PDF to PDF/A

A regular PDF is allowed to lean on its surroundings. It can use fonts that are only installed on the machine that made it, link to images stored elsewhere, run JavaScript, carry encryption, or hold form logic that calls out to a server. Today it looks right; on another system in fifteen years it may not.

PDF/A forbids exactly those dependencies. Everything the document needs to render has to live inside the file: every font embedded, a colour profile attached, no scripts, no passwords, no external content. That is why a converted file is usually a little larger than the original — the fonts and the profile have been added. The extra size is the point of the format, not a side effect.

The number is the generation. PDF/A-1 is based on PDF 1.4 and cannot hold transparency or JPEG 2000 images; PDF/A-2 and PDF/A-3 are based on PDF 1.7 and can. PDF/A-3 adds one thing on top of 2: it allows files of any type to be embedded as attachments.

The letter is how much is promised about the content. b guarantees the pages look the same. u also guarantees that every character maps to Unicode, so text search and copy-paste work in the archive. a goes furthest and requires a tagged structure for screen readers, which most documents only have if the software that created them added it.

If a portal or archive names a level, use that one. If nobody does, PDF/A-2b is the sensible default, and PDF/A-1b is the choice for older systems that reject anything newer.

Hybrid e-invoice formats such as ZUGFeRD and Factur-X pair a page a person can read with an XML file that accounting software can process. PDF/A-3 is what makes that pairing possible inside one archival file, because it is the only generation that permits attachments of any type.

This converter produces the PDF/A-3 container: pick 3b, or 3u when the visible text should also be fully Unicode-mapped. It does not generate the invoice XML or embed it for you — that comes from your invoicing software. If you only need a readable invoice PDF, the Invoice Generator makes one in your browser.

A scan is a stack of page images, and it converts to PDF/A without trouble — the images are kept exactly as they are. What it will not do is become searchable on the way, and an archive you cannot search is only half useful.

So send scans through OCR PDF before converting. OCR adds an invisible text layer under each page image without changing how the page looks. Convert that result to PDF/A — 2u if you want the Unicode guarantee — and the archived copy is both fixed and findable by full-text search.