Drag over what has to disappear, then export. Every page is flattened before it leaves the tab, so nothing underneath can be selected, searched or copied.
Free today: 3/3 runs, documents up to 2 pages. All-Access removes both limits.
The output is an image-only PDF: text, links and the searchable layer are gone by design. What you cannot select, you cannot accidentally publish.
Draw a box by dragging across the page. Tap an existing box to remove it, or use Undo. Use Repeat on all pages when the same name or account number sits in the same place on every page.
Most PDF editors implement "redact" as a filled shape drawn over the page. The shape lives in one layer, the characters live in another, and the characters are still there: still selectable with Ctrl+A, still returned by a search for the name you thought you had hidden, still present when the page is copied into a word processor. This is not hypothetical. In January 2019 a defence filing in a federal criminal case reached journalists with black bars across the sensitive passages; pasting the page into a text editor returned the words under the bars in plain sight, because the bars had been drawn over live text instead of replacing it. The same mistake keeps appearing in FOIA releases, in discovery productions and in HR files sent to a tribunal.
This tool takes the other route. The export does not edit the page: it renders each page to pixels at the resolution you choose, paints your boxes into those pixels, and builds a new PDF from the resulting images. The original text is not copied, because the output has no text layer to copy from. That is the whole mechanism, and it is why the file gets bigger.
A supplier sends 24 pages of invoices that have to reach an external auditor. The packet carries a bank account number in the top-right corner of every page, the name of an unrelated customer on page 7, and a signature with a printed name underneath it on page 24. Four marks in total: one box over the account number, repeated across all 24 pages; one over the customer name; two over the signature block.
Export at 200 dpi. A Letter page renders at 1700 × 2200 pixels and compresses to roughly 250 KB as JPEG, so the packet comes out around 6 MB against 480 KB for the original, which was mostly digital text. The panel reports the numbers that matter: 24 pages, 51 boxes, 6.1 MB, and 0 characters of extractable text in the output. Page 1 is now a picture of an audit trail. The account number is 24 identical black rectangles with nothing behind them.
An image-only PDF is a photograph of a document. It cannot be searched, its text cannot be selected, a screen reader has nothing to read out, links are no longer clickable and it cannot be edited without OCR. On a text-heavy file it is five to eight times larger. In exchange, a document with no text layer cannot leak a text layer — and that is the only guarantee worth having when the point of the exercise is that a name must not travel.
If the recipient has to be able to search the file, flattening is the wrong instrument. Produce a redacted copy from the source document in an editor that rewrites the content stream and deletes the characters, or keep the searchable original internally and send the flattened copy to the people who must not have the text. What you should never do is send a "redacted" file whose text was never removed and hope nobody presses Ctrl+A.
No, and this is the mistake that turns up in court filings and FOIA releases. In most PDF editors a redaction box is a filled shape placed on top of the page; the characters it covers remain in the content stream, so Ctrl+A copies them and a search for the name still finds them. Here the export does not add a shape: each page is rendered to pixels, the boxes are painted into those pixels, and a new PDF is built from the images. The output has no text layer at all.
Not from the file this tool produces. The covered area is solid black pixels; there is no text object, no OCR layer and no hidden copy of the original page. What remains recoverable is whatever you did not cover — a repeated header, a line just above the box, or the same detail in another section. That is why the Repeat button exists, and why boxes should be drawn generously.
Yes, a scan is just an image on a page. One caveat: scanners and phone scanning apps often add a hidden OCR text layer so the file becomes searchable, and that layer still contains the names you can see. Flattening drops it along with everything else. Use 200 dpi rather than 150 for a scan, because the page is re-compressed a second time and thin print degrades faster.
All of them are gone, because the output is built from page images rather than from the original page tree. A signature is invalidated the moment the bytes change, so a flattened copy is not a signed document and cannot be presented as one. Keep the signed original and send the flattened copy as a reading copy.
Digital text costs almost nothing: 40 pages of written content is often 300 to 600 KB. Rendered pages cost whatever their pixels cost. A Letter page at 200 dpi is 1700 by 2200 pixels, around 250 KB as high-quality JPEG, so 24 pages lands near 6 MB. At 150 dpi that halves; at 300 dpi it doubles.
Three runs a day, and a free run handles documents of up to two pages. The counter resets at midnight UTC. Both limits disappear with All-Access, which also covers every other tool on the site. A run that fails does not consume a run — the count is only spent once a file has actually been produced.
No. The file is read with the browser's own file API, parsed and rendered inside the tab, and the output is assembled in memory. Nothing is transmitted, which also means the tool keeps working on a train once the page has loaded. The only network requests are the ones that read your remaining free runs.
Both limits go away with All-Access, along with the counter on every other file tool. One payment, no subscription.