The Best File Formats for Long-Term Document Storage (Why PDF/A Wins, and What to Avoid)

CodexFuse Team · 2026-11-04T03:51:52.022Z · 13 min read min read

The best file format for long term document storage is PDF/A, the archival subset of PDF defined by the ISO 19005 standard. It wins because of one principle — self-containment: every font, color profile, and piece of metadata a PDF/A file needs is embedded inside the file itself, so it renders identically decades from now no matter which software opens it. The runner-up formats are simpler on purpose: uncompressed TIFF or PNG for scans, and plain UTF-8 text or CSV for data.

If you already have documents in plain PDF, the practical move is converting them now rather than hoping they survive: run our PDF to PDF/A converter to turn regular PDFs into ISO 19005 archival files, then validate them with a proper checker. This guide explains what PDF/A is, which version to pick, what to use for scans and data, and which formats to avoid — with every claim sourced from the standards and the institutions that write them, as of November 2026.

What Is the Best File Format for Long-Term Document Storage?

The best file format for long-term document storage is PDF/A, defined by ISO 19005, because it is the only mainstream document format engineered specifically to survive decades of software change. A PDF/A file carries everything it needs to render identically in fifty years: all fonts fully embedded, device-independent color, standardized XMP metadata, and zero dependencies on external files, proprietary codecs, or script execution. National archives in Germany, the US, and the UK mandate or recommend it for permanent electronic records, according to Mapsoft's 2026 guide to the standard.

That does not mean PDF/A is the answer for everything. The format-family decision depends on what you are storing: text documents go in PDF/A, scanned images go in uncompressed TIFF or PNG, and tabular data goes in plain-text CSV or TSV. The unifying rule, from the Library of Congress's Recommended Formats Statement 2025–2026, is that files should be non-proprietary, unencrypted, self-contained, and in common use.

What Is PDF/A and Why Does It Win?

PDF/A is an ISO standard — ISO 19005 — that defines a constrained subset of PDF built for the long-term archiving of electronic documents, and the "A" stands for archiving. First published as ISO 19005-1 in September 2005, it was created to solve a concrete problem: standard PDFs contain elements that may not render correctly, or at all, years down the line. The standard's answer is self-containment: a conforming document must carry everything needed to render it identically regardless of the environment or era in which it opens, with no external fonts, no references to external color profiles, no dynamic content, and no encryption that could prevent future access.

The self-containment rules are strict and they are the whole point. Fonts must be embedded (not merely referenced from the author's machine); color must be unambiguous through ICC output intents or fully tagged color management; metadata must use the XMP standard with the conformance level declared. Encryption, JavaScript, audio/video annotations, and external content references are forbidden in every level, because each is a proven failure point — an encrypted document "cannot guarantee future accessibility," and the appearance of a file must never depend on script execution.

This is also why plain PDF loses the comparison. A standard PDF may reference fonts installed on the author's system that no longer exist a decade later, use proprietary color spaces that become undefined, and rely on JavaScript whose API has changed. Those are exactly the failure modes PDF/A was designed to eliminate. The institutional backing follows from that engineering: the US National Archives accepts PDF/A for permanent-record transfers, US federal courts reference it in CM/ECF filing standards, and EU eIDAS technical standards reference PDF/A for electronic signatures on archived documents.

PDF/A vs PDF: What's the Real Difference?

The difference between PDF/A and regular PDF is that PDF/A trades away the risky conveniences of PDF — encryption, scripting, external references — for a self-contained file that renders the same decades from now. Here is the side-by-side:

The trade-off is genuine. PDF/A files are typically larger than regular PDFs because they embed everything, and you cannot password-protect a PDF/A file. Those are the costs of future-proofing — and for documents you need readable in twenty years, they are worth paying.

Which PDF/A Version Should You Choose? (PDF/A-1 vs PDF/A-2 vs PDF/A-4)

For most archiving today, PDF/A-2b is the widely recommended default — it is the most compatible variant, a sensible default for general business records (SBTEXMedia), and the level LibreOffice's own help recommends for most users. Four parts of the standard exist, and the choice between them is simpler than the version numbers suggest.

