What is PDF/A, and when do you need it?
Courts, universities and government archives ask for it by name. It is the same format with the unreliable parts removed.
A PDF that promises to still work later
PDF/A is an ISO standard for long-term archiving — a restricted subset of PDF, not a separate format. Any PDF/A file is a valid PDF and opens in any reader. The difference is what it is forbidden from containing.
The problem it solves is real. An ordinary PDF can reference a font installed on the author's machine rather than embedding it, so the file renders correctly today and substitutes something wrong on a computer that lacks that font. It can contain video, JavaScript, or a link to content on a server that will not exist in a decade. It can be encrypted with a key nobody kept.
PDF/A forbids all of that. The rule behind every restriction is the same: everything needed to render the document exactly must be inside the file, and nothing may depend on the outside world.
What it actually bans
The specifics are worth knowing, because they explain most conversion failures.
- External font references. Every font must be embedded, and embedded with permission to be embedded — some commercial fonts are licensed in a way that forbids it, which is a genuine reason a conversion can fail.
- Encryption. A PDF/A file cannot be password protected at all. An archive that cannot be opened is not an archive.
- JavaScript and executable content. No scripts, no launch actions.
- Audio and video. Nothing that needs a codec which may not exist later.
- Transparency, in the older PDF/A-1 level. This one surprises people, because drop shadows and soft edges are ordinary design features. PDF/A-2 relaxed it.
- Device-dependent colour. Colour must be defined against an embedded profile, so a colour means the same thing on any device rather than whatever the monitor decides.
The versions and conformance levels
You will see designations like PDF/A-1b or PDF/A-2u. The number is the generation of the standard and the letter is how strict the conformance is.
PDF/A-1 came first and is the most restrictive — notably no transparency, no layers. PDF/A-2 allows transparency, layers and JPEG 2000, and permits embedding other PDF/A files. PDF/A-3 adds one significant thing: it allows arbitrary file attachments, which is why it is used for electronic invoicing formats that pair a human-readable page with a machine-readable data file.
The letters matter more day to day. Level B (basic) guarantees the document will look the same. Level A (accessible) additionally requires the structural tagging that makes it navigable by a screen reader, which is a much higher bar. Level U sits between them: visual fidelity plus a guarantee that the text can be reliably extracted as Unicode.
If someone asks for PDF/A without qualifying it, they almost always mean PDF/A-1b or PDF/A-2b.
When you actually need it
Mostly when someone tells you to. Court electronic filing systems frequently mandate it, as do many university thesis submissions, national archives, and long-retention regulatory records in finance and healthcare.
For everyday documents it buys little. An ordinary PDF exported from modern software already embeds its fonts and avoids exotic features, and it will open perfectly well in twenty years. PDF/A is a guarantee, not an improvement — worth having when someone needs the guarantee, and unnecessary overhead when nobody does.
One consequence is worth planning around: because PDF/A forbids encryption, you cannot password protect an archival file. If a document needs both properties, they have to be separate copies.
Converting, and why it fails
Conversion works by embedding what is missing and stripping what is forbidden. Fonts get embedded, colour profiles get attached, JavaScript and multimedia get removed.
Failures cluster in two places. A font licensed against embedding cannot be embedded, and the only fix is to substitute a different font. And a scanned document converts cheerfully to PDF/A-1b while being useless at level A or U, because there is no text in it to tag or extract — it is a picture. Making a scan genuinely archival means running optical character recognition first.
Frequently asked questions
What is the difference between PDF and PDF/A?
PDF/A is a restricted subset of PDF designed for archiving. It forbids anything that could stop the file rendering identically later — external font references, encryption, JavaScript, audio and video — and requires everything needed to display the document to be embedded in it.
What do PDF/A-1a, PDF/A-1b and PDF/A-2b mean?
The number is the generation of the standard and the letter is the conformance level. B guarantees visual fidelity, U adds reliable Unicode text extraction, and A additionally requires full accessibility tagging. PDF/A-1 forbids transparency; PDF/A-2 allows it.
Can a PDF/A file be password protected?
No. The standard forbids encryption outright, on the grounds that an archived document nobody can open has failed at being archived. If you need both, keep an archival copy and an encrypted copy separately.
Do I need PDF/A for normal documents?
Almost never. It matters when an institution requires it — court filings, thesis submissions, regulatory retention. For everyday files an ordinary PDF from modern software already embeds its fonts and will open fine for decades.
Why did my PDF/A conversion fail?
Most often a font whose licence forbids embedding, which leaves substitution as the only route. Transparency also fails against PDF/A-1 specifically, and converting to that level means flattening the artwork.
Can a scanned document be PDF/A?
It can meet level B, since that only concerns appearance. It cannot meaningfully meet level A or U, because those require extractable and tagged text and a scan contains none — it is an image of a page. Run OCR first.