Security

How to redact a PDF permanently, not just black it out

Use a tool that deletes what is under each box instead of drawing over it: in Redact PDF you mark text, areas or search results, apply, and the marked content is removed from the file before a box is drawn where it was.

7 min read

Most redactions that go wrong do not go wrong because someone missed a word. They go wrong because the word was never removed at all. A black rectangle was drawn over it, the page looked finished, and the text underneath stayed in the file, ready to be copied out by anyone who tried.

So the one idea worth taking from this page: redacting is deleting. The black box is only there to show where something used to be.

Why a black box is not a redaction

A PDF page is a list of drawing instructions: put these characters here, this picture there, a line from here to there. When you draw a black rectangle with a shape tool, a highlighter set to black, or a box in a word processor before exporting, you add one more instruction to the end of that list. Everything before it is still there.

That is why "redacted" text so often turns out to be readable. Select across the box and copy, and the words paste out. Press Ctrl+F and the search finds them. A screen reader reads them aloud. A text-extraction script lifts them without ever noticing a box was drawn.

Real redaction takes the characters, the image pixels and the drawn lines under the mark out of the page's instructions altogether. Afterwards there is nothing under the box to find.

How to redact a PDF with KovaPDF, step by step

  1. Open [Redact PDF](/redact-pdf) and choose your file. Each page appears on screen, ready to be marked.
  2. Mark what has to go, in any of the three ways described below. You can mix them freely.
  3. Review the list. Every mark appears in a list beside the pages. Search results are only suggestions until you tick them and press Mark. You can move or resize a box, remove any mark, and undo or redo with Ctrl+Z and Ctrl+Y. Nothing in the file changes yet.
  4. Choose how the boxes look. Black, white or grey, with an optional label printed in each box, such as REDACTED. The colour and label only change the appearance. The content under the box is removed whichever you pick.
  5. Decide about hidden information (see below), then press Apply redactions and download the new file. Its name ends in -redacted.pdf.

The three ways to mark something:

  • Select text (press T) and drag across words, the way you would select them in any reader.
  • Draw a box (press A) over anything at all: a photo, a signature, a stamp, a chart, part of a scanned page.
  • Search for a name, number or phrase across every page, or find every email address, phone number, card and account number, IBAN, date, ID number in the 000-00-0000 form, or web address at once.

What is actually removed

Under every mark, the tool removes:

  • Text, including the invisible text layer that a scanned-and-OCR'd PDF carries. The characters around a removed word keep their exact positions, so the rest of the line does not shift.
  • Image pixels. They are overwritten in the image data itself. An image lying completely under a box is dropped.
  • Drawn shapes and lines. They are cut at the edge of the mark, so the part underneath no longer exists.
  • Links, comments and form fields whose area overlaps the mark. Form field values are cleared.

Two things are always discarded, whatever you mark: earlier versions of the document saved inside the file, and page thumbnails. Both are small pictures or copies of the pages as they were.

Hidden information that is not on the page

A page can be spotless while the file still gives the game away. The title in the document properties might be "Disciplinary notes. J. Smith". A bookmark might repeat a name you blacked out. Under Also remove hidden information you choose which of these go:

OptionWhat it removes
Document propertiesTitle, author, subject, keywords and the XMP metadata other apps store
Content outside the visible pageText and images beyond the page edge, where a viewer cannot show them
Scripts and actionsJavaScript and actions that run when the file is opened or a link is clicked
Comments and markupNotes, highlights, stamps and drawings added on top of the pages
Attached filesFiles embedded inside the PDF
BookmarksThe outline in a viewer's sidebar, which can repeat names and titles
Hidden layersLayers that are switched off but still in the file

The first three are ticked to begin with. The rest are your call, because a bookmark list or an attachment is sometimes exactly what the recipient needs.

The check before you download

When you apply, the finished file is read again by a different PDF reader from the one that wrote it, and two things are checked. First, that no text is left under any mark. Second, whether any word you searched for still appears anywhere else: on a page you did not mark, in the document properties, in bookmarks, comments, form fields or the names of attached files. If it does, the result screen says where, and for properties, bookmarks, comments and attachments it offers to remove them and apply again.

Very occasionally a page is built in a way that cannot be taken apart with certainty, for example text in an unusual font encoding. Rather than hand you a page where a marked word might survive, the tool turns that one page into an image with the boxes burnt in, and tells you which page it was. That page is no longer searchable. The rest of the document keeps its real text.

Redacting a scanned PDF

Drawing a box works on a scan exactly as it does anywhere else: the pixels under it are overwritten in the scanned image.

Search is different. A plain scan is a picture, and search cannot read a picture. When you open a scan, the tool notices which pages have no text and offers to recognise it. After that, search finds words on those pages too.

Password-protected and restricted PDFs

A PDF that needs a password to open has to be unlocked first with Unlock PDF, using the password you already have. A PDF that opens freely but blocks editing is handled automatically: the restriction is removed so the file can be redacted, and the redacted copy has none.

Where your file goes

Redaction runs on our server, so the file is uploaded. It is kept only while you work: deleted within 30 minutes, or straight away when you choose "Redact another file" or leave the page. The file on your own device is never changed, which matters, because a redaction cannot be undone in the new file. Keep your original somewhere safe in case you marked too much.

Three mistakes worth avoiding

Missing the variants. You searched for "Jonathan Smith" and redacted every match. The document also says "J. Smith", "Mr Smith", "Jon" and jsmith@example.com. Search each form, and use Whole words so a short name does not match inside longer words.

Letting the box size speak. A box exactly the width of a short surname, next to a job title only one person holds, can say more than you meant. Where the length of a name matters, draw a slightly wider box.

Sending the wrong file. The redacted copy is a new file. It is surprisingly easy to attach the original from the same folder. Check the file name before you press Send.

Common questions

Can a redacted PDF be un-redacted?

Not when the content was genuinely removed. What you marked is not in the new file, so there is nothing to recover from it. A black shape drawn on top of text is a different matter: that can be deleted or copied around in seconds, which is why drawing is not redacting.

Is a white box safer than a black one?

The removal is identical; only the look differs. A white box makes the page look as though nothing was ever there, which can mislead a reader about what the document said. Black, or a box with a label such as REDACTED, makes clear that something was withheld.

Can I just delete the text in Word before converting to PDF?

Yes, deleting is a genuine redaction, as long as you delete rather than shade. Text shaded black in Word is still text, and it travels into the PDF. Check comments, tracked changes and the document properties too, because they convert along with the page.

Will the rest of my document still be searchable?

Yes. Only what you marked is removed; every other word stays real, selectable text. The one exception is a page the tool had to turn into an image to be safe, and the result screen names any such page.

Is it free?

Yes, without an account. Files can be up to 200 pages and 80 MB.