What Is PDF/A? The ISO 19005 Standard for Archiving PDFs
Published by PDFToolz · Updated · Technical references listed below
PDF/A is a stricter version of PDF, defined in the ISO 19005 series, for files that must open and look the same decades from now. It bans features that can break over time, such as outside fonts, encryption and scripts. Each file must carry everything it needs to display itself.
This guide covers the four versions (PDF/A-1 to PDF/A-4), the conformance levels a, b and u, the rules a valid file must follow, how to create and validate one, and who has to use it.
What PDF/A is and the problem it solves
PDF/A is a restricted subset of PDF. The first part, ISO 19005-1, came out in 2005. A normal PDF can rely on fonts installed on the reader's computer, link to outside files, be encrypted or run scripts. Any of these can make the file display wrongly, or not at all, in a viewer that doesn't exist yet.
Archives, courts and regulators must keep records readable for decades, long after the software that made them is gone. PDF/A removes the risky features and requires every file to be self-contained. The page can then be drawn the same way in the future without any outside resource.
The four PDF/A versions
New parts don't replace old ones. Each part is still a valid ISO standard, so use the version your recipient asks for.
PDF/A-1 is the strictest and the most widely accepted. PDF/A-2 relaxes several technical limits. PDF/A-3 lets a file carry its own machine-readable source data, which is how European e-invoices such as Factur-X and ZUGFeRD work.
- PDF/A-1 (ISO 19005-1:2005) is based on PDF 1.4. It is the most restrictive, with no transparency, no layers and no JPEG 2000.
- PDF/A-2 (ISO 19005-2:2011) is based on PDF 1.7 (ISO 32000-1). It adds transparency, layers, JPEG 2000, embedded PDF/A files and PAdES digital signatures.
- PDF/A-3 (ISO 19005-3:2012) is also based on PDF 1.7. It matches PDF/A-2 but lets you embed any file type, such as XML, spreadsheets or CAD files.
- PDF/A-4 (ISO 19005-4:2020) is based on PDF 2.0 (ISO 32000-2). It is the newest part, with a simpler conformance model and variants for embedded files and engineering data.
Conformance levels b, u and a
A conformance level says how much a file guarantees. Level b (basic) is the minimum. It guarantees the pages will look right. Level a (accessible) is the strictest. It adds a full tag structure and Unicode text mapping, so the text can be searched, copied, reflowed and read by screen readers.
Level u (Unicode) arrived with PDF/A-2. It requires all text to map to Unicode but doesn't require tags. PDF/A-1 has only levels a and b. PDF/A-2 and PDF/A-3 add level u. PDF/A-4 drops the a, b and u names. It makes Unicode text mandatory for every file and adds PDF/A-4f for embedded files and PDF/A-4e for engineering content.
- Level b guarantees only that the pages look the same.
- Level u adds Unicode text mapping without required tags (PDF/A-2 and later).
- Level a adds tags and Unicode mapping for search, text extraction and accessibility.
Rules every PDF/A file must follow
All levels share a set of hard rules. A file can look like PDF/A and still break them, which is why you validate.
Color trips up many files. If a document uses device color such as DeviceRGB or DeviceCMYK, it must include an output intent with an embedded ICC profile. Future viewers then reproduce the exact colors instead of guessing.
- All fonts are embedded, including the standard 14, and licensed for embedding. Subsets are allowed.
- No encryption or passwords.
- No links to outside fonts, content or files needed to show the page.
- Device-independent color, or an output intent with an ICC profile when device color is used.
- XMP metadata that names the PDF/A part and conformance level.
- No JavaScript, launch actions, audio or video. PDF/A-3 may embed files but not play them.
How to create and validate a PDF/A file
Many apps export PDF/A directly. Word and Acrobat have a Save as PDF/A option, LibreOffice has one in its PDF export dialog, and Ghostscript or a preflight tool can convert an existing PDF. Converters embed fonts and flag what they can't fix. Run OCR on scanned pages first, so the archive holds real Unicode text instead of a flat image. Levels a and u need this, and so does future search.
