AI and creativity: better stories that look more alike
AI ideas can make one person's story better and make everyone's stories more alike. In a 2024 experiment with 293 writers, stories written with AI ideas were rated more novel and more useful, and were more similar to each other. A second study with 1,100 people found that direct answers from AI raised scores on one task while it was there. Afterwards, working alone, the groups that had coach-style AI scored lower than people who never had AI. Both are peer reviewed.
The two studies in one table
| Study | People | Design | Status |
|---|---|---|---|
| Doshi and Hauser, 2024 | 293 writers, 600 judges | Experiment with three groups | Peer reviewed. Science Advances |
| Kumar and others, 2025 | 1,100 across two experiments | Two experiments, planned and registered in advance | Peer reviewed. CHI 2025 conference |
The story experiment
Doshi and Hauser asked 293 people to write a short story. One group wrote with no help. One group could see a single story idea from an AI. One group could see five. Then 600 other people judged the stories.
- Better, one by one. With five AI ideas, stories were rated 8.1% more novel and 9.0% more useful than stories written with no help.
- The biggest gain went to the least creative writers. For them, novelty rose 10.7% and usefulness 11.5%.
- More alike as a set. The paper says stories written with AI ideas "are more similar to each other than stories by humans alone".
That is the trade in the title of the paper: more creativity for the individual, less variety across the group.
Why "more alike" matters to you
If you are the only person using AI ideas, a better story is simply a better story. But you are not the only one. When many people start from the same kind of suggestion, their work drifts toward the same place. For a school essay, a job application or a pitch, sounding like everyone else has a cost that a quality score does not show.
What happens when the AI is gone
Kumar and colleagues tested a different question. In two experiments with 1,100 people, some worked with a standard AI that gave direct answers, some with an AI set up like a coach, and some with no AI. The tasks covered two kinds of creative thinking: coming up with many ideas, and finding the one that fits.
Their finding, in the words of the paper, is that AI help boosts performance while people have it but may "inadvertently hinder independent creative performance" afterwards. In the detail, the clear drops afterwards were in the coach-style groups: lower originality in the first experiment, and 10.6 points lower accuracy in the second. The direct-answer groups did not differ significantly from people with no AI.
What these studies did not show
- Not that AI makes people less creative. In the story experiment, the stories got better ratings with AI ideas.
- Not a lasting effect. The story experiment was one session. Nobody was followed over months.
- Not all kinds of creative work. Short stories by people recruited online, with ideas from one AI model. Novels, music, design and science were not tested.
- Not a certain harm afterwards. The second paper says "may hinder". That is a careful wording, and it should stay careful.
A fair one-line summary. "In a peer-reviewed experiment, AI ideas made individual stories more creative and made the stories more similar to each other."
What to do with this
Have your own idea first. Write it down, even as one rough line. Then ask the AI for ideas and compare. Your idea is the part nobody else starts with, and the story experiment suggests it is the part to protect. The same habit shows up in a small MIT preprint on essay writing, where people who wrote alone first stayed more engaged when they later used AI.
Questions
Does AI make writing more creative?
In one experiment, yes for single stories: 8.1% more novel and 9.0% more useful with five AI ideas. The same experiment found the stories became more similar to each other.
Does AI make everyone's writing sound the same?
The experiment found more similarity between short stories written with AI ideas. It did not test other kinds of writing, so "everyone's writing" goes further than the evidence.
Who gained most from AI ideas?
The writers who were least creative to begin with. Their stories rose 10.7% in novelty and 11.5% in usefulness.
Will I be less creative without AI after using it?
One peer-reviewed study found lower scores afterwards for people who had used coach-style AI, and no clear drop for people given direct answers. It is one study with careful wording. Having your own idea first is a cheap way to stay on the safe side.