PDF/A-1b vs 2b vs 3b: Which Conformance Level Do You Need?

Pick the part your recipient named. If nobody named one, pick PDF/A-1b.

That answers most of the question, and the rest of this explains when the shortcut breaks. The parts of the standard are not quality tiers, which is the misconception that causes the trouble. They differ in what they allow, and a later part is not a better part — sending PDF/A-3 to a registry that accepts only PDF/A-1 gets you rejected just as firmly as sending something invalid.

PDF/A conformance levels compared, side by side

Notice how much stays constant. Fonts embedded, colour defined, nothing executable, nothing encrypted — those rules are identical in every part, and they are the ones durability actually depends on. What changes is tolerance for modern PDF features.

Which PDF/A version to use, by situation

The requirement names a part. Use exactly that one. This is the single most common reason a valid file gets rejected: it is perfectly good PDF/A-2b arriving at a system that only takes PDF/A-1b.

The requirement just says PDF/A. Use 1b. It is the strictest, so it satisfies the widest range of systems.

Your document is design-heavy and nobody specified. Here the shortcut breaks down. PDF/A-1 forbids transparency, so drop shadows and semi-transparent logos get flattened. Sometimes that is invisible. Sometimes you get hard edges where a gradient used to fade out. Convert, look at the result, and if it has degraded, ask whether 2b is acceptable rather than submitting something that looks wrong.

You need reliable text extraction. Look at level u rather than the part number. Level u is level b plus a guarantee that all text maps to Unicode, so it can be searched and copied dependably. It exists in parts 2 and 3 only.

What PDF/A-3 embedded attachments are really for

Part 3 differs from part 2 in exactly one respect: it lets the document carry any file type as an attachment. Part 2 allows attachments too, but only if they are themselves PDF/A.

That sounds minor and is not, because it enables the hybrid electronic invoice. Formats such as ZUGFeRD in Germany and Factur-X in France put a human-readable invoice on the page and a machine-readable XML file inside the same document. Accounting software reads the XML; a person reads the page; both are guaranteed to be the same invoice.

Two caveats we would flag before anyone gets excited. First, this is a genuinely specialist use case — if nobody has mentioned e-invoicing to you, part 3 is not what you need. Second, our own converter does not generate hybrid e-invoices. It can produce PDF/A-3b, but building a compliant invoice means embedding the XML with the right relationship key and filename, which is a job for dedicated invoicing software. We would rather say so than let you discover it at the tax office.

PDF/A level a vs level b, and why almost nobody has a

The letter after the number is a separate axis from the part, and it trips people up more than the parts do.

Level b means basic. It guarantees the document will look right — fonts embedded, colour defined, appearance preserved. This is what virtually every file described as PDF/A actually is.

Level a adds accessibility: a tagged structure tree, a defined reading order, alternative text on images, and text that maps to Unicode. A screen reader can navigate it properly.

Here is the part that matters practically. Level a cannot be manufactured by a converter. The structure has to come from the program that wrote the document, because a finished PDF simply does not contain the information — nothing in the file records that a particular line was a heading rather than large bold text. Anything promising to convert an arbitrary PDF to level a is either doing a poor job of guessing or misusing the label.

So if you are asked for PDF/A-1a, the work happens in your authoring tool. Use proper heading styles in Word, add alternative text to every image, then export directly.

Checking what you actually produced

Converters are not always honest about their output, and the failure is usually silent rather than loud. The file converts, reports success, and turns out to have no embedded colour profile at all.

Run the result through our PDF/A validator before you submit it. It reads the declared part and conformance level straight out of the file metadata and checks the structural requirements in your browser, which takes a couple of seconds and saves a rejection cycle.

If it comes back short on fonts, that is the most common failure of all and has its own guide in the fonts not embedded error.

Getting the level right the first time

The decision is smaller than it looks. Read the requirement; if it names a part, use that part. If it does not, use 1b and check the output has not been visibly flattened. Reach for part 3 only when something specifically needs attachments.

When you know which one you want, our PDF/A converter lets you choose the level rather than guessing on your behalf. New to the format entirely? Start with what PDF/A actually is. Already have an archived file you need to change? That is covered in editing a PDF/A file.