Redact a PDF properly
Black out text and images for good: affected pages are rebuilt as pictures, so redacted words cannot be copied back out of the file.
- Runs in your browser
- Files are never uploaded
- Free, no account
Why black rectangles keep leaking documents
Drawing a filled black box over a name in a PDF editor changes nothing about the name. The characters remain in the page's content stream, in reading order, at their original coordinates. Selecting the blacked-out area and pressing Ctrl+C copies them out. So does running the file through a text extractor, converting it to Word, or opening it in a browser's built-in viewer and using find.
This is not a theoretical risk. Court filings, government responses to freedom-of-information requests, redacted contracts published by regulators and pre-publication manuscripts have all been un-redacted this way within hours of release, repeatedly, for two decades. The tooling makes it look like the text is gone, and it is not.
What this tool does instead
- Each page is rendered to a bitmap at the resolution you pick, using the same engine your browser uses to display PDFs.
- Your black areas are painted into those pixels.
- A brand new PDF is assembled from the resulting images.
The original text objects, fonts, vector paths and embedded images are never copied into the output. There is no layer to peel back because nothing was layered — the black is the pixel value. And because the document is rebuilt rather than edited, the things that routinely leak a redacted name go with it: document metadata, the XMP packet, bookmarks and outlines, annotations and comments, form field values, attachments, and any incremental-update history the file was carrying.
The trade you are making
| Property | Before | After |
|---|---|---|
| Redacted text recoverable | Yes, trivially, if you only drew a box | No |
| Text selectable and searchable | Yes | No — the whole document becomes images |
| Screen-reader accessible | Yes | No |
| Metadata and comments | Preserved | Removed |
| File size | Original | Driven by the DPI you choose |
Losing the text layer is not a bug we would fix if we could — it is the mechanism. A tool that promises redaction and a searchable output is either doing selective text removal (which leaks through kerning, spacing and hyphenation artefacts) or it is not redacting at all.
Choosing a resolution
- 120 DPI — smallest file, fine for something that will only be read on screen.
- 150 DPI — the default. Comfortable screen reading, acceptable print.
- 200 DPI — the right choice if the document will be printed or filed.
- 300 DPI — archival quality, and several times the file size. Use it for anything going into a formal record.
Resolution has no bearing on how thoroughly the text is destroyed — it is gone at every setting. It only affects legibility and size.
Redacting every occurrence of a word
Type a name or number and the tool reads the document's text layer, computes where each match sits on the page and adds a box over it. It is a head start, not a guarantee, and you should always page through the preview afterwards. Matching happens within each run of text as the producer wrote it, so a name broken across a line, hyphenated, or split by unusual character positioning may be missed. Names also appear in ways a search will not catch: in a scanned signature, in a logo, in a photograph of a letterhead.
If the search finds nothing at all, the document has no text layer — it is a scan. Draw the boxes by hand; the redaction is exactly as effective either way.
Checking your work before you send it
- Use the verification panel: it renders a page straight out of the finished file, not the working preview.
- Open the downloaded PDF and try to select the redacted area. There should be nothing selectable anywhere.
- Run it through PDF to Text. An empty result is the proof.
- Check the document properties for a title or author that names the thing you redacted. The rebuild strips them, but confirm.
- Look for the same information elsewhere: a header, a footer, a page number scheme, a table of contents, an email thread quoted on page 9.
And keep the original somewhere safe. Redaction is irreversible by design.
Questions
Why is drawing a black rectangle not enough?
Because a rectangle is just more ink on top. The text is still in the content stream underneath, and selecting the area, running any extraction tool, or converting the file to text pulls it straight back out. This mistake has exposed witness names, salaries and unredacted evidence in real filings more than once. Real redaction has to destroy the data, not hide it.
What does this tool actually do to the file?
It renders every page to a bitmap at the resolution you pick, paints your black areas into those pixels, and builds a brand new PDF out of the images. The original text objects, fonts and vector paths are never copied across, so the redacted words do not exist anywhere in the output — nor do the metadata, XMP, bookmarks, attachments, annotations or form fields, all of which routinely hold copies of the very thing being redacted.
Will the result still be searchable?
No, and that is the point. Once pages are images there is no text layer left for anyone — including you — to search or copy. If you need a searchable redacted document you would have to run OCR on the output afterwards, which we cannot do in the browser and would not want to do silently.
Can I redact every occurrence of a name automatically?
Yes. Type the word and the tool reads the document's text layer, works out where each match sits on the page and adds a box over it. Always look through the pages afterwards: matching happens inside each run of text, so a name split across a line break, hyphenated, or stored with unusual character spacing may be missed. Automatic search is a head start, not a guarantee.
How do I verify it worked?
The tool renders a page straight out of the finished file so you can see the result rather than take our word for it. Then open the downloaded PDF and try to select the redacted area — there is no text there to select. Watching your browser's network tab while you work confirms the other half: not one byte is sent anywhere.