PDF/A-1 (ISO 19005-1:2005, based on PDF 1.4) was the first edition; its PDF/A-1b level — reliable visual reproduction only — remains the minimum many courts and archives accept. PDF/A-2 (ISO 19005-2:2011, based on PDF 1.7) added JPEG 2000 compression, transparency, layers, and conforming file attachments. PDF/A-3 (ISO 19005-3:2012) is identical to PDF/A-2 except that attachments may be any file type — useful for hybrid formats like ZUGFeRD invoices with embedded XML, but weaker for pure archiving because the attachments are not themselves archival.

PDF/A-4 (ISO 19005-4:2020, based on PDF 2.0, published November 2020) is the current edition, and it simplifies the old A/B/U conformance levels into three variants: the base level (Unicode mappings now required by default), PDF/A-4f for arbitrary embedded files, and PDF/A-4e for engineering documents with 3D content. As of November 2026, though, adoption is still early: the spec was released in 2020 but real-world PDF/A-4 documents remain scarce (GdPicture), and free tools lag — LibreOffice 26.x still exports only PDF/A-1b, 2b, and 3b. For a file you create today, PDF/A-2b is the safest practical pick.

What Are the Conformance Levels (a, b, u) and Which Do You Need?

The conformance levels inside PDF/A parts 1–3 describe how much of the document's structure is preserved: level B is visual reproduction only, level U adds searchable Unicode text, and level A adds full accessibility tagging — and for ordinary archiving, B is enough. Level B (Basic) is the minimum and the most commonly implemented: it guarantees the document looks right. Level U (Unicode) adds Unicode mappings for all text so content is searchable and extractable, available in parts 2 and 3. Level A (Accessible) requires tagged logical structure plus a declared document language, which is significantly more demanding to produce and usually only needed where accessibility compliance is the goal.

PDF/A-4 removes the A/B/U distinction entirely: all PDF/A-4 documents require Unicode mappings by default, so the old B-versus-U decision disappears. If you are choosing between PDF/A-2b and PDF/A-2u today, pick 2u when searchability matters and the file will be converted by software that handles the Unicode mapping properly; otherwise 2b keeps things simple and maximally compatible.

What Is the Best Format for Scanned Documents?

For scanned documents, the master scan should be uncompressed TIFF, with PDF/A as the container for the finished, searchable document. The Library of Congress's requirements, documented in Cornell's preservation tutorial, specify 300 dpi 1-bit TIFF with ITU-T Group 4 compression for printed text, and uncompressed TIFF for pictorial materials and manuscripts. The University of Washington Libraries' file-format guidance rates TIFF uncompressed, PDF/A, and PNG as high-confidence formats — defined as non-proprietary, open-source, and uncompressed or losslessly compressed.

JPEG is lossy — it discards detail to keep files small — which makes it fine for sharing copies but wrong for masters, since data taken away can never be put back. JPEG2000 in lossless mode is acceptable but less universally supported than plain TIFF. Keep the TIFF masters in cold storage, and if you need shareable copies from them, a TIFF to PDF converter builds the working copy while the master stays archived.

What Is the Best Format for Long-Term Data Storage?

The best format for long-term data storage is plain text: UTF-8 text files for prose, and CSV or TSV for tables. Plain text in UTF-8 is, as one 2026 archival-formats guide puts it, "the most durable format there is — no fonts, no layout and nothing to break." The Library of Congress Recommended Formats Statement 2025–2026 prefers platform-independent, character-based formats over native or binary ones for datasets, explicitly including line-oriented CSV, TSV, and fixed-width files, with UTF-8 as the top character encoding.

