Convert a PDF to PDF/A for Long-Term Archiving

Converting PDF to PDF/A rewrites a document to conform to ISO 19005, the archival profile of PDF: every font embedded, colors defined by an embedded ICC profile, XMP metadata identifying the conformance level, and no encryption, JavaScript, or external dependencies — so the file renders identically decades from now on software that has never seen your computer. Courts, national archives, and regulators require it for exactly that reason. This page explains the A-1, A-2, and A-3 levels and gives working LibreOffice and Ghostscript recipes, validated with veraPDF.

Published · Last updated

Why there is no one-click converter on this page

PDF/A conversion has to re-embed fonts your system provides, convert color spaces against ICC profiles, and rewrite the file to pass a formal validator — work that needs a full PDF rendering engine and font access that browsers do not expose to pages. A converter that cannot be validated would defeat the entire point of an archival format, so the desktop recipes below are the trustworthy route.

Disclosure: this page contains affiliate placeholder links. If they were live, we could earn a commission at no extra cost to you.

AFFILIATE PLACEHOLDER: Recommended desktop PDF suite (placeholder link)

How to convert a PDF to PDF/A (step by step)

Method 1: Ghostscript (free, works on existing PDFs)

  1. Install Ghostscript from ghostscript.com or your package manager. On Windows the command is gswin64c; on Linux and macOS it is gs.
  2. Run the PDF/A conversion. gs -dPDFA=2 -dBATCH -dNOPAUSE -sColorConversionStrategy=UseDeviceIndependentColor -sDEVICE=pdfwrite -dPDFACompatibilityPolicy=1 -sOutputFile=archive.pdf input.pdf. Ghostscript re-renders the file, embeds fonts, and writes PDF/A-2 conforming output. (-dPDFA=1 or -dPDFA=3 target the other levels.)
  3. Validate with veraPDF. Download the free veraPDF validator, open archive.pdf with the PDF/A-2b profile, and run the check. Fix reported failures or adjust the conversion flags — a file is only PDF/A when a validator says so, not when a converter claims so.
  4. Archive both files. Store the validated PDF/A alongside the original; if conversion changed any rendering detail, the original remains your reference copy.

Method 2: LibreOffice (when you still have the source document)

If the PDF came from a document you can still open (Writer, Calc, Word formats), export fresh instead of converting: File → Export as PDF → General tab → Archival (PDF/A) , choosing PDF/A-1b, 2b, or 3b from the version dropdown. Exporting from the source is strictly better than converting an existing PDF, because LibreOffice embeds the fonts and writes the metadata correctly at creation time rather than patching afterwards.

Method 3: Acrobat Pro's Preflight (when you need conformance level A or fixups)

Acrobat Pro's Preflight panel has "Convert to PDF/A" fixup profiles for each level and its own validator, and it is the practical choice when a court or registry demands level A (tagged) conformance, since it can add basic structure tags during conversion. Verify with veraPDF anyway; the two validators occasionally disagree, and veraPDF is the reference implementation.

How it works

PDF/A is a subset standard: it does not add features to PDF, it forbids the ones that make files device-dependent or mortal. The requirements cluster into four groups. First, self-containment — every font used must be embedded (with at least the glyph subset actually referenced), so rendering never depends on fonts installed in 2049. Second, deterministic color: device-dependent color spaces like plain DeviceRGB are only allowed if an embedded ICC profile (an OutputIntent) defines what the numbers mean. Third, prohibitions: no encryption of any kind, no JavaScript or launch actions, no audio/video, no external references, and no LZW compression — anything whose behavior depends on software or resources outside the file. Fourth, identification: an XMP metadata packet must declare the standard part and conformance level (for example pdfaid:part=2, pdfaid:conformance=B), and all document metadata must live in XMP, not just the legacy Info dictionary.

The three parts of the standard track PDF's own evolution. PDF/A-1 (2005, based on PDF 1.4) predates transparency support, so a modern file with soft shadows cannot conform without flattening. PDF/A-2 (2011, based on ISO 32000-1) permits transparency, JPEG 2000, and object streams, and lets you embed other PDF/A files. PDF/A-3 changes exactly one thing: arbitrary file attachments may be embedded — the mechanism German ZUGFeRD e-invoices use to carry machine-readable XML inside a human-readable PDF. Within each part, conformance level "b" (basic) guarantees visual reproduction, "u" adds Unicode mapping for all text, and "a" (accessible) requires tagged logical structure.

A converter like Ghostscript's pdfwrite therefore interprets the whole input, then re-emits it: fonts get embedded or substituted, colors get bound to the OutputIntent profile, forbidden objects are dropped or rasterized per the compatibility policy, and the XMP claim is written. Validation is the essential final step because the claim is just metadata — veraPDF checks the file against every clause of the standard and reports each violation individually.

Worked example: archiving a signed 31-page contract

A notary's office needed a 31-page, 1.4 MB scanned-plus-typed contract stored as PDF/A-2b for a 30-year retention requirement. veraPDF flagged the original with 3 rule violations: two unembedded fonts (Arial and Times New Roman, referenced but not included) and a missing OutputIntent. The Ghostscript command above ran in 11 seconds and produced a 2.1 MB file — 50% larger, the growth being 940 KB of newly embedded font programs and a 3 KB sRGB ICC profile. veraPDF then passed it: 0 failed checks against the PDF/A-2b profile, 8,244 checks total. A second attempt targeting PDF/A-1b failed with 62 violations because the scanner's pages used transparency groups, illegal in A-1 — a tidy demonstration of why A-2 is the right default target for modern files.

Frequently asked questions

What is the difference between PDF/A-1, PDF/A-2, and PDF/A-3?

PDF/A-1 (2005) is the strictest and oldest, banning transparency and JPEG 2000. PDF/A-2 (2011) allows those plus embedding other PDF/A files, and PDF/A-3 is identical to A-2 except it may embed attachments of any file type. For most new archiving, PDF/A-2b is the sensible default.

What do the conformance letters a, b, and u mean in PDF/A?

Level b guarantees only reliable visual reproduction; level u adds that all text must map to Unicode; level a additionally requires tagged logical structure for accessibility. Converting an existing untagged PDF can realistically reach b or u, but not a, because tagging cannot be invented afterwards.

Why is my PDF/A file bigger than the original PDF?

PDF/A requires every font to be embedded and a color profile to be included, which adds data that ordinary PDFs often omit. A file that previously relied on system fonts can grow by several hundred kilobytes.

Does converting to PDF/A remove passwords and JavaScript?

It must — PDF/A forbids encryption, JavaScript, embedded multimedia, and external content references, because an archive file has to be self-contained and openable forever. Decrypt the file first; converters cannot process a password-protected input.

How do I check that a PDF really is valid PDF/A?

Run it through veraPDF, the free open-source validator developed with the PDF Association, selecting the profile that matches the claimed level. The blue PDF/A banner some viewers show only reflects the file's metadata claim, not actual conformance.

Is a scanned document PDF a valid basis for PDF/A?

Yes — image-only PDFs convert to PDF/A-2b cleanly since there are no fonts to embed, and this is how national archives store scans. Adding an OCR text layer before conversion makes the archive searchable too.

Related tools