The difference in one sentence
PDF is a format for displaying documents; PDF/A is a profile of that format for keeping them. The Library of Congress, whose digital preservation programme documents both, calls PDF/A-1 a constrained form of PDF 1.4 (ISO 19005-1), created because large volumes of official records are kept as PDF and ordinary PDF is not well suited to archiving: too much of what a file needs to render can live outside the file.
What PDF/A forbids and requires
| Feature | Ordinary PDF | PDF/A | Why |
|---|---|---|---|
| Fonts | May reference fonts on the reader's system | Every font must be embedded, and licensed for unlimited rendering | A missing font changes the page |
| Encryption and passwords | Allowed | Not allowed | A lost key is a lost document |
| JavaScript, launching files | Allowed | Prohibited | Scripts depend on a viewer that may not exist |
| Audio and video | Allowed | Forbidden | Depends on codecs outside the file |
| External references | Allowed | Avoided | The linked file may be gone |
| Colour | May be device-dependent | Must be specified device-independently | The same colour on any screen or press |
| Metadata | Optional | XMP required for basic descriptive and identifying data | An archive has to know what the file is |
Nothing in the list changes how the page looks. A PDF/A file opens in any PDF viewer; the constraints only remove the ways a file could fail later. That is why PDF/A is described as a final-state format for delivery to end users, not a working format: you make it at the end, from the finished document.
PDF/A-1, 2 and 3
| Part | Standard | Based on | What it adds |
|---|---|---|---|
| PDF/A-1 | ISO 19005-1 | PDF 1.4 | The baseline: the rules in the table above |
| PDF/A-2 | ISO 19005-2 | PDF 1.7 (ISO 32000-1) | JPEG 2000 compression, transparency for graphics, embedding of other PDF/A files, portable collections, and the new level U |
| PDF/A-3 | ISO 19005-3 | PDF 1.7 | One feature: files in any format may be embedded, not only PDF/A files |
The Library of Congress is careful about PDF/A-3. Embedded files of arbitrary types are hard to characterise and manage for long-term preservation, there is no simple way to tell from the outside that a file contains them, and a 2014 preservation report it cites recommends PDF/A-2 over PDF/A-3 whenever the document has no embedded files or only PDF/A ones. In practice PDF/A-3 is used where the attachment is the point, such as an invoice PDF that carries its machine-readable data file; for a plain document, PDF/A-2 is the safer choice.
Levels a, b and u
- Level B (basic) guarantees the visual appearance: what you see is reproducible. It says nothing about structure.
- Level U (Unicode), from PDF/A-2 on, is level B plus the rule that every piece of text has a Unicode equivalent, so the text can be searched, copied and indexed reliably.
- Level A (accessible) meets all requirements, including tagged structure that records reading order, headings and alternative text for images, which is what screen readers and reflow need.
Level A is the best and the hardest, because the tagging has to come from the authoring application; the Library of Congress notes it is difficult to add after creation and that level B is usually acceptable when PDF/A is mandated. A scanned document can reach level B (as an image) or U (with an OCR text layer), but never meaningfully level A. For most submissions, PDF/A-2b or PDF/A-2u is what is meant.
When you will be asked for it
When a court, a records office, a university or a tender portal asks for PDF/A, it is for the reason the Library of Congress gives: they are keeping the file, not viewing it once. The requirement usually names the part and level, such as PDF/A-1b or PDF/A-2u, and a validator on the receiving end checks the file against it. The checks that most often fail are fonts that were not embedded, a password, and colour defined for a specific device.
Making and handling PDF/A files
- Export from the source document with the PDF/A option in the word processor or layout program, rather than converting a finished PDF, so fonts and tags come from the original.
- Merge before archiving. Combining archive-ready files is fine when the tool keeps fonts and metadata intact; the merge PDF tool joins pages in the browser, and the result still needs to be exported or validated as PDF/A if the recipient requires it.
- Compress the images, not the structure. The compress PDF tool shrinks the images inside a file; PDF/A does not limit image compression, only where the file's dependencies live.
- Scans made with the JPG to PDF tool are image-only PDFs, the raw material for a level B archive file; adding an OCR layer is what lifts them to level U.
- Do not password-protect an archive copy. The protect PDF tool is for files you send, not files you file; a PDF/A document cannot be encrypted.
Frequently asked questions
What is PDF/A?
A constrained form of PDF for long-term preservation, standardised as ISO 19005. The Library of Congress describes PDF/A-1 as a constrained form of PDF 1.4 that forbids audio, video, JavaScript and encryption, requires every font to be embedded, requires device-independent colour and XMP metadata, and avoids external dependencies, so the file renders the same decades later without anything outside it.
What is the difference between PDF/A-1, PDF/A-2 and PDF/A-3?
The base PDF version and what may be embedded. PDF/A-1 is built on PDF 1.4. PDF/A-2 moves to PDF 1.7 and adds JPEG 2000 compression, transparency and the embedding of other PDF/A files. PDF/A-3 adds one thing: files in any format may be embedded, which the Library of Congress notes is a mixed blessing for preservation.
What do the letters a, b and u mean?
Conformance levels. Level A meets every requirement of the standard, including tagged structure so the text order and headings are machine-readable. Level B covers the visual appearance only. Level U, added in PDF/A-2, is level B plus the requirement that all text has Unicode equivalents, so it can be searched and copied. The Library of Congress notes that level B is usually acceptable when PDF/A is mandated, because the tagging that level A needs is hard to add after creation.
Can a PDF/A file be password protected?
No. Encryption is one of the things PDF/A-1 forbids, because an archive must be able to open the file without a key that may be lost. If a document must be both archived and restricted, the restriction has to live in the archive system, not in the file.
Is PDF/A smaller or larger than a normal PDF?
Usually a little larger, because every font used must be embedded and images must be in a device-independent colour space. A compressor can still shrink the images inside it; the rule is only that nothing the file needs may live outside it.
Sources
- Library of Congress, Sustainability of Digital Formats: PDF/A-1, ISO 19005-1 (a constrained form of PDF 1.4; forbidden and required features; conformance levels a and b)
- Library of Congress: PDF/A-2, ISO 19005-2 (based on PDF 1.7; JPEG 2000, transparency, PDF/A attachments; levels a, b and u)
- Library of Congress: PDF/A-3, ISO 19005-3 (embedding of files in any format; the preservation caveats)