Redact personal data from your PDFs in a minute, not an afternoon.

Yes, and there is a consequence worth knowing. The final PDF is not produced by drawing a rectangle over the original text: each page is rebuilt as an image with the black already burned in. The document you download contains no text objects at all, which means nothing can be selected, copied or searched — neither the covered data nor the rest of the document. In exchange the file is larger (typically 2 to 5 times the original). On screen and in print it looks identical, and that is exactly what an anonymised document should be.

No. The PDF is opened by the browser on your own computer. Only the text of each page reaches the server — what the AI needs in order to decide which fragments are personal data. The page image never leaves your browser, and the redacted PDF is rebuilt locally. We do not keep the text you send or the results beyond the lifetime of the request.

It happens, and it is planned for: that is why the machine does not get the last word. The detection list is editable — every item toggles on and off with a click — and there is a tool to cover any area of the page by hand. Above all, before downloading you see the pages exactly as they will be saved and you have to confirm you checked them: until you approve, no file is produced. Poor scans and handwriting reduce reading accuracy. Always check the document before handing it over: responsibility for what you publish stays with you, and the tool is built to give you the controls to verify it.

Because on text they are not a redaction. Blurring and pixelation do not remove the information: they scramble it following a rule. If someone suspects a tax code or a date is underneath, they can generate every possible value, run it through the same filter and compare until the pixel pattern matches. On short, predictable data — exactly what this tool covers — it works, and it has already been done to real documents.

So there are two styles and neither of them blurs: the black bar, which covers, and replacement with a label, which writes in place of the data what kind of information was there ([NAME], [IBAN], [REDACTED]) and keeps the document readable — handy for a filing or a reusable template. Either way the original pixels are destroyed, not shuffled.

Yes. If a page is an image with no text, optical recognition kicks in automatically, and it runs in your browser rather than on our servers. The first scanned page takes a few extra seconds while the recognition engine loads.

One page processed. A 12-page contract uses 12 credits. Pages with no text cost nothing, and if a job fails the credit is returned automatically. You always see what a document will cost before you start.

No, and that is deliberate. An anonymised document has to stay usable: who issued it, for how much, under what terms. What gets covered is the data of natural persons. If in a specific case you want to cover the rest as well, you do it with the manual box.

Because the payment system is currently a simulation: you can walk the entire path — signup, credits, running out, changing plan, cancelling — without entering a card and without anything being charged. Everything else, including the AI work and the PDF redaction, is real and working.