Liquid Content: Why Content Atomization Needs Solid Atoms

Liquid content is content designed as structured knowledge — facts, quotes, context — that AI can reassemble into different formats, surfaces, and experiences depending on who’s consuming it and where. The term came out of news publishing in January 2026. The idea is bigger than publishing, and most of the conversation is missing its hardest problem: the atoms.

I’ve been atomizing one specific kind of content — customer interviews — since late 2024, before the term existed. It started as a production problem. I had transcripts, and I needed a better way to structure them: which quotes could carry a case study, which belonged in a testimonial video, which answered the objection a rep kept running into. So the quotes became rows in a spreadsheet.

Then came the part I didn’t expect. Customers would approve as many as sixty quotes out of a single conversation. The constraint was never their willingness to go on the record — it was that nobody had asked them properly, or kept track of what they’d already said yes to.

That spreadsheet is why I’m writing this. As AI got better at assembling content, and provenance stopped being an abstraction, it became clear we weren’t just making case studies faster to produce. We were accumulating a record of provenance for the material a company can’t source anywhere else. Some of that is your historical first-party data, in whatever shape it’s in. The part we work on is the voice of the people with something to say: for a SaaS company that’s usually the customer, and sometimes it’s the founding team — the ones who actually understand where the product sits in its industry. Everything we’ve built since comes back to the same question: how do you keep that material credible, and how do you prove it’s yours?

So I’ve been living with the problem publishers are just now naming. Here’s the term, where it came from, what it means for B2B, and the part of the stack nobody owns yet.

What is liquid content?

The Reuters Institute’s 2026 journalism trends report, published in January, defined liquid content as stories that are not static but adapt in real time based on the viewer’s context, location, time, or interaction — and noted it requires media companies to move away from authoring “articles” toward more flexible atomic objects.

Bauer Media’s chief product officer Marcel Semmler put it plainly in Digiday’s explainer: publishers have traditionally created content as a finished object; liquid content shifts that thinking toward content as structured knowledge that can flow into different formats, surfaces, and interfaces.

If you’ve met this family of ideas as content atomization, modular content, or structured content, you already have the right picture — liquid content is that school of thought with a generative assembly engine attached. The atoms are the same; what changed is how cheaply they can be reassembled.

In practice: one reported story becomes a newsletter item, a vertical video, a podcast segment, a chatbot answer, and a personalized briefing — each assembled on demand from the same underlying knowledge, rather than hand-produced as separate artifacts.

Is liquid content just content repurposing?

Not quite — but it’s the same instinct, systematized. Content repurposing takes a finished piece and manually cuts derivatives from it: the webinar becomes clips, the report becomes social posts. The output multiplies, but every derivative is still hand-made, and the source is still a frozen artifact. Liquid content moves the work upstream: structure the knowledge first, and let the derivatives be assembled — not crafted — on demand.

The instinct has a longer history than either term. In 2009, NPR’s Daniel Jacobson published the COPE philosophy — Create Once, Publish Everywhere — arguing that content systems should capture content agnostic to how or where it would be viewed, separating the knowledge from its presentation so one creation effort could feed every platform, including platforms that didn’t exist yet. COPE became the founding document of structured content and shaped a generation of content architecture.

What’s new in 2026 is the assembly engine. COPE could route the same structured content to different surfaces; it couldn’t reshape it. Generative AI can — a text story becomes audio, a long read becomes a briefing, a fixed artifact becomes an answer. Liquid content is COPE with a transformation layer, which is why it’s arriving now: the marginal cost of a new format just collapsed.

That history matters because it predicts where the hard problems will be. COPE’s lesson was that the constraint is never the distribution — it’s the discipline of the source system. Seventeen years later, the same is true, with one new twist.

What problem does atomization not solve?

Provenance. When machines reassemble content, attribution is exactly what breaks first — and this is measured, not hypothetical.

In March 2025, Columbia’s Tow Center for Digital Journalism tested eight AI search engines by feeding them direct excerpts from real news articles and asking for the source. Collectively, the tools answered more than 60% of queries incorrectly — fabricating links, crediting syndicated copies instead of originals, misattributing publishers — and did so, in the researchers’ words, with alarming confidence, rarely qualifying or declining to answer. Even publishers with licensing deals in place saw their work misattributed.

