PDF Privacy Scanner in the browser is not a generate
/free-tools/pdf-privacy-scanner never spends a credit. Do not pay a model to do an in-browser file job.
/free-tools/pdf-privacy-scanner never spends a credit. Do not pay a model to do an in-browser file job.
The scanner reads document metadata and the text layer, then lists emails, phone numbers and common ID-shaped strings a recipient could copy out. The file is read into memory on your machine, parsed there, and reported there. Nothing is sent to a server. There is no account. Generation on Versely is the opposite: every model call costs credits, and there is no free generation tier. Mixing the two is how a PII hunt becomes a billed extract you did not need.
A finder is not a generate
The free-tools surface has one rule: nothing there calls a model, an API, or the network. PDF Privacy Scanner is pdf-lib plus pdf.js in the tab. The work happens on the device. That is why it stays free, and why it keeps working with the network off.
PDFs up to 150 MB and 500 pages. That is a browser ceiling, not a credit ceiling. Password-protected files fail at open — the protection is enforced by the format, not by this tool. Remove the password first.
A generate is a catalog row: an OCR model asked to “read this scan,” a stills row asked to “redraw the page without the email.” That row is billed. Finding strings in a text layer is not a style. It is a parse of this PDF. If you needed new pixels, pay the quote you see before confirm. If you needed to know what a recipient can copy out, stay in the browser tool.
Metadata plus the text layer
The scan reports two surfaces a recipient can recover without your permission.
Document metadata. Title, Author, Subject, Creator, Producer, Created, Modified — the fields pdf-lib can read from the info dictionary. Those strings are scanned the same way as page text. An author email in metadata is still an email.
The text layer. Each page’s text content is joined into a line and run through regular expressions: emails, phone numbers (10–15 digits), SSN-like (000-00-0000), NINO-like, and card-like numbers that pass a Luhn check. Hits are grouped by kind, with page number and a short context clip. Page 0 in that list means metadata, not a folio.
It is a finder, not a guarantee. Scanned pages have no text layer. A product code that happens to look like a National Insurance number will show up. Treat every hit as “this string is in the file,” then decide whether it is actually sensitive. If the tool found nothing but you can see the info on the page, the page is almost certainly a scan. Raster text needs OCR, which is a model, not a parser.
Pattern matches are not an identity bureau
The scanner looks like a recipient would: regular expressions, not an identity bureau. A model asked to extract PII will invent as readily as it finds. A regex on the text layer will not invent an email that is not in the bytes. It will also miss every letter that is only pixels. This tool does not redact, rewrite, or strip metadata. It reports.
The cost hub is the map of how jobs bill: per second, per 1,000 characters, per megapixel, per call, per export. None of that arithmetic applies here, because no job is dispatched.
Credits are for models; this is a parser
The expensive confusion is sending a packed PDF through a vision model with the instruction “list every email and ID.” You will pay the model’s meter and you may get hallucinations. Do the scan first, in the free tool. Generate later if you actually need new pixels. The free-tools index is the rest of these jobs. Credits are for the ones that invent.
FAQ
Is PDF Privacy Scanner actually free?
Yes. No account, no credit, no limit, because it costs nothing to run — the work happens on your device. Generation on Versely is not free. This tool is not generation.
Do my files get uploaded?
No. The file is read into memory on your own machine, processed there, and reported there. Nothing is sent to a server.
It flagged something that is not a real ID. Why?
The scanner looks like a recipient would: regular expressions, not an identity bureau. A product code that happens to look like a National Insurance number will show up. Treat every hit as “this string is in the file,” then decide whether it is actually sensitive.
Why is this free if Versely has no free generation tier?
Because it is not generation. Free tools run on your device; they cost nothing to run. /cost is the catalogue of jobs that do cost credits. Do not mix the two to “save” a parse.