New research shows that large language models (LLMs), like those powering chatbots, align with human brain activity during creative thinking—and the bigger and more creativity-focused the model, the closer the match.
The Research
In a study published on arXiv (April 2026), researchers led by Mete Ismayilzada at EPFL, in collaboration with teams from the University of Trento, the University of Pennsylvania, and others, used fMRI data from 170 participants performing the Alternate Uses Task (AUT)—a classic divergent thinking test where people list unusual uses for everyday objects. They then extracted representations from various LLMs (ranging from 270M to 72B parameters) and compared their internal patterns to human brain activity using Representational Similarity Analysis (RSA).
They observed that brain-LLM alignment increased with model size, specifically in the default mode network (DMN), a brain network linked to creative thought. They also found that alignment was stronger when ideas were rated as more original—an effect that was most pronounced early in the creative process. Notably, post-training made a difference: a creativity-optimized Llama-3.1-8B-Instruct retained high alignment with high-creativity brain responses but lacked the positive alignment with low-creativity responses seen in other variants, while a reasoning-trained variant showed the opposite pattern—negative alignment with high-creativity brain activity.
The study was accepted at COLM 2026 and is available as arXiv:2604.03480.
Why It Matters
This research suggests that LLMs can serve as computational models of human creative cognition, offering insights into how creative ideas emerge in the brain. For you, this means that the same neural networks supporting flexibility and originality are being replicated in AI, and understanding this can help you appreciate the neural basis of your own creativity. Moreover, since larger models align better with brain activity, this hints that scale and training objectives shape how AI mimics human thought—and perhaps how you can optimize your own thinking.
By understanding that brain alignment varies with task and timing, researchers can develop better AI assistants for creative work and improve cognitive training methods to boost your divergent thinking.
What You Can Do
To boost your creative thinking, engage in practices that stimulate the default mode network: brainstorm unusual uses for common objects, take breaks to let your mind wander, and practice open-ended problem solving. These activities not only strengthen your creativity but also align your brain's activity with the patterns seen in high-performing AI models.
Source: arXiv q-bio.NC
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.
