How to Convert a PDF to PDF/A (and What Gets Stripped Out)

Somewhere between a court e-filing portal, a university thesis submission and a records-retention policy, someone will tell you the file has to be PDF/A. It's a PDF built to survive being opened in thirty years by software nobody has written yet, which means everything the page needs in order to render has to live inside the file. The conversion is one step. What it changes about your document is worth two minutes of your attention.

What the conversion actually does

Three things change, and all of them are about cutting the file's dependence on the machine that opens it:

• Every font gets embedded. An ordinary PDF is allowed to name Arial and trust your computer to supply it. PDF/A isn't, so the glyph outlines are copied into the document. This is the main reason the output is larger than the input.
• Colour gets pinned down. Colours are converted to a device-independent space, so a particular red is defined absolutely rather than left to whatever the screen or printer decides it means.
• Anything reaching outside the file is removed. JavaScript, embedded audio and video, links that point at other files on disk, and encryption are all forbidden by the standard.

The output is PDF/A-2b. The 2 is the revision of the ISO standard, and the b is the conformance level: basic, meaning the file guarantees it will look the same in future. Keep that word in mind, it matters again further down.

Conformance wins over fidelity

When the converter meets something PDF/A doesn't allow, it has two options: keep the feature and hand you a file that fails validation, or drop the feature and hand you one that passes. It's set to drop. You get a file that validates, which is the entire point of asking for PDF/A in the first place.

For most documents nothing is lost, because most documents are text and images. Open the result and check the specifics if yours contains:

• Form fields wired to scripts, which stop being interactive
• Embedded media, which is removed rather than converted
• Links to companion files sitting next to the PDF, which won't survive

Password-protected PDFs are a separate case. The standard forbids encryption outright, so the file has to be unlocked before it can be converted at all.

Basic conformance does not mean searchable

This trips people up often enough to be worth stating plainly. PDF/A-2b guarantees the page will look right. It says nothing about the page having real text behind it.

Run a scanned document through and you get a valid archival PDF whose pages are still pictures of paper. Nothing is selectable, nothing is searchable, and a validator will pass it happily. If what your archive actually needs is findable text, OCR the scan first and convert the result, in that order — converting to PDF/A won't add a text layer, and running OCR afterwards would rewrite the file you just made compliant.

Common questions

Why is my PDF/A file bigger than the original?

Embedded fonts, almost always. A document set in a couple of typefaces can gain a few hundred kilobytes, and one using a large CJK font can gain several megabytes, since the whole character set travels with the file. That weight is the archive doing its job. Compressing afterwards is possible but self-defeating: it rewrites the file and you'd need to re-validate it, so compress first if you need to, then convert.

Will the file still be PDF/A after I edit it?

Assume not. Adding a signature, filling a form or re-saving in an editor that isn't PDF/A-aware can quietly reintroduce something the standard forbids, and the file keeps its PDF/A label while no longer conforming. Treat the conversion as the last step before you file or archive, not a stage in the middle.