Two habits make this practical. First, save a CSV copy of spreadsheet data alongside the original — CSV loses formatting but keeps every value readable by any program, past or future. If your numbers still live in a spreadsheet, an Excel to CSV converter exports the values to the archival copy in seconds. Second, when you are unsure about any format's risk, check it against PRONOM, the UK National Archives' online database covering more than 2,000 file formats with persistent identifiers — its rebuilt site launched in October 2026, and its data feeds the free DROID identification tool.

For both documents and datasets, the Library of Congress adds two hard rules: files must contain no DRM or encryption, and they should be self-contained so they remain useful if external services change. A spreadsheet on a cloud service you stop paying for fails both tests.

Which Formats Should You Avoid for Archiving?

Avoid formats that depend on specific software, encryption, or external services — the self-containment test eliminates them one by one. Proprietary word-processing formats like DOC and DOCX carry no embedded-font guarantee, so a document authored today can reflow or misrender under different software tomorrow; PDF/A exists precisely to fix that. Encrypted and password-protected files fail the future-accessibility test — PDF/A forbids encryption in every level for this reason, and the Library of Congress's 2025–2026 statement requires datasets to have no DRM or encryption at all.

Lossy image formats like JPEG are wrong for document masters because compression permanently discards detail. Formats with external dependencies — documents that pull in web fonts, linked stylesheets, or referenced media — break when those resources vanish, which is why PDF/A prohibits external content references. And anything tied to a subscription or an account — a cloud-only document, a file in a proprietary backup format — is not archived; it is rented.

How Do You Actually Convert Documents to PDF/A?

Converting to PDF/A takes five steps: prepare the source, pick PDF/A-2b, export, validate, and store with a plain-text checksum record.

1. Prepare the source. Flatten anything dynamic: accept tracked changes, remove password protection (PDF/A cannot be encrypted), and confirm fonts are embeddable. For Word users, a TU Munich 2026 library guide documents the Microsoft path — Word's Options include a "PDF/A compatible" checkbox that its library expects to produce at least PDF/A-1b.
2. Export as PDF/A-2b. In LibreOffice, use File → Export As → Export as PDF → General tab → check "Archive (PDF/A, ISO 19005)" — LibreOffice's own help recommends PDF/A-2b for most users. In Adobe Acrobat Pro, the Preflight panel offers built-in "Convert to PDF/A-2b" profiles that embed fonts, convert color spaces, strip prohibited features, and add XMP metadata.
3. Validate the result. Run the exported file through veraPDF, the purpose-built open-source PDF/A validator maintained by the Open Preservation Foundation and the PDF Association — it covers every PDF/A part and produces detailed standards-referenced reports. A quick structural check with a PDF validator catches damaged files before you archive them.
4. Record what you did. Store a small plain-text note alongside the file: source filename, converter and version, the PDF/A level chosen, and the validation result. Future you — or the archive — will need this.
5. Keep two copies in two places. Format choice is only half of preservation; the other half is redundancy. One copy on local media and one copy offsite turns a good format into a good archive.

Sources

Every factual claim in this guide comes from a source fetched in November 2026.

