Creative talent: has AI knocked humans out?

Are generative artificial intelligence systems such as ChatGPT truly creative? A research team led by Professor Karim Jerbi from the Department of Psychology at the Université de Montréal, and including AI pioneer Yoshua Bengio, also a professor at Université de Montréal, has just published the largest comparative study ever conducted on the creativity of large language models versus humans.

Published in Scientific Reports (Nature Portfolio), the findings reveal that generative AI has reached a major milestone: it can now surpass average human creativity. However, the most creative individuals still clearly outperform even the best AI systems.

AI reaches the threshold of average human creativity

The study tested the creativity of several large language models (including ChatGPT, Claude, Gemini, and others) and compared their performance with that of 100,000 human participants. The results mark a turning point: some AI models, such as GPT-4, now exceed the average creative performance observed in humans on tasks of divergent linguistic creativity.

“Our study shows that some AI systems based on large language models can now outperform average human creativity on well-defined tasks,” explains Professor Karim Jerbi. “This result may be surprising — even unsettling — but our study also highlights an equally important observation: even the best AI systems still fall short of the levels reached by the most creative humans.”

Analyses conducted by the study’s two co-first authors — postdoctoral researcher Antoine Bellemare-Pépin (Université de Montréal) and PhD candidate François Lespinasse (Université Concordia) — reveal a new and intriguing reality. While some generative AI systems now surpass average human creativity, the highest levels of creativity remain distinctly human.

In fact, the average performance of the most creative half of participants exceeds that of all AI models tested, and the top 10% of the most creative individuals open an even wider gap.

“We developed a rigorous framework that allows us to compare human and AI creativity using the same tools, based on data from more than 100,000 participants, in collaboration with Jay Olson from the University of Toronto,” says Professor Karim Jerbi, who is also an associate professor at Mila.

How do you measure human and AI creativity?

To compare human creativity with that of AI systems, the research team relied on several complementary approaches. The main one is the Divergent Association Task (DAT), a tool used in psychology to measure divergent creativity — the ability to generate many, varied, and original ideas from a single starting point.

Developed by study co-author Jay Olson, the DAT asks participants — human or AI — to produce ten words that are as semantically different from one another as possible. For example, a highly creative participant might suggest: “galaxy, fork, freedom, algae, harmonica, quantum, nostalgia, velvet, hurricane, photosynthesis.”

Crucially, performance on this task in humans also reflects performance on other well-established creativity tests, used in idea generation, writing, and creative problem solving. In other words, although the task is language-based, it does not simply measure vocabulary skills: it engages general cognitive mechanisms of creative thinking, relevant far beyond the linguistic domain. Another major advantage is that the test is quick — taking only two to four minutes — and easily accessible online to the general public.

Following this logic, the researchers then asked whether AI performance on this very simple task — generating a small set of semantically distinct words — would generalize to more complex creative activities closer to real-world creative practices. They therefore directly compared AI models and human participants on creative writing tasks, including haiku composition (a short three-line poetic form), movie plot summaries, and short stories. Here again, the most skilled human creators retained a clear advantage, even though AI systems can sometimes outperform average human creativity.

Is AI creativity a matter of tuning?

These findings naturally led the researchers to a key question: can AI creativity be modulated? The study shows that it can — notably by adjusting the model’s temperature, a technical parameter that controls how predictable or daring the generated responses are. At low temperature, AI produces cautious and predictable outputs; at higher temperature, it introduces more randomness, takes greater risks, and encourages the system to move beyond well-trodden paths, generating more varied and original associations.

The study also shows that how instructions are phrased strongly influences AI creativity. For instance, a prompting strategy based on etymology — encouraging the model to draw on the origins and structure of words — leads to less obvious associations and higher creativity scores. Together, these findings highlight a central point: AI creativity depends closely on how humans guide and parameterize these systems, making human–AI interaction a key element of the creative process.

Will human creators be replaced?

These results provide a nuanced perspective on concerns about the potential replacement of creative workers by artificial intelligence. While some AI systems can now rival human creativity on specific tasks, the study also underscores the current limits of machines and the central role of humans in creativity.

“Even though AI can now reach human-level creativity on certain tests, we need to move beyond this misleading sense of competition,” says Professor Karim Jerbi. “Generative AI has above all become an extremely powerful tool in the service of human creativity: it will not replace creators, but profoundly transform how they imagine, explore, and create — for those who choose to use it.”

Rather than announcing the disappearance of creative professions, the study invites us to rethink AI as a creative assistant, capable of expanding possibilities for exploration and inspiration. The future of creativity may lie less in opposition between humans and machines than in new forms of creative collaboration, where AI enriches human ingenuity instead of replacing it.

“By directly confronting human and machine capabilities, studies like ours push us to rethink what we mean by creativity,” concludes Professor Karim Jerbi.

About the Study

The article “Divergent creativity in humans and large language models” was published in Scientific Reports on January 21, 2026. The study represents an exceptional collaboration among researchers from Université de Montréal, Université Concordia, University of Toronto Mississauga, Mila (Quebec AI Institute), and Google DeepMind.

The research was led by Professor Karim Jerbi, with co-first authors Antoine Bellemare-Pépin (Université de Montréal) and François Lespinasse (Université Concordia). The author team also includes Yoshua Bengio, founder of Mila and LoiZéro, and one of the world’s leading pioneers of deep learning — the technology underpinning modern AI systems such as ChatGPT.