A new study has shown that artificial intelligence is now capable of outperforming the average person in certain creativity tests—an important milestone in the evolution of generative models.

Can AI Generate Original Ideas?

The question of whether generative AI systems like ChatGPT can truly create new ideas has become the focus of a large-scale study led by Professor Karim Jerbi from the University of Montreal, with the participation of renowned AI researcher Yoshua Bengio. This research represents the largest direct comparison of human creativity and that of advanced language models.

An article published in Scientific Reports highlights a significant shift: modern generative AI systems have reached a level where, by several measures of creativity, they can surpass the average human. However, the most creative individuals still demonstrate a clear and consistent advantage, even over the most advanced models.

How the Comparison Was Conducted

The study evaluated leading language models, including ChatGPT, Claude, Gemini, and others, comparing their results with the responses of more than 100,000 people. The findings underscore a turning point: some AI systems, such as GPT-4, outperformed average human scores in tasks measuring divergent linguistic creativity.

Professor Karim Jerbi notes: “Our research shows that some AI systems based on large language models can now exceed the average level of human creativity in well-defined tasks. This may seem surprising—even unsettling—but it’s important to remember: even the best AI systems still lag behind the most creative people.”

Further analysis by Antoine Bellemare-Pepin and François Lespinasse revealed an interesting pattern: while some AI models already outperform the average person, the peak of creativity still belongs to humans. The most creative participants in the study consistently achieved higher scores than any of the tested models, and the gap was even more pronounced among the top 10% most creative individuals.

How Creativity Was Measured

To objectively assess creativity in both humans and AI, researchers used several methods. The main tool was the Divergent Association Task (DAT)—a widely used psychological test that measures divergent creativity, or the ability to generate diverse and original ideas in response to a single prompt.

In the DAT, participants—whether human or AI—were asked to name ten words that are as unrelated in meaning as possible. An example of a highly creative response: “galaxy, fork, freedom, seaweed, harmonica, quantum, nostalgia, velvet, hurricane, photosynthesis.”

DAT results closely correlate with other recognized creativity tests used to evaluate written work, idea generation, and creative problem-solving. Although the task is language-based, it goes beyond vocabulary, engaging broader cognitive processes related to creative thinking across different domains. DAT is also convenient: it takes only 2–4 minutes to complete and is available online to a wide audience.

From Word Lists to Creative Writing

Researchers then examined whether AI’s success in simple word association tasks extends to more complex and realistic creative challenges. To do this, they compared the performance of AI and humans in creative writing tasks: composing haiku, writing brief film plot summaries, and creating short stories.

The results were similar: while AI sometimes outperformed average humans, the most talented human authors consistently produced stronger and more original work.

Can AI Creativity Be Controlled?

These findings raised another important question: is AI creativity fixed, or can it be adjusted? The study showed that the level of AI creativity can be regulated through technical parameters, particularly the “temperature” setting, which determines how predictable or bold the model’s responses will be.

At low temperature values, AI produces more standard and safe answers. At higher settings, responses become more varied, less predictable, and more exploratory, allowing the system to move beyond conventional ideas.

It was also found that AI creativity is highly influenced by the way instructions are phrased. For example, prompts that encourage the model to reflect on the etymology and structure of words lead to more unexpected associations and higher creativity scores. This highlights that AI creativity largely depends on human guidance, making interaction and prompt formulation a central part of the creative process.

Will AI Replace Human Creators?

The study offers a balanced perspective on concerns about AI replacing creative professionals. While AI can now match or surpass the average level of human creativity in some tasks, it still has clear limitations and relies on human direction.

Professor Karim Jerbi emphasizes: “Even if AI can now reach human levels of creativity in certain tests, we need to move beyond the false sense of competition. Generative AI has become a powerful tool in service of human creativity: it won’t replace creators, but it will profoundly change how they imagine, explore, and create—for those who choose to use it.”

Rather than signaling the end of creative professions, the study’s results point to a future where AI acts as a creative assistant. By expanding ideas and opening new avenues for exploration, AI can enhance human imagination rather than replace it.

About the Study

The article “Divergent Creativity in Humans and Large Language Models” was published in Scientific Reports on January 21, 2026. The research involved scientists from the University of Montreal, Concordia University, the University of Toronto in Mississauga, the Mila Institute (Quebec Artificial Intelligence Institute), and Google DeepMind.