A team of neurologists and AI specialists at MIT’s Media Lab has led a study looking into the brain impacts of large language model (LLM) use among people who engage with them for study or work. They report evidence that the use of LLMs may lead to an erosion of critical thinking skills. In their study, posted on the arXiv preprint server, the researchers asked groups of volunteers to write essays while connected to EEG monitors.
Over the past few years, the use of LLMs such as ChatGPT has become commonplace. Some use them for fun, while others use them to help with school or work responsibilities, and the team at MIT wondered what sort of impact LLM use might have on the brain.
To find out, they recruited 54 volunteers. The initial group was then split into three small groups, all of whom were asked to write a 20-minute essay on the topic of philanthropy—one group was asked to use ChatGPT for help, the second was asked to use Google Search, and the third “Brain-only” group was given no tools or resources at all. The participants remained in these same groups for three writing sessions.
Each of the volunteers was fitted with an EEG device to monitor brain activity, such as cognitive engagement and mental workload, while they wrote. The researchers also performed natural language processing analysis and interviewed participants after each session. Essays were scored by human teachers as well as an AI agent.
For these first three sessions, the EEG analysis showed clear differences in brain connectivity between the groups. The Brain-only group showed the strongest and most widespread brain network activity, the search engine group showed intermediate levels of engagement, and the ChatGPT group showed the weakest overall brain connectivity.
Out of the 54 volunteers, 18 also returned months later to complete a fourth session. Those who had used ChatGPT now went the Brain-only route, and vice versa. In this session, those who had originally used ChatGPT in the first three sessions showed weaker neural connectivity, while those from the previous Brain-only group demonstrated higher memory recall.
Overall, the interviews also revealed that volunteers who used the LLM felt less ownership over their essays compared to the other groups. ChatGPT users also struggled to recall or quote from their own essays shortly after writing them. Across all measures—brain activity, language analysis, and essay scoring—participants who relied on ChatGPT performed worse than the Brain-only group
These findings highlight the potential educational impact of relying on LLMs for writing tasks. While they can provide immediate benefits, they run the risk of reduced learning outcomes over time and frequent use may hinder the development of critical thinking skills.
Written for you by our author Bob Yirka,
edited by Lisa Lock
, and fact-checked and reviewed by Robert Egan —this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive.
If this reporting matters to you,
please consider a donation (especially monthly).
You’ll get an ad-free account as a thank-you.
More information:
Nataliya Kosmyna et al, Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task, arXiv (2025). DOI: 10.48550/arxiv.2506.08872
Project page: www.brainonllm.com/
Journal information:
arXiv
© 2025 Science X Network
Citation:
Using ChatGPT to write essays may be eroding critical thinking skills (2025, June 20)
retrieved 20 June 2025
from https://phys.org/news/2025-06-chatgpt-essays-eroding-critical-skills.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no
part may be reproduced without the written permission. The content is provided for information purposes only.
