Redact text in a PDF
in your browser.

Covering text you don't want shared

The reasons to redact a PDF are usually about hiding parts of a document while leaving the rest readable. A contract with a counterparty's home address is going to a third reviewer who shouldn't see that address. A medical referral is being shared with an insurer where the patient's identifier should be obscured. A financial statement is being attached to a board pack but the unit-level account numbers should be covered. A project proposal is being shown to candidate vendors with the budget line blacked out so vendors propose without anchoring on the figure. The job here is small and predictable: type the words or phrases you want hidden, and the tool draws an opaque black rectangle over each occurrence on every page.

Read this carefully before redacting anything sensitive: this is visual redaction, not forensic redaction. The black rectangles cover the text on screen and in print, but the underlying text stays in the file. Anyone with a PDF tool can extract it via copy-paste, search, or content-stream inspection. This is appropriate for situations where casual readers shouldn't see the text — sharing a draft, hiding non-public account numbers from a colleague, masking your phone number on a screenshot. It is not appropriate for releasing documents under FOIA, court filings, or anything where an adversary will look hard.

How the matching works

You enter one or more patterns. Each pattern is matched as a case-insensitive whole-substring search against text fragments inside the PDF. When a fragment contains your pattern, the entire fragment's bounding rectangle gets a black box drawn over it.

Practical consequences:

• Substring match, not whole-word. Pattern art covers «art», «artist», «smart», «martin». Make patterns specific enough — full names, full account numbers, full email addresses.
• Case-insensitive. Pattern Anna covers «Anna», «anna», «ANNA».
• No regex, no wildcards. What you type is the literal sequence to match.
• Coverage is per-fragment. The PDF stores text in fragments; the bounding rect drawn covers a whole fragment. If your pattern hits one word in a fragment that includes more text, the surrounding text gets covered too. Usually this is what you want; occasionally you'll see more covered than expected.
• Cross-fragment patterns are missed. «John Smith» is one fragment in some PDFs and two in others. If it's split across fragments, the pattern won't match on the joined string. Add the parts as separate patterns: John Smith, Smith, John.

The black rectangle has 1 point of padding around the detected bounding box, to account for sub-pixel placement of glyphs and avoid stripes of the original text peeking out at the edges.

What's hidden vs. what's still in the file

This is the section that matters more than any other. The black rectangle is a drawing instruction in the page's content stream. The original text is also in the content stream, underneath. PDF readers display them in order: text first, then rectangle on top, so the rectangle wins visually. But the text itself isn't deleted.

What this means concretely:

• Copy-paste retrieves the original text. Selecting the redacted area in Adobe Reader and copying gives you the words you tried to hide. The black rectangle isn't selectable; the text under it is.
• Search engines find the text. If the redacted PDF is uploaded to a website, a search for the redacted name will surface the document.
• Re-saving in a different PDF tool may keep both layers. Some tools rebuild the content stream and could drop redactions; others keep them.
• Forensic tools extract the original easily. Any PDF parser can list all text content streams. The rectangle is just a separate drawing.

For situations where the underlying text really must be gone — government releases under transparency law, evidence in court, anything that goes to an adversary — you need a different workflow: re-export the source document with the sensitive text deleted, or convert the page to an image and re-OCR (which permanently removes the text layer).

What survives and what doesn't

• Visible page content stays the same. Everything not matched stays exactly as it was — same fonts, same layout, same images.
• Bookmarks, annotations, links, form fields all survive. The rectangles are drawn additions; nothing else is rebuilt.
• The original text is preserved underneath. See the section above. This is the central limitation.
• Digital signatures are invalidated. Any modification to the document, including adding rectangles, breaks the signature binding.
• Encrypted PDFs need to be unlocked first. Run through unlock-pdf with the password before redacting.

If something looks off

• "At least one redaction pattern required." The pattern list was empty. Type at least one word or phrase.
• The pattern matched but covered more text than expected. The PDF stored that line as a single fragment, and the rectangle is per-fragment. Usually acceptable; if it's a problem, the source needs to be re-exported with the text removed properly.
• The pattern didn't match what I expected. Three usual causes: (1) the text is in a scanned image, not a text layer — the tool can't see image text; first OCR via pdf-to-txt or a dedicated OCR tool. (2) The pattern spans fragments — split into shorter patterns. (3) The PDF uses non-Latin characters with broken encoding; copy-pasting from a viewer to check what's actually there is the fastest diagnostic.
• I redacted but a colleague extracted the text via copy-paste. Working as documented — visual redaction only. For real removal, re-export from the source.

Practical notes

1. Use this for visual-grade hiding, not for adversarial scenarios. Hiding a name in a draft you're sending to a friend who won't try to extract it: fine. Hiding a name in a court filing where opposing counsel will run forensic tools: not fine.
2. Specific patterns over generic ones. Redact +1-555-0123, not 0123. Redact [email protected], not jane. Specific patterns are less likely to over-match and equally hide what you intended.
3. Verify by copying. After redacting, open the result in a viewer, select the redacted region, and copy. If the original text comes out, you've confirmed the limitation — proceed only if visual hiding is sufficient.
4. For lasting removal, use the source. If the document was generated from Word, edit the Word file and re-export. If it was from a database, re-query without the sensitive fields. The PDF redaction tool is for cases where you don't have the source.
5. The original file stays untouched. What you download is a new redacted PDF; the source on your disk is unchanged.

What happens to your file

Redaction runs in your browser. The patterns you type and the file content never leave this tab. Open DevTools and watch the Network tab during the operation — there are no outbound requests carrying the file content or the patterns. The PDF stays on your disk; the redacted version is a new download alongside it.