Any PDF opens without trouble today, but there is no guarantee that fifteen years from now it will display the same way. It may depend on fonts installed on the computer, or contain links to external resources, scripts or multimedia content that future readers may not handle. PDF/A was created to remove these uncertainties: it is a version of PDF, defined by the ISO 19005 standard, designed so that the document stays self-contained and viewable over time.
What changes compared with an ordinary PDF
PDF/A does not add features: it removes them. A compliant file must follow a few rules:
- All fonts embedded. The document cannot rely on the fonts present on the computer of whoever opens it.
- Device-independent colours, defined through a colour profile included in the file.
- No encryption and no opening password, because the key may not be available in the future.
- No JavaScript, audio or video, and no content that depends on external resources.
- Metadata in XMP format, meaning readable descriptions of the document (title, author, date) inside the file.
The result is a slightly heavier PDF, because it carries its fonts with it, but one that contains everything needed to be displayed correctly.
The parts
The standard has several parts, each based on a different version of the PDF format:
| Part | Main characteristics |
|---|---|
| PDF/A-1 | The first; most restrictive, no transparency and no layers |
| PDF/A-2 | Allows transparency, layers, JPEG 2000 compression and attachments as long as they are PDF/A |
| PDF/A-3 | Like part 2, but allows attachments in any format |
| PDF/A-4 | Based on PDF 2.0, with simplified conformance levels |
PDF/A-3 is used, for example, when you want the readable PDF to include a structured data file as well, as happens in some electronic invoice formats. Be careful, though: the attachment itself is not guaranteed to be preservable, only the PDF that contains it.
Conformance levels
In parts 1, 2 and 3, letters indicate the level:
- b (basic): guarantees that the visual appearance can be reproduced. It is the minimum level and the most common.
- u (unicode): in addition, the text can be extracted correctly as Unicode, so it is searchable and copyable.
- a (accessible): in addition, the document is structured with tags, so it can also be read by assistive technologies.
If the document comes from a word processor, aiming for at least level u is reasonable. Level a requires the source document to be well built, with real headings and alternative text for images.
How to produce it
LibreOffice Writer. From File › Export As › Export as PDF, on the General tab you enable the PDF/A archiving option and choose the version. It is worth ticking the tagged PDF option as well.
Microsoft Word. In the Save As dialog, choosing PDF as the format, the Options button contains the checkbox for ISO 19005 (PDF/A) compliance.
From other formats. Tools such as Ghostscript or professional PDF software convert existing files to PDF/A. Converting an existing PDF, however, cannot always recover missing information, such as the structure needed for level a.
Scans. Many scanning programs save directly to PDF/A; if they also apply text recognition, the file becomes searchable.
Checking conformance
A program claiming to have produced a PDF/A does not mean the file really is one. To check, there is veraPDF, an open-source validator that analyses the file and reports every rule violated. It is the reference tool when an archive or an institution requires compliant files.
The most frequent errors are fonts not embedded (often in images or charts pasted from other applications), transparency in a file declared as PDF/A-1, and inconsistent metadata between the document information and the XMP block.
When you really need it
PDF/A is the right choice for documents that must be kept for a long time and not modified: signed contracts, minutes, delivered reports, closed project documentation, correspondence archives. Many digital preservation systems list it among their preferred formats.
It is not needed, on the other hand, for working documents still in progress, for interactive forms or for files that must contain video. In my work I always keep two versions: the editable source and, on delivery, the PDF/A.
In short
- PDF/A is a self-contained PDF: embedded fonts, defined colours, no encryption or scripts.
- Parts 1-4 differ in what they allow; part 3 permits attachments of any type.
- Levels b, u and a guarantee appearance, extractable text and accessible structure respectively.
- Word and LibreOffice produce it directly; conformance should be checked with veraPDF.
- It pays to keep both the editable source and the final PDF/A version.
