An MIT brain-scan study found that using AI to write essays slashed brain activity by 47% — and 83% of those users couldn’t even remember what they had just “written.”
Story Snapshot
- MIT researchers measured brain activity across 32 regions and found AI writing users showed a 47% drop compared to people who wrote on their own.
- A joint Microsoft and Carnegie Mellon University study of 319 office workers found that higher trust in AI led to less critical thinking and weaker independent problem-solving.
- A researcher at the Swiss School of Business tested 666 people in the UK and found frequent AI users scored lower on critical thinking than older adults who rarely used it.
- Scientists warn that leaning on AI without staying mentally engaged risks “deskilling” — slowly losing the ability to think through hard problems on your own.
What the Brain Scans Actually Showed
MIT researchers used electroencephalogram (EEG) technology to track brain activity across 32 regions while people wrote essays. The group using ChatGPT showed 47% less brain activity than the group writing without AI help. Two English teachers who reviewed the AI-assisted essays called them “soulless” — full of the same phrases and patterns, with little original thought. The results raised a clear red flag: when AI does the thinking, the brain largely checks out.
The memory results were just as alarming. Eighty-three percent of people who used AI to write their essays could not remember the content they had just produced. That makes sense — if you didn’t actually think through the ideas yourself, your brain has little reason to store them. Researchers say this points to a serious disengagement problem: AI may be doing so much of the mental work that users stop processing information at any meaningful depth.
Corporate-Funded Research Confirms the Risk
Microsoft — one of the biggest investors in AI technology — co-funded a study with Carnegie Mellon University that found its own products may be hurting how people think. The study tracked 319 office workers and found they put in less mental effort on tasks like comprehension, analysis, and applying knowledge when using generative AI tools. Workers who trusted AI the most showed the sharpest drop in critical thinking. That’s a striking finding, especially coming from a company that profits from AI adoption.
Michael Gerlich, a researcher at the Swiss School of Business, ran a separate study with 666 people in the United Kingdom. He found that people who used AI tools most often scored lower on critical thinking tests than older adults who rarely used them. That age comparison matters. Older adults — who grew up without these tools — outperformed younger, frequent AI users on the very skills that help people solve problems, evaluate arguments, and make sound decisions.
The “Brainrot” Warning Researchers Are Raising
A 2026 research paper published on the academic preprint server arXiv used the term “brainrot” to describe what happens when people over-rely on AI. The paper argues that deskilling and even addiction to AI tools are serious risks that the tech industry has largely ignored. Productivity may go up in the short term, but the long-term cost could be a workforce that struggles to think independently when AI isn’t available — or when it gets something badly wrong.
The Echovesper Research Foundry
The system we have been circling is not a single artificial mathematician. It is a governed synthetic research institution built around replaceable AI models. Its intelligence does not live in one enormous prompt or one supposedly omniscient…
— Aaron (@aaronnagy1987) July 13, 2026
Some researchers argue that AI can actually boost thinking skills when used carefully, with structure and human oversight guiding the process. That’s a fair point — a hammer doesn’t hurt you if you use it right. But the studies above show most people aren’t using AI that way. They’re handing over the thinking entirely. No major university or school system has yet required students or workers to stay “in the loop” when using AI tools. Until they do, the risk of a slow mental checkout — one essay, one work task, one decision at a time — remains very real.
Sources:
insiderpaper.com, chosun.com, polytechnique-insights.com, raconteur.net, cmu.edu, medium.com
