PDF/A, PDF/X, PDF/UA: the alphabet after PDF

At some point a court filing system, a print shop or a procurement portal rejects your perfectly good PDF and demands something called PDF/A, and the natural questions follow: what is that, why does my PDF not qualify, and which one am I actually being asked for? The short answers: it's a restricted PDF, probably because your file depends on something outside itself, and there are three suffixes you might meet.

The pattern behind all of them is the same and makes the rest easy to remember. Ordinary PDF is permissive: a file may carry encryption, scripts, references to fonts it doesn't include, video, whatever. Each suffixed standard takes ordinary PDF and forbids the features that would break one specific use. A conforming file is a normal PDF with a shorter list of permitted features, and it opens in any reader; it has promised not to do certain things.

PDF/A: the archival standard

PDF/A (ISO 19005, first published 2005) answers an archivist's question: will this file still render, identically, in fifty years? Anything that depends on the outside world is banned. Fonts must be embedded, not referenced. Encryption is forbidden, since a password-protected archive is an archive one lost password from being garbage. JavaScript and multimedia are out, along with any content the file references but doesn't contain, though plain hyperlinks to the web are still allowed, because they're not something the file needs in order to draw itself. Colour must be unambiguous: either specified in a device-independent space, or accompanied by an output intent naming the device it was meant for. The file must, in effect, be a sealed box containing everything needed to draw itself.

You'll meet it wherever documents are kept for decades under legal obligation: court e-filing systems, government records, universities archiving theses, banks and insurers with retention rules. The letter after the version number matters, because it's what the rejection message names. b (basic) guarantees only that the file reproduces visually; this is what portals nearly always want, and what an ordinary export dialog can produce. a (accessible) additionally requires tagged structure and reading order, roughly level b plus the accessibility rules below, and is much harder to reach after the fact. u (available from PDF/A-2 on) sits between them, requiring that all text maps to Unicode. Successive parts (A-1, A-2, A-3) track newer PDF features, and PDF/A-3 adds the one people build products on: it allows arbitrary files to be embedded inside the archival document, which is the basis of hybrid e-invoices like Germany's ZUGFeRD and France's Factur-X: one file that is simultaneously a human-readable PDF and machine-readable XML data.

Conformance is fragile. Every PDF/A file declares what it conforms to in its XMP metadata, so a tool that strips metadata (including ours) removes a PDF/A file's claim to be one, and edits can break the underlying rules too. If an authority gave you a PDF/A and you only need to read or print it, don't edit it at all.

PDF/X: the print production standard

PDF/X (ISO 15930, first published 2001) comes from the graphic-arts industry, where a file that renders ambiguously costs a plate and a print run. The X profiles force every question a press could ask to have one answer inside the file: fonts embedded, images actually embedded rather than placeholders to be swapped in later, colour a press can reproduce predictably, and explicit page geometry: where the bleed ends, where the trim happens. The two profiles print shops name are PDF/X-1a, which restricts colour to CMYK and spot inks, and PDF/X-4, which allows ICC-managed colour including RGB alongside live transparency.

Unless you send work to commercial printers you may never touch it, and if you do, the print shop will tell you which profile they want. It's produced from design software at export time (InDesign and its peers have presets) rather than retrofitted onto an existing PDF. For ordinary printing on ordinary printers, the useful operations are a level below the standard: imposition, sheet layouts and page sizing.

PDF/UA: the accessibility standard

PDF/UA (ISO 14289, published 2012) is the profile for files a screen reader can navigate. Where A and X restrict, UA mostly requires: every piece of content tagged with its role (heading, paragraph, table cell), a defined reading order, alternative text on images, the document's language declared. It's the formal version of what accessible PDFs need in practice, and it's gaining teeth as accessibility law tightens; public bodies in particular increasingly publish to it.

How to produce and verify a conforming file

• You'll be told when a standard applies. The system that requires one names it, usually down to the level, like "PDF/A-1b", in its instructions or its rejection message, so there is no guessing involved.
• Conformance is created at export. Word's Save as PDF has an "ISO 19005-1 compliant (PDF/A)" option under its PDF settings; LibreOffice's Export as PDF has an Archive (PDF/A) checkbox with a version selector. The originating application is the reliable route.
• With no source document, convert then validate. Acrobat's Preflight and Ghostscript can convert an existing PDF toward PDF/A. Expect missing fonts and transparency to be the usual failure points.
• Verify rather than trust. A file can claim conformance in its metadata and still break the rules, so serious workflows run a validator. The best-known free one for PDF/A is veraPDF, an open-source validator built by the archival community.

The standards exist so a press operator, an archivist or a screen reader can rely on a PDF for one exacting purpose. For the ordinary contracts, scans, forms and reports of everyday life, plain PDF is what you want.