You can unredact a PDF only when the words are still in the file under a black box: copy and paste, pdftotext or a plain text search brings them back. When the words are removed and the file is rewritten, there is nothing left to recover. Below is a test PDF with made-up data and every step to check OpenPdfEdit's redaction yourself, in your browser, with nothing uploaded. We ran it on 25 October 2026 and every check came back empty.
All three are one-page PDFs we wrote by hand for this test. Every name, address and number in them is made up, and the files are released into the public domain (CC0 1.0), so you can use them to see whether any tool lets someone unredact a PDF it saved. The secret to look for is PURPLE-OTTER-7731. It appears twice on the page.
SHA-256: original 1b3417c6dc2d6d3f51f14bf0b624bb3b10f226451a79072a0b5a7c64e9cb9e3d · box only bb4fc9a66cd32068c378008ee79e45f065e6af63e8ec57e85265e5ecd400fb6a · redacted 0d25c3b98ada22c4270059a8871d597ebde69ca159ccb77528cb6c90a0197ac8
There is no special trick to unredact a PDF that was redacted badly. A rectangle drawn on top of text hides it from the eye and from nothing else. Select the area and copy it, run a text extractor, or search the file's bytes, and the words come straight back. A second, quieter failure is the incremental save: the file is changed by adding a new version at the end, and the earlier version, words included, stays in the same file.
So the useful question is not whether someone could unredact a PDF in general. It is whether this particular file still carries the words. The checks below answer that for any file you have, and the box-only test file shows what a failure looks like, so you know the checks work before you trust an empty result.
redaction-test.pdf. The file is read by the page itself; it is not uploaded.PURPLE-OTTER-7731 under Words or phrases and press Find. It finds 2 matches on 1 page.The app's own message is a check it runs on the saved bytes, but it is still the app checking itself. In Step 2 you try to unredact a PDF without taking its word for anything.
Each check below is a way people try to unredact a PDF. Run each one on the file you saved, and then on redaction-test-box-only.pdf. The box-only file should fail every one of them; that is how you know the check can see a covered secret.
pdftotext redaction-test-redacted.pdf - | grep -c PURPLE-OTTER-7731
qpdf --qdf --object-streams=disable --preserve-unreferenced redaction-test-redacted.pdf expanded.pdf
grep -a -c PURPLE-OTTER-7731 expanded.pdf
grep -a -c PURPLE-OTTER-7731 redaction-test-redacted.pdf
grep -a -c %%EOF redaction-test-redacted.pdf
A count of 0 means the secret is not there, and nobody can unredact a PDF from text that is not in it. The second command in the last check counts end-of-file markers: 1 means the file was written once, from scratch, with no earlier revision appended behind it. On macOS, brew install poppler qpdf installs both tools; on Debian or Ubuntu, sudo apt install poppler-utils qpdf. On Windows, the Xpdf command-line tools from xpdfreader.com include pdftotext, and qpdf publishes its own Windows builds.
We ran Step 1 in the live web app in headless Chrome, then the checks below on all three files. The number is how many times PURPLE-OTTER-7731 was found.
| Check | Original | Box only | Redacted |
|---|---|---|---|
| pdftotext (Xpdf tools 4.06) | 2 | 2 | 0 |
| Text layer in pdf.js 3.11.174, the engine of Firefox's PDF viewer | found | found | not found |
| qpdf QDF, every stream decompressed (libqpdf 12.3.2, through pikepdf) | 2 | 2 | 0 |
| Raw bytes of the file | 2 | 2 | 0 |
| pypdf 6.21.2 and pdfminer.six text extraction | found | found | not found |
In the redacted file the rest of the page was untouched: the customer name, the email address, the account number and the label “Secret phrase:” all still extract, and the file has one end-of-file marker. Inside the page's content stream, the phrase was replaced by a gap of the same width, so the words after it on the line did not move, and two black rectangles were drawn where it had been.
OpenPdfEdit is an open source PDF editor (AGPL-3.0-or-later), so the reasons these checks cannot unredact a PDF it saved can be traced to code on GitLab:
redact_areas_impl.save_full.This test PDF is plain text in a standard font, the easy case. It does not test scanned pages, text inside form XObjects or redacting part of an image, which the code handles but this page does not demonstrate. Redaction also does not touch the document properties, XMP metadata, attached files, embedded JavaScript or hidden optional-content layers, and it does not apply clipping paths when deciding what is visible. A file can be clean on the page and still carry an author name or an attachment, so check those separately before you send a sensitive document on.
The fairest test is to try to unredact a PDF of your own: redact it, save a copy, and run the same four checks for the words you removed. If any check finds them, please tell us at hello@openpdfedit.com or open an issue on GitLab.
Download the test PDF, redact it in the browser, and see for yourself that nobody can unredact a PDF whose words are gone.
Guides: Two ways to save a PDF · All guides
Not from the file it saves, as far as these checks can tell. The redacted words are removed from the page's content stream and the file is rewritten from scratch, so there is no hidden text and no earlier revision to recover. The test on this page lets you confirm that on our test PDF and on your own documents.
Search the saved file for the words that were meant to go: copy and paste the page into a text editor, run pdftotext, expand the file with qpdf and search it, and count the %%EOF markers. If any of these finds the words, anyone could unredact a PDF like that, because the redaction only covered them.
No. Redaction removes what is drawn on the page. The document properties, XMP metadata, attached files, embedded JavaScript and hidden optional-content layers are left as they were, so check them separately.
A white box reads as a gap in the page rather than a mark saying something was taken out, which suits the ordinary case of tidying a document. Pick Black under Box fill when the reader should see where something was removed. The colour does not change what is removed.
Yes. The test files contain only made-up data and are released into the public domain (CC0 1.0), so you can redact them in any tool and run the same checks to see whether you can unredact a PDF each one saved.