• Mapsoft, "PDF/A: The Archival PDF Standard" (published 2026-03-17, updated 2026-03-24) — https://mapsoft.com/posts/pdf-pdfa-standard.html — ISO 19005 parts, self-containment rules, conformance levels, PDF/A-4 details, prohibited features, institutional mandates.
• Mapsoft, "PDF ISO Standards Explained" (2026-05-30) — https://mapsoft.com/posts/pdf-iso-standards.html — PDF/A-4 (ISO 19005-4:2020) as the current edition.
• Wikipedia, "PDF/A" (updated ~October 2026) — https://en.wikipedia.org/wiki/PDF/A — part publication dates: 19005-1:2005, 19005-2:2011, 19005-3:2012, 19005-4:2020.
• Nutrient, "What Is PDF/A?" — https://www.nutrient.io/blog/what-is-pdf-a/ — PDF/A-2b as the most compatible, widely recommended variant for scanned documents.
• SBTEXMedia, "PDF/A Document Archiving" — https://www.sbtexmedia.com/blog/pdf-a-document-archiving.html — PDF/A-2b as a sensible default for general business records.
• LibreOffice Help, "PDF Export General" (26.2) — https://help.libreoffice.org/26.2/hsb/text/shared/01/ref_pdf_export_general.html — "PDF/A-2b is recommended for most users"; 26.x GUI offers only A-1b/2b/3b.
• verapdf.org (home + software pages) — https://verapdf.org/home/ — open-source PDF/A validator by the OPF and PDF Association consortium; PDF/A-4 support since v1.20; current line 1.26 with 1.30 release candidates (Feb 2026).
• Library of Congress, Recommended Formats Statement 2025–2026 — https://www.loc.gov/preservation/resources/rfs/ — PDF/A preferred for textual works; character-based CSV/TSV preferred for datasets; UTF-8 first; no DRM/encryption; self-contained files.
• Cornell University Library, "Representative institutional requirements for conversion" — http://preservationtutorial.library.cornell.edu/tutorial/conversion/table3-1.html — LoC scan requirements: 300 dpi 1-bit TIFF G4 for text; uncompressed TIFF for pictorial/manuscript materials.
• Recollection Wisconsin, "File Formats" (citing University of Washington Libraries Preferred File Formats) — https://recollectionwisconsin.org/the-toolkit-file-formats — high/medium/low confidence formats; TIFF uncompressed, PDF/A, PNG high-confidence.
• MIT Libraries, "Store: File Formats" — https://libraries.mit.edu/data-management/store/formats/ — plain text (UTF-8), CSV/TSV for tabular data; non-proprietary, software-neutral, unencrypted.
• PRONOM, National Archives (UK), "About" — https://pronom.nationalarchives.gov.uk/about — 2,000+ format records, PUIDs, feeds DROID; new site launched October 2026.
• Henëdo, "Archival File Formats" (2026) — https://www.henedo.com/blog/archival-file-formats/ — LoC sustainability factors; plain-text UTF-8 durability; CSV alongside spreadsheets.
• IronPDF, "PDF/A Archiving" tutorial (updated 2026-08-02) — https://ironpdf.com/tutorials/pdfa-archiving-csharp/ — NARA acceptance; US courts CM/ECF; European Commission mandates (vendor claims, attributed).
• TU Munich University Library, "PDF/A Guide" (edited January 2026) — https://mediatum.ub.tum.de/doc/1840064/1840064.pdf — Word "PDF/A compatible" path.

Frequently Asked Questions

What is the difference between PDF/A-1 and PDF/A-2?

PDF/A-1 (ISO 19005-1:2005, based on PDF 1.4) was the first edition; PDF/A-2 (ISO 19005-2:2011, based on PDF 1.7) extends it with JPEG 2000 compression, transparency, layers, and conforming file attachments. For new archiving, PDF/A-2b is the widely recommended default.

Can a PDF/A file be encrypted or password-protected?

No. Encryption is forbidden in every PDF/A conformance level because an encrypted file cannot guarantee future accessibility. If a document must be protected, secure the storage location or the archive system instead of the file.

Is PDF/A better than TIFF for scanned documents?

They serve different roles: keep the scan master as uncompressed TIFF, and use PDF/A-2b as the container for the finished, searchable document. The Library of Congress specifies 300 dpi 1-bit TIFF for printed-text scans, and PDF/A is a high-confidence format for the combined document.

How do I check whether my file is really PDF/A?

Validate it with veraPDF, the open-source PDF/A validator maintained by the Open Preservation Foundation and the PDF Association — it checks every PDF/A part and reports exactly which rule a file violates. Always validate after converting.

Should I use PDF/A-4 yet?

PDF/A-4 (ISO 19005-4:2020) is the current edition of the standard, but real-world adoption is still early and free tools like LibreOffice do not yet export it. Stick with PDF/A-2b for maximum compatibility today, and revisit PDF/A-4 as tooling matures.