How to prepare a PDF for archiving
What PDF/A actually demands, what a browser can honestly do about it, and when you need a different tool entirely.
PDF/A is the format you are told to use when a document has to still open in twenty years. It is a real ISO standard, and it is strict in ways that surprise people: no encryption, no external dependencies, every font embedded, colour defined absolutely rather than left to the viewer. The idea is a file that carries everything needed to render itself, so a machine in 2046 with no access to your fonts or your colour profiles still shows what you saw.
What this tool honestly does
It is a metadata cleanup and a clean re-save. That is the whole thing, and you should know it before you use it rather than after someone rejects your filing.
The file is fully re-serialized — read in and written back out from scratch — which is genuinely useful and is the part people undervalue. It rebuilds the internal structure, which repairs a lot of quietly broken files: bad cross-reference tables, inconsistent object trees, the accumulated damage of a document that has been through six tools. Then it sets four pieces of document metadata. Producer, Creator, a creation date if the file has none, and a modification date of now.
Four setters and a re-save. There are no options — a file field and one button, no conformance level picker, nothing to configure — and that is honest, because there is nothing here to configure.
Run the archival cleanup
Upload the PDF
Drop in the document. The page count confirms it has been read.
Read the notice
The workspace says plainly that this is not a certified ISO PDF/A-1B converter. That notice is the most important control on the page even though it is not a control.
Click Convert for archival
The file is rebuilt and its metadata set. Your copy downloads with _archival on the name.
Validate it if it matters
If anyone is going to check this file against the actual standard, check it yourself first with a real PDF/A validator. This tool cannot tell you whether it passes.
Why it is not certified PDF/A-1B
Two requirements a browser cannot honestly meet for an arbitrary file, and both are structural rather than a matter of effort.
The first is the OutputIntent — an embedded, verified ICC colour profile that pins down what the colours in your document actually mean, so a viewer decades from now renders them the same way. Embedding bytes that claim to be a profile is easy. Establishing that the profile is correct for this specific document's colour spaces, and verified, is not something you can do to a file you know nothing about.
The second is font embedding. PDF/A requires that every font actually used is fully embedded, because a font on your machine in 2026 is not a font on a machine in 2046. Proving that for an arbitrary uploaded PDF — finding every font, confirming each is embedded, and embedding the missing ones from somewhere — is a hard problem, and the honest answer is that this tool does not attempt it. It does not add fonts. If your source file references a font it does not embed, the output references a font it does not embed.
So: if a court, a government filing system, a records department or a regulator requires PDF/A, use a tool built to produce and validate it. A desktop PDF suite with a PDF/A export, a print-to-PDF/A driver, or a validator like veraPDF that will tell you specifically what fails. Do not send this output somewhere that will run a conformance check on it and reject you. That is not modesty, it is the actual recommendation.
It overwrites your Producer and Creator. Permanently.
This is the consequence worth understanding, and it cuts against the reason you are here.
Producer and Creator are set to BabaPDF unconditionally. Not filled in if empty — overwritten. If your PDF said it was produced by Microsoft Word 16.0, or a scanner's firmware, or a specific version of a legal filing system, that string is gone from the output copy and cannot be recovered from it.
For most files this is nothing. But you are archiving, and archiving is frequently about provenance — the record of what made this document and when. A scan whose Producer field identifies the scanner model is carrying evidence. Running it through here to make it more archival strips exactly that. If provenance is why you are keeping the file, note those fields somewhere first, or do not run this at all. A cleanly re-saved file with a falsified toolchain is worse for an archive than an untidy file that says where it came from.
The creation date behaves better and is worth the contrast. It is only set if the document does not already have one — an existing creation date is preserved exactly as it was. The modification date is set to now, which is accurate, because you are modifying it now.
Encrypted files
PDF/A forbids encryption outright. An archival file that needs a password is a file that becomes unreadable the moment the password is lost, which is the opposite of the goal. Two cases, and they get different treatment.
- Permission-restricted files — the kind that open freely but claim to forbid printing or copying — are opened, decrypted, and written out clean. That is what PDF/A wants and it happens without you doing anything.
- Files that need a password just to open are turned away with a message. Nothing here can guess a password. Run it through Unlock PDF first, with the password you know, then archive the result.
So when is this the right tool?
- You want a tidy, structurally sound copy of a file for your own long-term storage, and nobody is going to run a conformance checker on it → yes, this is exactly it.
- A file has been through several tools and behaves oddly in some viewers → yes. The clean re-save fixes more than you would expect.
- Someone specified "PDF/A" and will validate it → no. Use a certified converter.
- The file's Producer metadata is part of what makes it worth keeping → no. You would be destroying the thing you are preserving.
The rebuild runs through pdf-lib in your browser, so the document never leaves your device — which for an archive of anything sensitive is the point.
