The worry is in the air: AI makes humans uncreative. It delivers prefabricated answers, everyone uses the same tools, and in the end everything looks the same. Two studies put this worry to numbers.
What the studies actually measured
In 2024, the economics researchers Anil Doshi and Oliver Hauser published a writing experiment in the journal Science Advances. 293 people each wrote a short story of eight sentences. One group could call up one idea from an artificial intelligence, another up to five ideas, the rest wrote alone. 88.4 percent of the authors with AI access pulled at least one idea. 600 readers rated the stories, each person six randomly selected pieces, without knowing who used AI. Randomly assigned, rated blind: a clean experiment.
The core numbers: stories with access to five AI ideas were rated 8.1 percent more novel than stories without AI. Usefulness, rated as audience fit and publication potential, rose by 9.0 percent. The strongest effect showed among authors with low scores in a prior creativity test: their rating for how well written the story is rose by up to 26.6 percent.
The second study delivers the counterpart. A team around Barrett Anderson had 33 people collect ideas, once with ChatGPT, once with a classic creativity tool, a deck of cards with prompts. With ChatGPT, more and more detailed ideas emerged on average. But: the ideas of different people resembled each other more with ChatGPT. Whoever asks the same machine gets answers that converge.
- Rated more novel, with five AI ideas
- +8.1 percent (Science Advances, July 2024)
- Rated better written, less creative authors
- +26.6 percent (Science Advances, July 2024)
- Ideas of different users with ChatGPT
- more alike in content (C&C conference 2024)
- Freelancers in the QLTY network
- more than 200
Why AI makes individuals more creative
The effect is measured, the explanation is not. Plausible: creativity often depends on the start. Anyone sitting in front of a blank page loses time to doubt. An idea as a starting point removes the blockage. The rest is craft: execute, discard, improve.
The authors of the study describe the effect as professionalisation: stories feel more rounded. Presumably the AI idea carries structure with it. Those who already bring structure gain less. That explains why less creative authors benefit the most. AI lifts the level from below.
For companies this means: AI lowers the entry barrier. Anyone who never dared to think through a campaign, a video, a game now gets a drafting partner that never tires. The number of people who can turn ideas into reality grows.
Where the uniformity comes from
The convergence is not a rumour. Stories with one AI idea lay 5.2 percent closer to the AI idea, stories with five ideas 5.0 percent. Whoever lets the AI supply the concept as well hangs on its suggestion. Everyone gets faster. And the results resemble each other more.
Both experiments hold the task fixed. Doshi and Hauser had everyone write the same kind of short story, Anderson's team had ideas collected on given topics, such as product improvements. Within a fixed task there is a limited idea space. That is exactly where it narrows.
Add a finding for everyone who sells results under their own name. The readers of the study attributed at least 25 percent less own share to AI users after disclosure. Anyone claiming everything was their own idea loses credibility the moment someone asks.
How orchestration shifts the limit
This is where what the studies prove ends. They test text against text and do not measure what happens when someone combines media. A computer game as a project presentation. An explainer video with its own music. A website that adapts to its visitors. Such combinations used to be team projects over weeks. Today one person coordinates the tools and production steps. We call that orchestration.
This reading rests on an old logic: the synthesizer made studio production affordable. Digital editing software made films independent of the laboratory. Every tool that cut production costs created forms that were previously too expensive. AI continues this series, this time across media boundaries.
Within a task the idea space narrows. The number of possible form combinations grows with every tool. Whether diversity really increases in the end is open. Combinable forms can converge too. Nobody has measured that.
Measured: with AI, individual work is rated more creative, and the results resemble each other. Open: the level of combination. That is exactly where new things are emerging.
What this means for your projects
First: keep the concept with the human. The AI delivers building blocks, selection and review remain your task. That reduces the risk of hanging on the first AI suggestion and keeps the attributed own share high. Not a guarantee, a working recommendation.
Second: the gain lies not in the tenth variant text, but in combining video, game, sound and software into one experience. That is our position from project work, not a study result. Third: label where labelling is required. Article 50 of the EU AI Act has applied since August 2026.
QLTY media works exactly at this point. We orchestrate film, music, web and AI content from Cologne, for mid-sized companies and industry. Editing, sound and programming are handled by specialists from a network of more than 200 freelancers.
Free initial consultation: If you have an idea that was too expensive so far: we check in a conversation which parts AI can take over and which not. No quote on call, no budget before the first result.
Free initial consultation






AI in Marketing for Mid-Sized Companies: What Actually Works in 2026
AI labelling: what the EU Code of Practice regulates