When a thesis portal, an archive or a court's e-filing system asks for a document "in PDF/A format", it's fair to wonder why an ordinary PDF won't do. The file is already a PDF; it opens and reads fine. The difference with PDF/A isn't about today. It's about twenty years from now.
What PDF/A is for
PDF/A is a subset of PDF defined for long-term preservation, published as the international standard ISO 19005. Its goal fits in one sentence: however the file looks today, it should look exactly the same decades from now, on a different computer, in a different program.
To get there, PDF/A forbids or requires a few things that ordinary PDFs leave open:
- Every font must be embedded in the file. Fonts installed on the computer can't be relied on.
- Encryption is not allowed. An archived document whose password has been forgotten is useless.
- No JavaScript, audio or video. Nothing whose behavior in the future is uncertain.
- No dependence on outside resources. Images and fonts can't be loaded from a web address.
- Colors must be device-independent. The file carries a color profile, so a red is the same red on every screen.
- The document declares itself PDF/A in its own metadata.
Versions and letters: what do 1b, 2b and 3b mean?
PDF/A comes in several versions. The number is the version; the letter is the conformance level.
| Version | Year | Notable feature |
|---|---|---|
| PDF/A-1 | 2005 | The first and strictest; no transparency or layers |
| PDF/A-2 | 2011 | Allows transparency, JPEG2000 and layers |
| PDF/A-3 | 2012 | Like PDF/A-2, plus files of any type can be attached |
| PDF/A-4 | 2020 | The newest, based on PDF 2.0 |
The letters:
- b (basic): the visual appearance is preserved. This is the level most often requested.
- a (accessible): on top of appearance, the document's structure (headings, paragraphs, reading order) is tagged, which screen readers need.
- u (Unicode): all text can be extracted as Unicode; available in PDF/A-2 and 3.
PDF/A-3's attachment feature matters for e-invoicing: formats such as ZUGFeRD in Germany and Factur-X in France embed a machine-readable XML invoice inside the human-readable PDF.
Which version should you pick?
If the organization names a version, use it. If not:
- PDF/A-2b is the right choice most of the time. It handles most modern PDFs without loss.
- PDF/A-1b only when it's explicitly required. Documents with transparency (drop shadows, semi-transparent logos) can give it trouble.
- PDF/A-3b when you need to embed other files, such as an XML file or the source spreadsheet.
PDF to PDF/A offers these three levels and suggests 2b by default.
Prepare the document before converting
The things that cause the most trouble in a PDF/A conversion are easy to clean up beforehand:
- Remove the password. An encrypted file can't be PDF/A. If you know the password, take it off with Unlock PDF.
- Flatten forms and comments. Fillable fields and some comment types can cause conformance problems. Flatten PDF makes them part of the page.
- Strip scripts and attachments. Sanitize Document removes JavaScript, actions and embedded files (keep attachments if you're going for PDF/A-3).
- Fill in the title and author. Archives search on these. You can edit them in Document Properties.
If the document is very complex or the conversion keeps failing, the "Convert pages to images" option rebuilds each page as an image. It clears most conformance problems, but the text can no longer be selected or searched, so treat it as a last resort.
Always validate afterwards
A file that says "PDF/A" in its name, or even declares PDF/A in its metadata, isn't necessarily conformant. If you're handing it to an institution, check it with an independent validator. veraPDF, an open-source validator backed by the PDF Association, is free and widely used for exactly this. Submission portals often run their own check on upload as well.
If the validator reports errors, read them: "font not embedded", "transparency" and "encryption" are the three you'll see most, and the preparation steps above usually take care of them.
In short
- PDF/A is PDF with stricter rules so the document still looks the same years from now.
- If no particular version is required, choose PDF/A-2b.
- Before converting, remove passwords, flatten forms and strip scripts.
- Validate with an independent tool such as veraPDF before you submit.