A converted file isn't always a valid one, so validate before you archive. veraPDF is the open-source reference validator, built by the Open Preservation Foundation and the PDF Association with early funding from the EU PREFORMA project. It checks a file against every clause of the standard and writes a machine-readable report. Acrobat Preflight also reports PDF/A errors. A file labeled PDF/A is only proven valid once a validator passes it.
- Create the file by exporting from your authoring app, or convert it with preflight or Ghostscript.
- Run OCR on scans first so the text is real and mapped to Unicode.
- Validate with veraPDF or Acrobat Preflight and fix each reported error before archiving.
Who needs PDF/A
PDF/A is usually a requirement, not a choice. Groups with legal retention duties or long-term record rules use it, and many ask for it in submissions.
Recipients usually state the exact part and level, such as PDF/A-1b or PDF/A-2u. Check the target before you convert.
- Courts and e-filing systems that require or recommend PDF/A.
- Government agencies that keep official records.
- Libraries, archives and museums with digital collections.
- Companies with records management and regulatory retention rules.
- E-invoicing, where Factur-X and ZUGFeRD are built on PDF/A-3.
Key takeaways
- ✓PDF/A (ISO 19005) makes a PDF self-contained so it displays the same decades later.
- ✓PDF/A-1, 2, 3 and 4 are all valid. Later parts add features and don't replace earlier ones.
- ✓Level b guarantees appearance, level u adds Unicode text and level a adds tags for accessibility.
- ✓A valid file embeds all fonts, has no encryption, defines its color, carries XMP metadata and contains no JavaScript or media.
- ✓Validate with veraPDF or Acrobat Preflight. A PDF/A label on its own proves nothing.
Tools for the job
Frequently asked questions
What is PDF/A and how is it different from a normal PDF?
PDF/A is the ISO 19005 standard for archiving documents as PDF. It is a restricted subset of PDF that bans outside font links, encryption, JavaScript and media, and requires embedded fonts and metadata. Every PDF/A file opens in a normal PDF reader, but most PDFs don't qualify as PDF/A.
What is the difference between PDF/A-1a and PDF/A-1b?
Both guarantee the pages look right. PDF/A-1b stops there. PDF/A-1a also requires tags and Unicode text mapping, so the text can be searched, copied, reflowed and read by screen readers. PDF/A-1a is harder to produce and is usually needed only when accessibility is required.
Can a PDF/A file have a password?
No. The standard bans all encryption. A password or algorithm could be lost over the decades a record must last, which would leave the file unreadable. If you need access control, apply it in the storage system, not inside the file.
How do I check if a PDF is valid PDF/A?
Run it through a validator instead of trusting the label. veraPDF is the free, open-source reference validator from the Open Preservation Foundation and the PDF Association. It tests the file against every clause and lists each error. Acrobat Preflight can also check and fix PDF/A files. Converters don't always produce a valid file, so check before you archive.
Which PDF/A version and level should I use?
Use what the recipient asks for. They usually name an exact target, such as PDF/A-1b or PDF/A-2u. If you can choose, PDF/A-1b is the most widely accepted for plain archiving. PDF/A-2 suits files with transparency or JPEG 2000. PDF/A-3 is for files that must carry their own source data, such as e-invoices.
Does converting to PDF/A change how my document looks?
It shouldn't. The point of the standard is that the page looks the same. Conversion works behind the scenes. It embeds fonts, adds a color output intent and XMP metadata, and removes banned features like JavaScript. Sometimes it flags a problem it can't fix, such as a font that isn't licensed for embedding. That's why you validate the result.
Related terms
Sources and further reading
- ISO 19005 (PDF/A) from PDF Association
- PDF/A basics from PDF Association
Browse all PDF guides, look up a term in the PDF glossary, or head back to the PDFToolz toolkit.