Buyers have noticed. In Gartner’s 2026 survey of 645 B2B buyers, 45% used generative AI to research vendors and products — and 69% said they prefer to validate AI-generated insights with a sales rep. TrustRadius’s 2026 buying research found the same reflex hardening into habit: the share of buyers who always or very often fact-check AI-generated information jumped from 58% to 72% in a single year, and 94% of buyers who used AI verify at least some of the time.

Read those together and the pattern is clear: the more machines assemble our information, the more the humans on the receiving end demand a way to check the source. The industry is responding at the infrastructure layer — the C2PA standard (backed by Adobe, Microsoft, Google, the BBC, and dozens of others) now attaches cryptographically signed “Content Credentials” to media files, and the EU AI Act’s machine-readable marking requirements apply from August 2, 2026. But note what those systems certify: the file — where an image or video came from and how it was edited. Nothing in the emerging stack certifies the statement: who said this, whether they actually said it, and whether they approved this use of it.

So the liquid-content stack has three layers: the reservoir of structured knowledge, the assembly engine, and the verification layer that certifies each atom. Publishers and vendors are racing to build the first two. The third barely exists.

What does liquid content mean for B2B marketing?

B2B has the same disease with a different frozen object. The publisher’s finished object is the article. B2B’s is the case study.

We analyzed 56 of the fastest-growing B2B SaaS companies for our social proof research report. 88% publish case studies — an average of 45 per website. The average case study runs 965 words. And 93% of the ones we reviewed follow the same Challenge–Solution–Impact arc. That is a monoculture of frozen objects: one format, one length, one story shape, produced by hand, one at a time.

Meanwhile the source material is the richest structured knowledge a B2B company owns. A recorded customer interview contains the buyer’s pain in their own words, the alternatives they evaluated, the objections they had, the outcome numbers, and the way they categorize you against competitors — often in language your team would never have guessed. In our own delivery work, a single 30-to-60-minute interview yields anywhere from 25 to 60 distinct quotes, and we typically submit the majority of them for approval. We started in late 2024 by treating those quotes as rows in a spreadsheet — and the moment you do that, the machine-readable layer suggests itself. Each quote picks up a statement type, a sentiment, the point it’s making, synonyms for search, and a strength rating against the value propositions and use cases it could support. That enrichment is what makes reassembly possible at all. We wrote about mapping one interview across the full buyer funnel back in January — before “liquid content” gave the pattern a name.

One of our customers already runs this way. Spendgo, a restaurant-loyalty platform, used one interview program to feed direct sales, partner marketing, and ultimately an acquisition. Neal Dubisky, then Spendgo’s VP of Sales & GTM Strategy, described the mechanic:

Different versions for different audiences, assembled from the same approved atoms — and his summary of what sits underneath it is the whole thesis of this piece in a customer’s words:

“The interview became the asset that created everything else.”

Neal Dubisky, former VP Sales & GTM Strategy, Spendgo

One recorded interview fans out into the whole asset set. We broke the same pattern down across the case studies of the 56 fastest-growing B2B SaaS companies.

The video data makes the frozen-object cost visible. Only 36% of the companies in our analysis include video in their case studies — but when companies designate “featured” case studies, video appears in more than 90% of them. Everyone agrees the richer atom is more valuable. Almost nobody has a system that captures it. That’s not a preference gap; it’s an acquisition gap — the exact gap a liquid pipeline is supposed to close.

So the B2B translation of liquid content: stop treating the case study PDF as the asset. The interview is the reservoir. The case study is one crystallization of it — one of many the same atoms should produce.

What makes a content atom safe to reassemble?

This is where B2B can learn from journalism’s oldest standard rather than its newest term: the record.

When someone goes on the record, they stand behind their words, and those words carry authority precisely because the provenance is explicit. An atom is safe to liquefy when it carries four things everywhere it flows:

Two refinements from actually running this in production.

First, the record is something you return to, not something you open with. Journalists ask their subjects whether they’re on the record before the conversation starts; in B2B, that kills candor. Your customer is often operating under an NDA and a data processing agreement with you — the best material comes when they can speak freely with a partner they trust. So we invert the mechanic: interview first, then go back with the specific quotes we’d like to put on the record — “is this okay to publish and reference again?” Same standard journalism built, resequenced for the B2B relationship.

