There is a specific kind of embarrassment that follows a leaked redaction: a court filing, a settlement, or a contract where someone drew neat black rectangles over the sensitive parts and published it, only for a reader to select the text, paste it into a document, and read every word underneath. The black box was a graphic drawn on top. The text layer never moved. If you are removing sensitive information from a PDF, the goal is not to cover it up. It is to make sure it no longer exists in the file at all.

Why a black box isn't enough

A PDF is not a flat picture. On a typical text page there are two things stacked together: the visible glyphs you read, and an invisible, selectable text layer that carries the actual characters. When you drop a black rectangle over a paragraph in a normal editor, you add a new shape to the page. You do not delete anything beneath it. The characters are still in the file, still selectable, still searchable, and often recoverable by simply moving the box or extracting the text. The same trap catches images: lowering the opacity of a shape, using a highlighter, or placing a colored bar can all look opaque on screen while leaving the underlying content perfectly intact in the data. Anyone with a few minutes and a free PDF viewer can undo the illusion. Redaction is not a visual effect. It is the permanent removal of data.

How proper redaction destroys the text

The Redact tool takes a firmer approach: after you mark the regions to remove, it permanently flattens the affected pages to images. Flattening means the page is re-rendered as a picture with the marked areas blacked out, and the original selectable text layer for those pages is discarded. There is nothing left to select, copy, or search, because the words are simply no longer present as text. What remains is an image showing black where your sensitive content used to be. This is a deliberate trade-off. Because the page becomes an image, text on redacted pages is no longer selectable or searchable, and file size can grow. That is the cost of certainty. If you later need machine-readable text from the surrounding, non-sensitive content, you can run the OCR tool on the flattened result to rebuild a text layer from the visible pixels, without ever resurrecting what you removed.

What actually needs to come off

Redaction discipline is about being thorough, not just hitting the obvious line, because attackers and journalists reassemble people from fragments. Before you share a document, scan for every category of identifier.

  • Full names, and partial names that combine with other details to identify someone
  • Account, card, policy, case, and reference numbers
  • Government identifiers such as Social Security, tax, or national ID numbers
  • Addresses, phone numbers, and email addresses
  • Dates of birth and other pinpoint dates that narrow down an individual
  • Handwritten or drawn signatures, which are themselves sensitive biometric-style data
  • Salaries, balances, and figures that reveal more than the story needs

Watch the places sensitive text hides in plain sight: repeated headers and footers, watermarks, tables of contents, and cross-references that quote a redacted value elsewhere in the document. Redacting page five is pointless if the same account number appears again in the appendix, and email threads pasted into a document often repeat a name or address dozens of times. Work through the whole file, not just the passage that first caught your eye, and verify your work afterward by trying to select text over each black area before you send anything. If your cursor highlights anything you thought you had removed, the redaction did not take, and you need to run it again rather than trust the appearance of the page.

Don't forget the metadata

Even a perfectly redacted body can betray you through the file's metadata: the invisible properties that travel with a PDF. Author names, the original file title, the software and template used, and creation or modification timestamps can all leak the identity, department, or history you were trying to protect. A document titled with a client's name in its properties has already given the game away before anyone reads a page. Use the Edit Metadata tool to review and clear the author, title, subject, and keywords fields after you redact. Treat it as the final step of the same job: redaction removes what is on the page, and metadata cleanup removes what is attached to the file. You want both gone before the PDF leaves your machine. One habit makes all of this safer: because these tools run entirely in your browser, the document you are redacting is never uploaded anywhere during the process, which matters most for exactly the kind of file you would redact in the first place.