. Open the finished file. Select the text under a redaction. Copy the whole page into a plain text editor. Search the result for every item on your list. If anything you meant to remove reappears, start again from the marked file — not from the "finished" one.
Finding every occurrence, including scans
On one page, finding what to remove is obvious. Across a few hundred pages it becomes the whole job: every occurrence, every variant spelling, every phone number in a signature, every reference in a footer. Miss one and a name blacked out on page four and left visible on page forty tells the reader exactly what was hidden.
That mechanical sweep is what automated detection is for. It will not decide whether a particular third party should come out, and it should not. It changes the shape of the work from "read 400 pages and hope" to "review what was found."
eleyed does that sweep on native PDFs and on scans. It reads the words on the page, takes what it finds off the page itself, and discards the file when the work is done. You still do steps 3 and 6 — the judgement and the check. You do not have to hunt.
A scanned page is an image. There is no text object to search unless the tool reads the pixels. eleyed does: built-in OCR, no separate conversion step, including photographed pages and faxes saved as PDF. A desktop editor that only searches visible text will miss a scan unless you OCR it first.
Covering a scan with a rectangle in a PDF editor can work if the rectangle is burned into the image and there is no hidden text layer underneath. Many "searchable scans" have both: an image you see, and an invisible text layer from an earlier OCR pass. Redacting only the image leaves the text layer extractable. Always do the copy-and-paste test. If words appear that you cannot see, there is a text layer you have not removed.
If you are marking boxes by hand
Some people already live in a desktop PDF editor and have a short, known list. That path still has to remove content rather than cover it.
Adobe Acrobat Pro has a Redact tool that removes content rather than covering it. Acrobat Reader can display a redacted file but cannot produce one.
In current Acrobat Pro the flow is:
- Open the PDF and find Redact in the tools (sometimes under Edit or Protect).
- Mark each piece of content, or search and mark every occurrence of a term.
- Apply redactions. Until you apply, the marks are only instructions. The text is still there.
- Run Sanitise / Remove hidden information so author names, previous versions and attachments do not travel with the file.
- Save as a new file and run the verification steps above.
Acrobat Pro does not find personal data for you, does not handle plain text, and does not hook into a pipeline. If you already pay for it and you are marking a handful of known boxes, it will do that job.
Apple Preview, since macOS Sonoma, has a real Redact tool in the markup toolbar. It removes content. That is a different control from drawing a rectangle in the same toolbar — the rectangle is still just an annotation. Use Redact, not the shape. Then check document properties, rename the file, and run the copy-and-paste test. Preview only handles PDFs, only on a Mac, and only one document at a time.
Other paid PDF editors (Foxit, Nitro and similar) need the same two checks: an Apply / commit step that deletes content, and a way to remove hidden information. "Draw a black box and flatten" is not that feature.
If the file must never leave the device, a local editor is the architecture for that constraint. For anything hosted, the questions worth asking first are in what to ask before putting a document into a redaction tool.
Free online "PDF editors" are the high-risk end of this. Many store the upload, and many only add an annotation. If you use one, assume the file has been retained, and still run the copy-and-paste test on what comes back.
After the file looks right
The page is the part everyone checks. These are the parts people miss:
- Document properties — author, title, software, sometimes the original path.
- The filename —
Smith-disciplinary-final.pdftells a story before anyone opens it. - Comments, attachments, form fields, previous revisions.
That list is expanded in what still leaks after you redact a PDF.