Second, not every atom should be publishable — and the system has to know the difference. A customer riffing on their read of the industry may never be citable in an article, but it’s exactly what you study to distill buyer language, voice, and tone. So attribution has grades: approved for marketing use versus internal intelligence. You need a platform that marks that line on every atom, because the two failure modes are symmetric — publish what you shouldn’t, or bury what you could.

Without those, liquefaction degrades trust at machine speed — the Tow Center data shows what reassembly does to attribution even when the source is a major newsroom. With them, the same atoms compound, and they compound in exactly the direction the AI-search era rewards. The Princeton-led GEO study (Aggarwal et al., KDD 2024) measured which on-page tactics lift a source’s visibility in generative-engine answers: adding quotations produced roughly a 41% visibility gain — the largest single-tactic lift measured — with statistics around 30% and citing sources around 28%. (Per-tactic findings; they don’t sum.) The single highest-value atom in the liquid stack is the attributed quote. Which means the verification layer isn’t compliance overhead — it’s the value.

That verification layer is what we build at Proofmap: a proof of record where every atom is on-record, dual-consent, attributed, and traceable to the recorded source. But the argument stands whether or not you use us: if your atoms can’t carry their provenance, your liquid content is a liability with good distribution.

The question both industries are answering

Publishers are asking whether the article should remain the unit of production. B2B marketers should be asking the same thing about the case study. Underneath both is one question: is your knowledge an asset, or an archive of frozen outputs?

If the answer you want is “an asset,” the next question is practical: what would it actually take to make your knowledge machine-readable — not as a metaphor, but as a working system?

How do you make content machine-readable?

At the atom level, you’ve already seen our answer: every quote travels inside its wrapper — attribution, consent grade, enrichment, and the trail to source. That’s the Proofbase, and it’s what makes atoms safe to hand to an assembly engine.

But atoms alone don’t decide what your content should say. Marketing can feel like it’s all about perception, which is why people struggle to see how positioning becomes operational with AI. It becomes operational as a canon — the premise everyone leveraging AI in your organization operates from and can reference: your core positioning, your guardrails, your voice and tone. The more liquid your content gets — the faster dynamic versions get created — the more necessary the canon becomes. The atoms supply what’s true; the canon supplies what you stand for.

Tracing every statement back to the record is one half of the discipline. Governing what all of it adds up to is the other.

Where should you start?

Honest answer: this is a content operations decision before it’s a tooling one. And there are two projects here, different in kind.

The first is migration — taking the archive you already have, the published case studies and decks and webinar recordings, and reverse-engineering it back down to the atom level. It’s worth doing, and it’s an exercise in itself: most of the provenance you need was never captured, so you’re reconstructing attribution and consent after the fact.

The second is capture — deciding that from today, everything new gets collected in its most liquid form at the source. In practice that means interviews and conversations, recorded, attributed, and consented in the moment: the same capture discipline journalism has always run, applied to your customers and your subject-matter experts.

I won’t argue one is more valuable than the other. But they build different things. Migration makes your past usable. Capture compounds: every interview from this point forward adds atoms that are liquid from birth — and a competitor who starts later can’t backfill years of on-record capture they never did. That’s a moat, and you start building it the day your capture changes, not the day your migration finishes.

And when you choose what to capture, weight video — because liquid content runs headlong into a second force: the AI trust gap. The two curves are moving against each other. On one side, every team is adopting exactly the tactics this piece describes: more AI, more assembly, more versions of everything. On the other, the people consuming it all trust machine-assembled content a little less every year — the misattribution numbers and the buyer fact-checking data earlier in this piece are that curve made visible. Liquefy without offsetting the trust gap and you are accelerating into skepticism: more output, met with less belief.

Video is the offset. The strongest answer to “did a human actually say this?” isn’t a disclosure label; it’s footage — a real person, on camera, saying the words, with the approval and the trail attached. It’s the same instinct behind the featured-case-study pattern earlier in this piece. When any content can be generated, watching someone stand behind their words may be the most durable trust layer there is.

The organizations that win the liquid era won’t be the ones that melt their content fastest. They’ll be the ones whose atoms are solid enough to survive the melting.