In January 2026, a team writing in Scientific Reports compared more than 100,000 humans against frontier language models on standard divergent-creativity tasks. The headline: AI now beats the average person. The footnote, which matters more: the top 10% of humans still leave AI behind, and the gap widens on longer-form creative work — poetry, storytelling, anything requiring a throughline of taste.
That is the actual story of creativity in 2026. Not that AI replaces you. Not that AI is a toy. That there is now a measurable line between people who use AI to stay average and people who use AI to get further ahead — and which side of that line you sit on is almost entirely a function of how you collaborate with the model, not which model you use.
Stanford innovation expert Jeremy Utley calls the moment of realisation “the bathtub moment” — the point where you stop treating generative AI as a vending machine and start treating it as a thinking partner. A year on from his original framing, the research has caught up. Here is what changed, and what didn't.
What the 2026 research actually says
Three findings from the last six months matter for anyone doing creative work with AI:
- AI beats the median, not the mode of excellence. The Scientific Reports study (“Divergent creativity in humans and large language models”, January 2026) confirms that GPT-class models now outperform the average human on idea-generation tests — but the top decile of humans still dominates on originality, elaboration and narrative coherence.
- Pairing beats solo use, by a wide margin. Stanford's March 2026 work on generative-AI creative collaboration with visual artists found that humans working with well-conditioned AI produced measurably more novel output than either humans or AI working alone. The effect is largest when the human brings domain expertise and the AI is tuned to ask, not just answer.
- Consumers can tell — and they're starting to pay for the difference. Getty Images' 2026: Human Craft or AI Potential? report found that 66% of global consumers think creative work made by humans alone should be priced higher than AI-generated work, even as they accept AI-assisted work as the new default.
The productivity gap hasn't closed. Less than 10% of knowledge workers report capturing the 25–40% gains that studies say are available. The bottleneck is no longer the model. It's the method.
The Bathtub Moment: AI as Your Creative Companion
Creativity is not a talent reserved for painters, novelists and ad agency creative directors. It is the daily act of solving problems with original thought — writing a brief, framing a pitch, choosing what to cut from a deck. What changed in 2026 is not the definition of creativity; it is that every marketer, founder and operator now has a collaborator on call 24 hours a day who can keep up with any half-formed idea, at 3am, in any language.
The bathtub moment is the one where you stop typing instructions and start having a conversation. The research above is the quantitative version of that shift. The people getting 25–40% lift aren't prompting harder — they are thinking out loud with a partner who never gets tired.
Chapter 1: Don't Just Ask AI — Let It Ask You
Low performers treat AI as an oracle: one prompt, one answer, accept or reject. High performers treat it as an interviewer. They ask the model to interrogate the brief before generating anything. “Before you write this landing page, what five questions would you ask the founder?” is a more productive first prompt than “Write me a landing page.”
This inverts the power relationship in a useful way. You become the source of intent; the AI becomes the pressure-tester. That pattern — letting AI ask — is the single biggest behavioural difference between the 10% capturing gains and the 90% who aren't.
Chapter 2: AI Works Better When You Treat It Like a Teammate
Teammates get context. Teammates get the backstory, the constraints, the boss's weird preference for em-dashes, the three previous drafts that didn't work. Prompt engineering as a discipline has quietly died in 2026; in its place is context engineering — the practice of giving the model the same briefing you'd give a senior hire on day one.
This is why the bare chat window underperforms. It has no memory of your brand voice, your audience research, your last six campaigns. A system that holds that context for you — a Marketing OS, not a prompt box — is where the productivity gap actually closes.
Chapter 3: Inspiration Is a Discipline
The romantic view of creativity is that ideas strike. The research view is that people who generate a lot of good ideas are, overwhelmingly, people who generate a lot of ideas. AI flips this from a rate-limited process (how fast can you type?) to a volume game (how well can you critique?).
In 2026 the discipline is no longer producing — models will out-produce you on any given day. The discipline is selecting. You need a sharper taste, a clearer point of view, and a faster way to kill mediocre options. Inspiration becomes a filtering problem, not a generation problem.
Chapter 4: Go Beyond Good Enough
Satisficing is the creative killer of the AI era. Models are exceptionally good at producing work that is 70% of the way there. The temptation to ship at 70% is now enormous, because the marginal cost of another iteration is almost zero.
The market has noticed. The Adobe Digital Trends 2026 report finds that consumers are increasingly able to identify homogenised, AI-default output — the “ChatGPT voice,” the Midjourney glow, the stock-photo uncanny. Brands that accept 70% are training their audiences to tune them out. The 10% who go beyond good enough are becoming the only ones worth reading.
Realising Human Agency in an AI World
Agency, in the sense this post uses it, is the capacity to direct your own creative work — to know what you want, to recognise when you've got it, and to refuse what you haven't. AI does not remove agency. It exposes whether you had any to begin with.
The people who were already clear thinkers in 2023 are five times more productive in 2026. The people who were muddy thinkers are now muddy thinkers with unlimited output. That is the real inequality generative models have introduced, and it is one of the few inequalities that can be closed deliberately, by practice, in months rather than years.
What this means for marketers in 2026
For marketing teams specifically, three practical shifts follow from the 2026 research:
- Stop buying individual AI tools. Buy workflows. The gap isn't between GPT-5 and Claude; it's between teams whose context lives in one place and teams whose context is scattered across 14 browser tabs. Anjin is built for the first kind of team.
- Hire for taste, not for typing. The job description of a marketer in 2026 is closer to “creative director for a stable of AI agents” than “person who writes copy.” Taste — the ability to recognise the 10% output and reject the 70% default — is the scarce resource.
- Price on outcome, not effort. If AI has collapsed the cost of producing a campaign, clients and internal stakeholders will keep compressing budgets unless you price on what the work moved, not how many hours it took.
None of this is philosophical. It is a specific set of operational choices about how creative work is organised.
Anjin: The Marketing Operating System for AI-native creative work
This is where the “bathtub moment” stops being a mindset and starts being a piece of infrastructure. Anjin is the Marketing Operating System built for the era the research above describes. Most teams reading this post already believe the thesis — that AI is a partner, that context matters, that taste wins. They just don't have a system that lets them operate that way.
Anjin holds your brand context — positioning, audience, voice, prior campaigns — in one place, then deploys it across every AI-assisted task: briefs, drafts, channel plans, performance reviews. The agents ask you the right questions before generating. The context compounds rather than evaporating between chats. The output belongs to you, with an audit trail you can hand to a client or a CFO.
It is, specifically, the system for people who want to sit in the top 10% the Scientific Reports study describes — not by typing faster, but by thinking in a better environment.
The £888 Lifetime License — Offer Closing Soon
Lifetime access to Anjin for a one-time payment of £888. Not a subscription. Not a seat. Not a trial. One payment, unlimited use, for as long as Anjin exists.
The average marketing team spends £888 in about three working days on tooling, freelancers and coordination software. You're buying the platform that replaces most of it — once.
This price will not be offered again once we close our early-access cohort.
Claim your £888 Anjin lifetime license →Founders, agency owners and in-house marketers — this is how you run marketing at AI speed without the team, the burn, or another year of waiting.
Sources: Scientific Reports (Jan 2026), Stanford News (Mar 2026), Getty Images Creative Trends 2026, Adobe Digital Trends 2026, World Economic Forum (Jan 2026)
