Pairs of people beat people working with ChatGPT on creative ideas in a 208-person experiment
ChatGPT users wrote longer answers, but human pairs produced more original and more varied ideas, according to the authors' own account.
In brief
Two people out-brainstormed ChatGPT
In a test with 208 participants, human pairs gave more creative and more varied ideas than people paired with ChatGPT. ChatGPT users wrote longer answers. The data comes from the authors' own article.
People who brainstormed with another person came up with more creative ideas than people who brainstormed with ChatGPT. That is the result of an experiment described by two cognitive scientists, Clin Lai of Brandeis University and Roger Beaty of Penn State, in an article published on The Conversation on Oct 28, 2026.
How the test worked
ChatGPT offered the same clever idea across many separate sessions, making its overall pool of ideas more uniform.
The team randomly assigned 208 participants to one of two setups. Some worked with another person. Others worked with ChatGPT. The task was the alternate uses task.
It is a classic creativity test. You get a common object, such as a brick or a book, and list as many different uses for it as you can. Scientists then judge how many ideas there are and how original they are.
What they found
ChatGPT users wrote longer responses than people taking turns with a partner. Length did not translate into quality. Human pairs produced more creative ideas.
Their ideas were also more distinct and more diverse. One outlet that republished the article, Marsmag, adds that giving ChatGPT a persona to play raised the variety of its ideas, but it still fell short of human writers. Another, Singularity Hub, reports that the advantage of working with people grew as the collaboration went on.
Good ideas, repeated
The authors do not say ChatGPT is dull. Single suggestions could be quite creative. One example was using a book as a sound dampener.
The catch is repetition. ChatGPT offered that same idea across many separate sessions. Seen as a whole, its pool of ideas was more uniform. Imagine ten friends who each ask the same assistant for party themes and all get the same clever one.
Why it matters
AI is mostly sold on speed and output. The authors point out that models can produce far more ideas than people, and faster. This experiment asks a different question: what happens to originality when a machine replaces a human partner?
The article also cites other work, which it does not detail here, showing that AI outputs tend to be samey, including images that turn bland when generated repeatedly.
Limits
The source for the experiment is the authors' own article. The material reviewed shows no journal paper, university press release or independent coverage of the study. The other outlets carry the same text. The excerpts also do not give effect sizes, participant details or how creativity was scored, and they test only one task with ChatGPT.
Sources
- AI may quickly generate ideas, but working with people leads to more creative ones (theconversation.com)
- AI May Quickly Generate Ideas—but Teams of People Are Still More Creative (singularityhub.com)
- Replacing Hollywood’s Human Creative Teams with AI is Doomed, Here’s Why | MARS Magazine (marsmag.com)