Can generative artificial intelligence systems, such as ChatGPT, demonstrate creativity comparable to that of humans? This question was the focus of a new large-scale study, the results of which were published in the scientific journal Scientific Reports.
The findings indicate that certain contemporary AI systems are now able to surpass average human performance on specific measures of creativity. At the same time, the study shows that the highest levels of creative ability continue to be found exclusively in humans.
Analysis of the data revealed a clear pattern. While some AI models outperformed the average of human participants, the most creative individuals—particularly those in the top 10 percent—achieved significantly higher scores than all AI systems tested.
“We developed a rigorous methodological framework that allows for the direct comparison of human and machine creativity using the same tools, based on data from more than 100,000 participants,” explains the study’s lead author, Karim Zerbi, Professor of Psychology at the University of Montreal.
Methodology and Assessment of Creativity
To ensure comparability, the research team employed multiple evaluation methods. The primary tool was the Divergent Association Task (DAT), a psychological test designed to measure divergent creativity.
Developed by study co-author Jay Olson, the task asks participants—either humans or AI systems—to generate ten words that are as semantically distant from one another as possible. High scores are achieved when the produced words belong to entirely different conceptual domains.
Previous research has shown that performance on this task is strongly correlated with established measures of creativity used in creative writing, idea generation, and problem-solving. Although the task is language-based, it does not simply assess vocabulary breadth; rather, it reflects broader cognitive processes associated with creative thinking.
From Words to Complex Creative Tasks
Building on the results of the initial word-based test, the researchers examined whether AI performance extended to more complex forms of creativity. To this end, they directly compared humans and AI systems on creative writing tasks, including haiku composition, film plot summarization, and the production of short narrative texts.
Once again, the same pattern emerged: while some AI systems approached or exceeded average human performance, the most capable human creators maintained a clear advantage.
Factors Influencing AI Creativity
An additional research question concerned whether AI creativity can be regulated or shaped. The data suggest that it can. A key factor is the model’s “temperature,” a technical parameter that affects the variability and predictability of generated responses.
At low temperature settings, models tend to produce more conservative and predictable outputs. Higher temperature values promote greater diversity and less expected associations, increasing originality.
At the same time, the formulation of prompts proved to be critical. Instructions that encourage, for example, etymological analysis or the exploration of conceptual structures lead to higher creativity scores. These findings underscore that the creative output of AI systems depends to a large extent on human guidance.
Conclusions and Broader Implications
The study provides a nuanced and evidence-based response to concerns that artificial intelligence may replace human creators. While some AI systems can now compete with the average human in specific creativity tasks, the results reveal clear limitations and reaffirm the uniqueness of high-level human creativity.
The researchers evaluated several large language models, including ChatGPT, Claude, and Gemini, comparing their performance with data from more than 100,000 human participants. Some models, including GPT-4, achieved higher scores than the average human in tasks measuring linguistic divergent creativity.
Nevertheless, as Zerbi emphasizes, “even the most advanced AI systems still fall short of the most creative humans.” Overall, the findings support the view that the future of creativity will be shaped not by competition between humans and machines, but by new forms of collaboration in which artificial intelligence serves as a powerful tool for augmenting human imagination.