"AI brain rot" and cognitive offloading Metacognition: knowing what you know
"Your Brain on ChatGPT": what the MIT essay-writing study did and didn't show
In short: People who wrote essays with ChatGPT showed the weakest brain connectivity on EEG, felt the least ownership of their essays and often couldn't quote what they had just written, compared with people using a search engine or no tools. It is one of the most-shared "AI brain rot" studies, but it is a small, unreviewed lab study about essay writing, not proof of lasting brain damage.
| Paper | Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task |
| Authors | Nataliya Kosmyna, Eugene Hauptmann, Ye Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beresnitzky, Iris Braunstein, Pattie Maes |
| Published | June 2025 · arXiv preprint 2506.08872 (MIT Media Lab) |
| Study type | Lab experiment with EEG · Preprint (not peer-reviewed at the time of summary) |
| What we read | Full text read. We read the full preprint, including its limitations section. |
| Summarized |
What they did
Participants wrote essays in three sessions in one of three conditions: ChatGPT, a search engine, or "brain-only" (no tools), while EEG recorded brain activity. In an optional fourth session, the ChatGPT group wrote without tools and the brain-only group used ChatGPT. Essays were analysed with language-processing methods and scored by teachers and an AI judge, and participants were interviewed.
Who took part
54 adults aged 18–39 from five universities in the Boston area (MIT, Wellesley, Harvard, Tufts and Northeastern); 18 completed the fourth session.
What they found
- Brain connectivity was strongest in the no-tools group, moderate with a search engine and weakest with ChatGPT.
- ChatGPT users reported the lowest ownership of their essays and struggled to quote their own work accurately.
- Essays within each group were similar to one another in wording and topics.
- When ChatGPT users switched to writing unaided, they showed lower alpha and beta connectivity, which the authors read as under-engagement. Brain-only users who switched to ChatGPT showed higher recall.
Effect size: No single headline effect size; results are reported as EEG connectivity differences and behavioural measures across groups.
Limitations
- Small sample (54, and 18 in session four) from a few nearby universities.
- One task (essay writing) and one model (ChatGPT); the authors say results may not generalise.
- EEG shows activity patterns, not learning or long-term change.
- A preprint, not peer-reviewed at the time of summary.
What this means for learners and parents
- If AI writes the first draft, you may remember less of it and feel it isn't yours. That matters when the point is to learn.
- Order seems to matter: thinking and writing on your own first, then using AI, looked better than the reverse.
- Treat this as a reason for "think first" habits, not as proof that AI damages brains.
Put it into practice (free tools)
- Who is doing the thinking? Check a prompt
- Family AI rules: try first, then ask
- Write what you remember first: the blurting method
Citation
Nataliya Kosmyna, Eugene Hauptmann, Ye Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beresnitzky, Iris Braunstein, Pattie Maes (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. arXiv preprint 2506.08872 (MIT Media Lab). https://doi.org/10.48550/arXiv.2506.08872
Related summaries
- Unrestricted ChatGPT help raised practice scores but lowered exam scores · Hamsa Bastani et al., 2025 · Preregistered randomized controlled trial
- ChatGPT made research feel easier but students' reasoning was weaker than with Google · Matthias Stadler et al., 2024 · Randomized experiment
- With AI, people scored better on reasoning puzzles but overestimated how well they did · Daniela Fernandes et al., 2026 · Two online experiments
- The classic study: testing yourself beats rereading after a week · Henry L. Roediger III et al., 2006 · Lab experiments