Liquid Content is a promising idea, but flexibility without structure means AI fills the gaps with plausible-sounding guesses.
Imagine writing a piece of text just once. Then one person reads it on their phone while riding the tram, another listens to it as a podcast while jogging, a third receives it as an interactive learning module, and an AI extracts the two sentences that someone just asked for in a chat. That's Liquid Content.
No rewriting, no copy-paste, no "we'd need this in another version." The content takes the form that the moment demands.
That's the idea behind Liquid Content. And I still think it's great.
But being liquid alone isn't a mark of quality. Water is liquid. So is spilled coffee.
If content adapts to any container without anything underneath holding it together, flexibility quickly turns into arbitrariness. The meaning gets watered down, a little more with each format, until in the end no one can say exactly what was actually meant.
For a long time, this was mainly a convenience issue. Today, it's a trust issue.
Because the reader is no longer just a human being. More and more often, it is a machine that processes, summarizes, and passes on our content. Georges-Simon Ulrich, Director of the Federal Statistical Office, captured this aptly in his "Data Literacy" series: when AI lacks reliable data, it does not flag the gap. It fills it. With something that sounds plausible. He calls this "apparent certainty."
👉 https://www.netzwoche.ch/news/2026-06-14/gute-daten-helfen-mensch-und-maschine
An AI that reads fluid but unstructured content does exactly that. It makes guesses where it should know. And it sounds convinced while doing so.
It is no coincidence that everyone is talking about "AI-ready data" right now. Switzerland's AI Action Plan has dedicated a separate pillar to the topic. It's about data quality, metadata, interoperability, and platforms like metadata.swiss. Important, long-overdue work.
There's just one level missing from this discussion for me. It almost always revolves around datasets, that is, tables, registers, numbers. But a huge portion of our knowledge isn't in datasets at all. It lives in content: in documents, standards, teaching materials, manuals.
And that is precisely where AI readiness is at its worst: at the content level. A dataset can be neatly documented, and the accompanying text can still be a jumble of PDF text from which no machine can reliably tell what is normative and what is an example, what is a definition and what is a footnote.
A backbone for content is less spectacular than it sounds. At its core, it comes down to just a few things.
The content is broken down into meaningful building blocks, not a jumble of running text. In bitmark we call them bits. Each one carries its own meaning and context. Each is individually addressable, and therefore citable, rather than just "somewhere on page 23." And if something changes, that exact bit can be updated without putting the entire document through the wringer again.
That sounds technical, but it's really just consistent organization. The same organization the Action Plan calls for at the data level (machine-actionable metadata, versioning, semantic clarity), applied one level deeper, to the content itself.
Anyone who knows me can guess where this is going. This very backbone is what we've been building with bitmark for years.
bitmark is an open Swiss standard for content. Machine- and human-readable at the same time, built AI-native rather than retrofitted for AI. Every bit carries its context, its metadata, and its address. The AI doesn't have to guess; it can point directly to the source.
And because content and presentation are cleanly separated, the same bit can appear as an app card, a printed page, a chat response, or an audio clip. Liquid Content is therefore not the opposite of structure. It is its reward. Only the solid backbone underneath makes the fluid form on top possible in the first place, without the meaning getting watered down.
Perhaps that is the real point of the whole AI-ready debate: we talk a lot about the form content should ultimately take, and too little about the skeleton that holds it up.
An AI that recognizes our content as a reliable source doesn't just fall from the sky. We have to build it. Bit by bit.
Liquid on top. Stable at the bottom. Both need each other.
This post was originally posted on Get More Brain's blog:
https://about.getmorebrain.com/liquid-content-needs-a-backbone/
Links:
