04
The file is open and someone needs it this afternoon. Here is the whole sequence, in order. It works in whatever PDF application you already have, because the order matters more than the software, and there is one step in it that almost nobody does.
01
Start with the original file
Track down the PDF as it was produced, rather than a copy someone printed and scanned back in. A native PDF still has its text layer, which means automatic detection can read it and the redaction tool can work out exactly which characters sit inside the region you mark. Once a document has been through a printer and a scanner it is a stack of pictures, and everything below gets slower and less certain.
If the file has been edited and re-saved a few times, use the newest copy, but expect it to carry history. More on that further down.
02
Scanned pages need a different method
Try to select a line of text with your cursor. If nothing highlights, the page is an image, and there is no text to delete. Redacting it means changing pixels: the tool has to burn the covering box into the image and rebuild the page from the modified image. Draw a box over the picture instead and the picture underneath is untouched.
Scans feel safe for the wrong reason. Copy-paste finds nothing, so the page looks clean. Run it through OCR and the names come back as text in about a minute.
03
Decide what has to go before you touch the tool
Two minutes with a pen saves a second pass. Write down who receives this document and what they already know, because that is what decides how much has to come out.
Then list the categories rather than hunting item by item. Names, initials, signatures. Email addresses and phone numbers. National ID, passport, case and employee numbers. IBANs and card numbers. Dates of birth. Salary figures. Home addresses. And the sentences that identify someone without naming them, like the contractor who walked out halfway through the March audit, who is exactly one person to everybody inside the organisation. If the document falls under GDPR, that list runs wider than most people expect.
04
Let detection do the first pass, then read the document yourself
Most tools worth using will scan the text and offer candidates: emails, phone numbers, ID numbers, card-like numbers, IBANs, dates, addresses, names. Take the offer. It is faster than forty pages by eye, and it tends to catch the account number sitting in a footnote on page 31.
Then read the document. Every rule set produces false positives and misses things, and the misses are the ones that hurt. Pattern matching finds a phone number reliably. It does not recognise that paragraph four describes one identifiable person in plain prose. Deciding on each candidate is the work, not a formality on top of it.
05
Use the redaction tool, not the rectangle tool
This is the step that ruins documents. The rectangle, the highlighter and the fill tool all draw a shape on top of the page. The text carries on existing underneath, and anyone can select it and copy it out. The mechanism is worth understanding once.
What you want is the feature called Redact, or Mark for Redaction, or something close. It normally sits apart from the drawing tools, and it will warn you that the change cannot be undone. That warning is how you know you have found the right one. Mark every region you decided on, and add manual boxes over anything detection did not offer you.
06
Export a new file
Redaction marks are not redactions until they are applied. In most tools there is a separate Apply step, and until you run it you are looking at an annotation.
Then export, or Save As under a new filename, rather than saving over the original. Saving in place can append the change to the existing file and keep the earlier version inside it, which is an unpleasant way to undo everything you just did. Keep the original somewhere you can find it, because you will not get it back out of the redacted copy.
07
Check what the page doesn’t show you
Open the document properties. Author, title, subject, keywords, original filename: any of those can name a person or a client matter you did not intend to send. Clear the ones that do.
Then look for comments, sticky notes, form fields and attached files. They live alongside the page rather than in it, so a box on the page does nothing to them. Cropped images deserve a thought too, since a crop normally hides part of the picture instead of deleting it.
08
Test the file you are about to send
Here is the step people skip.
Close the editor. Open the exported file in a plain reader. Select across an area you redacted, copy, and paste it into a text editor or an empty email. If your words appear, the document is not redacted, no matter what it looks like. Then search the file for a name or a number you took out. A hit behind a black box means the same thing.
Do this on the exported file, never on the one still open in front of you. The editor shows you what you meant to do. The export is what the other side receives.
hddn runs this sequence in the browser, so the document is not uploaded to a server to be processed. It flags candidates for you to approve or reject, takes manual boxes and text selections for everything else, and applies confirmed redactions as true redactions, so the covered content comes out of the exported PDF rather than sitting under a shape. Run the copy-paste test on the export anyway.
Nobody will ever see how carefully you marked the document up. They will notice the one line you left underneath it.