How to Redact Sensitive Information from a PDF

Quick answer: Real redaction removes the underlying text and objects from a PDF, not just paints over them. A black rectangle drawn as an annotation or shape typically leaves the original text layer intact underneath, meaning it can still be selected, searched, or recovered by removing the shape. Genuine redaction flattens the page so no text objects remain behind the blacked-out area at all.

A black box over a sensitive line looks like redaction, but whether it actually is one depends entirely on what happened to the text underneath it. This distinction has caused real, public leaks, documents released with "redacted" sections that were fully recoverable by simply copying the hidden text out. The same failure mode shows up whether the document is a court filing, a government disclosure, or an internal report, anywhere a black box is treated as the finished product rather than a visual cue that real removal still needs to happen underneath it.

What real redaction needs to do

Real redaction removes the underlying content, not just its visual appearance. A properly redacted PDF has no selectable text, no searchable content, and no way to recover the original characters in the redacted area, because they're genuinely gone from the file, not covered up.

How flattening to an image achieves this

PDF Redact removes the content itself from the areas you mark, then flattens the page to an image, so there's no text layer left underneath at all, only pixels. That distinction matters: a page that's still made of selectable text objects with a shape drawn on top can always have that shape peeled back off, but a page that's been rendered down to an image has no text objects left to recover in the first place, redacted or otherwise, on that page.

The classic mistake: covering instead of removing

Many general-purpose PDF editors let you draw a black rectangle as an annotation or shape on top of a page. Visually, this looks identical to real redaction. Functionally, it's completely different: the original text layer is untouched underneath, anyone who selects that area of the page, copies it, or deletes the black shape gets the original content back, fully intact. This exact mistake has exposed sensitive names, figures, and case details in documents that were genuinely believed to be redacted before release.

Why annotations don't touch the underlying content

A PDF page's actual text and graphics live in what the format calls a content stream, essentially the recorded instructions for what to draw and where. An annotation, or a new shape added on top, is typically stored as a separate object layered over that content stream rather than replacing anything inside it. The two coexist in the same file: the original content stream is exactly as it was, and the new shape is just drawn after it, which is why removing the shape, or opening the file in a viewer or tool that ignores it, reveals the original text was never actually gone.

What else can leak besides the page content

Visible text isn't the only place sensitive information hides in a PDF. Document metadata, author name, company, software used, sometimes even a revision history, can carry information you didn't intend to share, and it's invisible unless you specifically go looking for it. PDF Metadata Editor lets you inspect and clear these fields, worth doing as a second pass on any document you're redacting before sharing, since a redacted page with an author field reading someone's full name defeats part of the purpose.

Revision history you might not know is still in the file

The PDF format supports incremental updates, where an edit can be appended to the end of a file as a new revision rather than rewriting the whole document. Depending on how a document was edited and saved, older revisions can, in some cases, still be physically present in the underlying file even after the visible page looks fully updated. This is a separate concern from an annotation covering text, it's about the file's own edit history rather than what's drawn on a page, and it's part of why a genuinely clean export (rather than an incrementally saved edit) matters for anything sensitive.

Checking your work before you share

Before sending a redacted document anywhere, do the same check an attacker would: try selecting text in the redacted area, and try searching the document for a word you redacted. If either one still finds the "removed" content, the redaction didn't actually work, it was covered, not removed, and needs to be redone properly before the file goes anywhere.

Using a text comparison as a second check

Beyond manually selecting and searching, running the original file and the redacted version through PDF Compare shows exactly what text differs between the two documents, line by line. If a phrase you meant to remove still shows up unchanged in that comparison, the redaction didn't take, which is a more systematic check than relying on spot-checking a page by eye.

Common mistakes worth avoiding

Drawing a shape instead of using a dedicated redaction tool. It looks finished on screen and is exactly the mistake that has caused real public leaks of "redacted" documents.

Redacting the visible pages but not the metadata. An author name, company field, or embedded comment can carry the same sensitive information the visible redaction was meant to remove.

Assuming a scanned page can't leak text. True for the image itself, but not if OCR added a hidden searchable text layer on top of it.

Sharing a file whose cloud version history still contains the unredacted original. Redacting the document itself doesn't clear a separate revision history kept by the storage platform.

The short version

A black box on a PDF page only counts as redaction if it actually destroys the underlying content, not just visually covers it, if the text is still selectable or searchable afterward, nothing was really removed. Check both the visible content and the document's metadata before considering a redaction complete, and verify by trying to select or search for what you redacted, or by running a text comparison against the original, rather than just judging by how the page looks. PDF Redact handles the removal itself, PDF Metadata Editor clears what the visible page never touches, and PDF Compare confirms the result.