Alligator Design
Notes
Propose a project

PDF: the format that will not bend

Public bodies publish in PDF almost by reflex. The format is good at exactly one thing, and that thing is not the web.

October 2026

You are on a web page from a municipality, a ministry, an agency. Somewhere in the middle of a sentence there is a link, and behind the link is a file: twelve pages, four megabytes, a cover with a logo, a table of contents, and on page nine the two lines you actually came for: the deadline, the amount, the address you have to send it to.

What PDF is genuinely good at

Let us be fair, because the format is not a mistake. It was invented to solve a real problem: a page that looks the same everywhere, on every printer, forever. It is an open standard, it can be signed, it can be archived in a version built for exactly that, and it keeps a layout intact when the layout is the content — a map, a plan, a score, a form somebody must sign and file.

Those are the cases. A record that must be frozen, proved and printed. Nearly everything most administrations publish is none of those things.

When a PDF is the right answer

Five cases, and they have one thing in common: the page itself is the point, not the text on it.

  • It will be signed or relied on as a record. A contract, a certificate, an official extract, a decision. Something whose exact wording and layout must be provable later.
  • It must survive decades. Archiving has its own flavour of the format, built to open long after the software that made it is gone.
  • Paper is the destination. A form to fill in by hand, a poster, a booklet, something handed over at a counter.
  • The layout is the content. A map, a technical drawing, a plan, a score, a scientific paper whose figures and formulas cannot be pulled apart and reflowed.
  • Everyone must be able to cite the same page. A tender, a regulation, a report where “page 14, third paragraph” has to mean the same thing to every reader.

Even then, four conditions:

  1. it is never the only route to the information;
  2. the page that links to it states the essentials in plain text;
  3. it is tagged and made of real text rather than a scan;
  4. it keeps one address that still works next year.

Why it fails as a web page

It cannot bend. A PDF is A4. A phone is not. So the reader gets a shrunken page, pinches, zooms, drags sideways to finish a sentence, and loses their place. More than half the visitors to a public website are on a phone; the format hands them the worst version of the document.

HTML is a bargain: the author gives up control of the layout and gets every reader in return. PDF is the opposite deal.

It is usually unreadable to a screen reader. A PDF only works for blind readers if it has been tagged — headings marked as headings, a reading order, alternative text, tables that know they are tables. Most are not, and scanned ones are pictures of text, which means the content does not exist at all for anyone who cannot see it. There is a standard for accessible PDFs. It is rarely applied, because applying it takes work that the export button does not do for you.

It breaks the things the web is made of. You cannot link to the paragraph that matters, only to the file. You cannot search inside it from the site’s own search box. It opens in a viewer that is not your website, or lands in a downloads folder, where it goes stale the moment the rules change. The back button behaves oddly. None of the site’s navigation follows the reader in.

It ages badly, in public. A page is edited; a file is replaced. So there are two versions in circulation, one of them wrong, and the wrong one is in somebody’s downloads folder, attached to somebody’s email, cited in somebody’s letter. The web’s one real advantage — that the current version lives at a fixed address — is thrown away.

And it is heavy. Four megabytes against forty kilobytes for the same information as a page. On a slow connection that is the difference between an answer and a spinner.

The web already has a language

HTML was designed for something PDF was designed to prevent: text that rearranges itself for whoever is reading. It reflows on a small screen. It respects the size somebody has set because their eyes need it. It can be read aloud, translated, quoted, searched, linked to halfway down, and printed if anyone still wants paper. The author gives up control over where the line breaks fall, and in exchange the document reaches everybody. That is the deal the web is built on, and it is a good one.

Put the two side by side and PDF turns out to be a printout that escaped onto a website. It looks like a document because it was one — made in a word processor, laid out for A4, approved as a printed page, and then “published” by uploading the file. Nothing about that chain has the reader in it.

Why it happens anyway

Not out of laziness. A PDF is the shape of the internal process. It carries the letterhead, so it looks official. It is finished, so nobody can edit it afterwards. It matches what was signed off in the meeting. It can be produced by the person who wrote it, without asking the web team for anything — and that last one explains more of the web than any design theory.

Every one of those reasons is about the organisation. None of them is about the person on the bus trying to find out whether the deadline is Friday.

What to do instead

Make the page the real thing. The content goes into a web page, at a permanent address, with headings, a clear date and the practical facts near the top. If a printable version is genuinely needed, generate it from that page — the page is the original, the file is the copy.

Never let the file be the only route. Where a PDF must exist — a form to sign, a plan, a formal record — the page that links to it says in plain text what it contains, who it is for, what it costs and when it is due. Nobody should have to open a document to find out whether it concerns them.

Say what the reader is about to get. A link that reads “Application form (PDF, 2 pages, 180 kB)” respects the reader; one that reads “click here” does not.

And if a PDF stays, make it a proper one. Tagged, with a real document title, text rather than a scan, and a stable address that still works next year.

The test is simple enough. If the answer to a common question lives inside a file, it is not published. It is stored, on a website, where people are welcome to come and look for it.

Read the originals

The first one is a government explaining, in public, why it stopped.