Overview

• An MIT Media Lab preprint measuring EEG in 54 students showed that those writing essays with ChatGPT exhibited weaker and less widespread neural connectivity than peers using Google Search or no online tools.
• A University of Pennsylvania Wharton School study of over 4,500 participants found ChatGPT users provided less detailed and original gardening advice than standard search users, signifying more passive learning.
• When habitual ChatGPT users switched to manual writing, their neural engagement increased but their essays became less clear, suggesting that the timing of AI integration affects both brain activity and output quality.
• Researchers warn that overreliance on AI assistants can lead to cognitive off-loading and memory weakening, prompting calls to pair digital tools with pen-and-paper tasks to sustain deep learning.
• Critics argue the MIT experiment’s small sample size and lack of peer review limit its conclusions and stress the need for larger, rigorously validated studies on long-term AI impacts.