What is measured here, and what is not. Everything claimed about Lyonite on this page was tested against a published corpus with known values planted in it. Everything about Adobe Acrobat is read from Adobe's own documentation — we have not run it, because it is paid software and we are not trying to contest it. So this page does not rank the two, in either direction. It says what Adobe documents, what we measured, and lets you put those side by side. The argument is narrow on purpose: you should not need a subscription to get a document properly cleaned.
Why Adobe Acrobat is the one worth comparing to
Most free PDF metadata tools clear the document information dictionary — Title, Author, Producer — and call the file clean. That is the visible layer, and it is not where a document usually gives itself away.
Adobe Acrobat's Sanitize Document has been the exception for a long time. It goes after the things people do not think to look for: the EXIF block inside a photograph, the name attached to every review comment, files attached to the document, scripts set to run on open. Until recently, if you wanted those gone you either paid for Acrobat or did it by hand.
What Lyonite now matches it on
- Document properties, including custom application fields
- XMP metadata at document and page level
- EXIF and GPS inside embedded JPEG images
- Annotation author names and timestamps
- Embedded file attachments, deleted rather than unlinked
- Embedded JavaScript and automatic actions
- The trailer document ID that links revisions together
- A folder at a time — drop many files and each is swept, then zipped
Each of these is checked by a reproducible test with known values planted in a corpus, not by reading our own feature list. For context on how unusual that is among free tools, none of the six others we tested removed the GPS from an embedded photograph.
Batch used to be on the list below, and is not any more. Adding several PDFs to the remover now runs the same sweep over every one and returns them as a zip, with a line per file saying what came out of it. Up to fifty at a time, still without anything leaving the device — which is the part a desktop application gets for free and a web tool usually does not.
One difference worth naming, because it cuts both ways: Acrobat removes review comments entirely, and Lyonite removes the author and timestamp while leaving the comment text. If the comments themselves are the sensitive part, Acrobat is doing what you want and Lyonite is not. If you need the comments to survive and their authorship not to, it is the other way round.
Three things Adobe Acrobat does that we do not
These are not close calls we lost. They are jobs Lyonite does not attempt, and if one of them is your job, no amount of the above helps you. Better to read it here than find out after.
Encrypted and password-protected files
Acrobat decrypts and sanitizes in one pass. Lyonite refuses these outright rather than producing something it cannot guarantee. Remove the password first, then clean the decrypted copy.
Applying redactions
Acrobat can black out text and actually remove the characters underneath. Lyonite does not redact at all — and a rectangle drawn over text in any tool leaves the text fully selectable underneath it.
Unusual image encodings
Acrobat re-encodes JPEG 2000 and Flate-wrapped images to strip their metadata. Lyonite reports those as skipped instead, because rewriting them means re-compressing and changing how they look.
The one thing Acrobat does not do
Acrobat sanitizes and then tells you it sanitized. Confirming that a specific document properly lost a specific thing means opening the result and checking, which almost nobody does — and, as the removal test shows, the two obvious ways of checking both under-report.
Lyonite re-opens the file it just produced, reads it with the same inspector the viewer tab uses, and reports from that. It also names what it did not touch, and lists any image it skipped along with the reason. A cleaner that only lists its wins teaches you to trust it further than it has earned.
Check this yourself. If we got it wrong, tell us.
Every number here came from files you can download and a checker you can run, so you do not have to believe us — clone the corpus and get your own result.
If it disagrees with ours, or if you build one of the tools named here and we measured it unfairly, out of date, or with a setting you would not have used, send it to hello@lyonite.com.
We reply within 48 hours. If you are right, the page is corrected with the date on it and your correction credited, and the old number stays visible so the change is legible. If a tool has since been fixed, that is the update we most want to publish. Nothing here is worth defending past the point it stops being true.