Artificial intelligence has reached a new milestone in the long-running debate over machine creativity, according to the largest study ever conducted comparing human creativity with that of generative AI systems
While advanced AI models can now outperform the average person on certain creative tasks, the most creative humans remain well ahead.
The findings come from a large-scale international study led by researchers at the Université de Montréal and published in Scientific Reports. The research compared the creative performance of several leading AI models with more than 100,000 human participants,
AI reaches average human creativity
The study focused on large language models such as ChatGPT, Claude, and Gemini, testing their ability to generate original ideas using the same methods applied to human participants.
The results showed that some advanced AI systems, including GPT-4, now overtake the average human score on specific measures of linguistic creativity.
Before this study, AI creativity was always seen as inferior to human originality. The new evidence suggests that, at least on structured tasks, AI has crossed a threshold that has always been unique to humans,
A continued gap in creativity
Although AI models performed strongly, the study shows there is still a creativity gap. The most creative half of human participants outperformed every AI system tested, and the top 10% of individuals demonstrated a particularly wide advantage.
This pattern suggests that while AI can match or exceed average performance, exceptional creativity remains firmly human. The results highlight creativity as a spectrum rather than a single benchmark, with machines clustering around the middle while humans dominate the upper range.
Measuring creativity
To make fair comparisons, researchers used the Divergent Association Task (DAT), a psychological test designed to measure divergent thinking. Participants are asked to generate ten words that are as unrelated to one another as possible, encouraging originality and conceptual distance rather than vocabulary size.
The DAT is widely used in creativity research because performance on the task correlates with results on more complex creative tests involving writing, problem-solving and idea generation. It is also quick to complete, making it suitable for large-scale studies involving both humans and AI.
To decide whether AI’s creativity extended beyond a simple word association, researchers also compared humans and machines on creative writing tasks, including haiku poetry, short stories and film plot summaries.
For these more complex tasks, AI sometimes exceeded average human performance, but once again fell behind highly creative individuals.
Can AI creativity be adjusted?
The study also explored whether AI creativity can be influenced through technical settings and instructions. Researchers found that adjusting a model’s temperature significantly affected creativity. Higher temperatures encouraged more varied and unexpected outputs, while lower temperatures produced safer, more predictable responses.
Instruction design also played a huge part; prompts that encouraged models to draw on word origins or unconventional associations led to higher creativity scores. This suggests that AI creativity is not fixed, but depends heavily on how humans configure and guide the system.
What does this mean for human creators?
Rather than the study suggesting that human creativity is coming to an end, it suggests that AI is an emerging creative tool that can assist exploration, ideation, and experimentation, particularly in terms of early-stage creative work.
Originality at the highest level remains distinctly human. The study suggests that the future of creativity is less about competition between humans and machines and more about collaboration, where AI expands the creative landscape without replacing the people who shape it.
