PDF/A for Long-Term Archival: 2026 Conversion Guide
The differences between PDF/A-1, -2, and -3, what counts as compliant, and how to produce archival PDFs for legal and government submissions.
PDF/A for Long-Term Archival: 2026 Conversion Guide
PDF/A is not a different file format from PDF. It is a constrained subset of PDF defined by ISO 19005 that guarantees a document will render the same way decades from now. Fonts are embedded. Color spaces are declared. JavaScript, external links, encryption, and audio are all forbidden. You end up with a file that an archivist's great-grandchild can still open.
If you are submitting to a court, a national archive, or a regulator, PDF/A is usually mandatory. The federal courts in the United States require PDF/A for all CM/ECF filings since 2021. The UK National Archives mandate PDF/A-2b or PDF/A-3b for transfer. The European Commission's eIDAS framework references PDF/A-2 for long-term signature preservation.
PDF/A-1, -2, -3: the practical differences
The three conformance levels most people will encounter:
| Version | Based on | Year | Notable allowances |
|---|---|---|---|
| PDF/A-1 | PDF 1.4 | ISO 19005-1:2005 | No transparency, no layers, no JPEG2000 |
| PDF/A-2 | PDF 1.7 | ISO 19005-2:2011 | Transparency, layers, JPEG2000, PDF/A files embedded |
| PDF/A-3 | PDF 1.7 | ISO 19005-3:2012 | Any file type may be embedded (XML, CSV, source docs) |
Each version has conformance sub-levels: b (basic visual fidelity), u (Unicode text extraction), and for PDF/A-2 and -3 also a (accessibility, full structure and tagging).
Most legal filings accept PDF/A-1b or PDF/A-2b. Government tenders that want the original source attached (a filled XML form, a signed XBRL financial statement) require PDF/A-3. Accessibility mandates like the EU Web Accessibility Directive pull you toward the a level.
If you do not know which version to pick, PDF/A-2b is the safe default. It handles modern PDF features that PDF/A-1 chokes on, and conformance is straightforward.
What breaks during conversion
The common failure modes when converting a normal PDF to PDF/A:
- Unembedded fonts. Any font the document uses must be embedded. Standard PDF viewers will substitute Helvetica if a font is missing; PDF/A forbids this. Scan the source with a tool that reports font status before you convert.
- Undeclared color spaces. Every color must reference either a calibrated color space or an output intent. Setting a document-level ICC profile (sRGB IEC61966-2.1 for screen, FOGRA39 for European print) fixes this for most files.
- Transparency. Allowed in PDF/A-2 and -3, forbidden in PDF/A-1. If you need PDF/A-1 compliance, transparent objects get flattened, which can change how layered graphics look.
- JavaScript and forms. Any interactive JavaScript must be stripped. AcroForm fields are allowed, but the validation logic attached to them is not.
- External references. Hyperlinks to external URLs are technically allowed (as annotations), but anything that loads external content (remote images, streamed video, external CSS) is out.
A PDF with all five problems is not unusual. A marketing brochure exported from InDesign typically has transparency, web-optimized fonts subset incompletely, and links to campaign landing pages. Running it through a compliance check first saves a round of rejection later.
Embedded fonts, specifically
Fonts are the thing that catches most people. Three common mistakes:
- Subsetted fonts without the full glyph table. A PDF can embed only the glyphs actually used. This is allowed in PDF/A as long as the CIDSet or CIDToGIDMap is present and correct. Some older export pipelines omit the map and fail validation even though the glyphs are physically in the file.
- Type 3 fonts. Bitmap fonts generated by some scanning software are not archival-safe. Re-OCR the document or re-render with a real font.
- Licensing flags. A font that has its embedding bit set to "no embedding" (some foundry fonts) legally cannot go in a PDF/A. You have to substitute a licensed alternative.
A validator like veraPDF (free, open-source, run by the PDF Association) will catch all three. Run it on the output before you submit.
Compliance for legal and government work
For US federal court CM/ECF: PDF/A-1b or PDF/A-2b, under 50 MB per document, no password protection. Each document gets an OCR text layer if it started as a scan.
For EU public procurement (TED): PDF/A-2b minimum, PDF/A-2a preferred for bids that include structured data. Signatures must be PAdES-LTV.
For UK National Archives transfer: PDF/A-2b or -3b, documented in the Digital Records Infrastructure guidance.
If your documents contain sensitive metadata (authors, revision history, comments), strip it before you convert. PDF/A preserves XMP metadata by design, and a redacted-looking document with a 400-revision edit history is a frequent accidental disclosure. A local converter like the one at /convert runs the conversion in your browser, so the pre-compliance copy never leaves your machine. For ingesting a case folder of a few hundred files at once, /batch handles the full set without per-file reuploads.
A minimal correctness checklist
Before submitting a PDF/A anywhere, verify:
- Fonts embedded (full or subset with CIDSet)
- Output intent present, ICC profile declared
- No JavaScript, no encryption, no audio/video
- Metadata in XMP, synchronized with DocInfo
- Validates clean in veraPDF against the target conformance level
If all five pass, the file will survive the rendering engines of 2050 just as cleanly as today's.