A document that still opens in twenty years, and that everybody can read

Two things are asked of a serious document nowadays. That it can still be opened and read a very long time from now, by an auditor, a public body or a court. And that somebody who cannot see the page can still be told what is on it. Both have written rules, both are checked by a program that has never heard of us, and our documents pass.

Why an ordinary PDF may not reopen in fifteen years

A PDF does not have to carry everything it needs. It is allowed to lean on the machine that opens it: on a typeface installed there, on a picture fetched from somewhere else, on a small program tucked inside it, on a password held by somebody who has since left. The day one of those props is gone, the page comes back wrong, or does not come back at all.

That is not a fault in one program or another. The format allows it, which is exactly why a separate set of rules had to be written to forbid it.

In practice, this is what goes wrong:

• A typeface that was never put into the file, and is no longer on any machine: the text is redrawn in something else, and the whole layout shifts.
• A colour described as whatever this particular screen calls red, rather than as a measurement: the document comes out a different colour.
• A picture or a typeface fetched from an address that no longer answers.
• A password nobody has any more, on a file nobody can open any more.
• Little programs inside the document that today's reader refuses to run.

What an archival PDF is

There is a family of rules written for exactly this. It is called PDF/A, and it is one standard in four parts, published in 2005, 2011, 2012 and 2020, each built on a later version of the PDF format. They all say the same thing: everything the document needs is inside the document. The typefaces travel in it. The colours are stated as measurements rather than left to the machine. A block of information says what the file is and which rules it claims to follow. Passwords, outside links and little programs are simply not allowed.

Each part comes in levels. The plain level promises the page will always look the same. The higher level asks for more: the document must also carry the structure of its own text — which is precisely what a reading aid needs, so the two subjects on this page meet there.

National archives ask for these by name. The United States National Archives put PDF/A-1 and PDF/A-2 among the formats they prefer for text that was born digital, and PDF/A-4f and PDF/A-3 for collections of PDFs that carry attachments. Their table of formats was last brought up to date in August 2025.

What this engine writes is the third part, PDF/A-3, at both of its levels. Not the first, not the second, not the fourth. If somebody asks you for one of those three by name, this is not yet the tool for it, and we would rather you heard that here than found it out later.

What an accessible PDF is, and who needs one

Somebody who cannot see the page listens to it instead. The program reading it aloud has to know what is a heading, what is a paragraph, what is a table, which cell belongs to which column, and above all in what order to say it all. A PDF that only records where the ink went says none of that, so the reading aid works through the page in the order the marks were drawn — which is rarely the order a person would read them.

The rules for this are called PDF/UA. The edition of its first part that counts today was published in December 2014, and a second part followed in March 2024. They ask that every real piece of content be marked up and every piece of decoration be marked as decoration; that the reading order make sense; that the document state its own title and its own language; that a picture carry a description in words; and that nothing be said by colour alone.

This is not only a kindness. In Europe it is the law:

• A European directive of 26 October 2016 obliges public bodies to make what they publish usable. Office documents are in it: the only ones it leaves out are those published before 23 September 2018, so everything put online since that day counts — PDFs included.
• The European standard those rules point at, EN 301 549, gives a whole chapter — its clause 10 — to documents that are not web pages, which means the PDF a visitor downloads.
• In France a decree of 24 July 2019 set the fine at 20 000 euros, or 2 000 euros for a commune of fewer than five thousand people, and it can be handed down again every year the failing lasts. Since an order of 6 September 2023 it is the regulator Arcom that fines, up to 50 000 euros, and it may fine again six months later if nothing has changed.
• And a further European law reaches past public bodies — into banking, transport, selling online and digital books — for anything put on the market or supplied after 28 June 2025.

Who decides whether a document really complies

Not the company that wrote it. There is a free program called veraPDF whose whole job is to open a file and check it, rule by rule, against the published texts. It is run by the Open Preservation Foundation and the PDF Association, two bodies whose business is keeping documents readable, and it was built with European Commission money under a programme that ended in 2017. Anybody may read its workings. It knows nothing about who wrote the file in front of it, and it answers pass or fail.

It checks all four parts of the archival rules and both parts of the accessibility ones, which is rather more than this engine claims. That is exactly why it is worth quoting: the checker was not built around what we can do.

What our documents pass

The archival examples published on this site are put through veraPDF, and it accepts them:

• as PDF/A-3 at the plain level, so the page will look the same for as long as anybody keeps it;
• as PDF/A-3 at the higher level, so the structure of the text is in there as well;
• as PDF/UA-1, so a reading aid can work through them properly.

And here is what they are not. The engine writes the third part of the archival rules and the first part of the accessibility ones. It does not write PDF/A-1, PDF/A-2 or PDF/A-4, and it does not write PDF/UA-2. The checker would examine all of those; we claim none of them.

Almost none of it is done by hand. You say the document is an archival one, and the engine sees to what that entails: the typefaces go in — only the letters actually used — the colours are described, the block of information is written, each paragraph is marked up as it is laid out, and a table builds its own structure, headings and all. What stays with you is the outline: which part of the document is a section, and in what order it should be read.

See a document that carries its own structure See an archival form

Where to start

If you answer calls for tender from public bodies, or if you keep documents for ten years, this is the difference between a file that is accepted and one that comes back. It costs you nothing extra: it is the same engine, and it is on in every offer.

See the prices See the examples