A contract signed in 1998 still opens today, looks identical, and prints the same. Open the Word file it was exported from and you may get a warning about missing fonts, a broken table, or a refusal to open at all. That difference is the reason PDF/A exists.
What the standard removes
PDF/A (ISO 19005) is not a new file format. It is a strict profile of PDF — a list of features you must avoid so that a document renders the same decades from now, on software that does not exist yet. The requirements mostly remove dependencies:
- Fonts must be embedded. No relying on the reader’s system having the typeface.
- No encryption. The file must open without a password, now and forever.
- No external references. Audio and video playback, and external content streams, are out; everything needed lives inside the file.
- Defined color. Device-dependent color spaces need an output intent, so colors do not drift.
- XMP metadata required. Title, author, and document info sit in a standard, machine-readable block.
- No JavaScript, no launch actions. An archive file should not be able to run anything.
Versions and levels
| Profile | Year | Based on | Adds |
|---|---|---|---|
| PDF/A-1 | 2005 | PDF 1.4 | The original: no transparency, no JPEG2000 |
| PDF/A-2 | 2011 | PDF 1.7 | Transparency, JPEG2000, better compression |
| PDF/A-3 | 2013 | PDF 1.7 | Embedded files of any type (e.g. invoices with XML) |
| PDF/A-4 | 2020 | PDF 2.0 | Modern baseline for new archives |
The first three generations also come in conformance levels: b for visual appearance, a for accessibility and tagged structure, u for Unicode mapping. “PDF/A-2b” is the everyday archive target.
What it does not promise
PDF/A guarantees appearance, not usability. A scanned document with no text layer can satisfy PDF/A-1b perfectly — it is still a picture you cannot search. Accessibility is covered only by the a level. Files grow, because every font is embedded. And the label on a file proves nothing; only a validator such as veraPDF decides whether it conforms.
The rule: keep a normal PDF for daily use and a validated PDF/A copy for the archive. Convert at the end, not the beginning — and run the validator before you delete anything.