AI Makes Your Brain Chill Out (But Maybe Too Much): Why ChatGPT Users Forget Their Own Essays

Overview

Paper Summary › Explain Like I'm Five › Conflicts of Interest › Identified Limitations › Rating Explanation › Good to know › Topic Hierarchy › File Information ›

Paper Summary

Paperzilla title

This study utilized EEG and NLP to show that using LLMs for essay writing leads to significantly reduced brain activity related to memory, critical thinking, and creativity, compared to using search engines or no tools. Participants relying on LLMs also demonstrated lower essay ownership and a diminished ability to recall what they had just written, suggesting an accumulation of "cognitive debt." While LLMs provided initial efficiency, over multiple sessions, LLM users performed worse overall in neural, linguistic, and scoring metrics. A key limitation for the strongest claims is the smaller sample size for the fourth session.

Explain Like I'm Five

When you use AI like ChatGPT to write, your brain doesn't have to work as hard to think up ideas or remember things. This makes writing seem easier, but it might also make your brain less good at deep thinking and remembering over time.

Possible Conflicts of Interest

Dr. Nataliya Kosmyna, the corresponding author, holds a Visiting Researcher position at Google as of the publication date (June 2025). Although the work was completed prior to her affiliation, Google has significant vested interests in AI/LLM development.

Identified Limitations

Small Sample Size for Session 4

Only 18 participants completed Session 4, which is crucial for the "LLM-to-Brain" and "Brain-to-LLM" comparisons, making findings from this session preliminary and less robust for generalization.

Limited Generalizability of Participants

The study recruited participants from a specific geographical area (greater Boston) and five academic institutions, primarily students, which may not represent the general population or diverse professional backgrounds.

Single LLM Model Used

The study exclusively used OpenAI's ChatGPT. Results may not be directly generalizable to other LLM models with different architectures, training data, or interaction paradigms.

No Subtask Differentiation

The essay writing task was not broken down into subtasks (e.g., idea generation, drafting, revising), which limits the ability to analyze specific cognitive processes affected at different stages of writing.

EEG Spatial Resolution Limitations

While EEG provides excellent temporal resolution, its spatial resolution is limited, making precise localization of deep cortical or subcortical brain activity challenging.

Context-Dependent Findings

The results are focused on essay writing in an educational context and may not generalize to other tasks or professional settings.

Preliminary Nature of Findings

The paper is a preprint and under review, and explicitly states that some findings are preliminary and require a larger sample to confirm.

Rating Explanation

The study employs rigorous methodologies (EEG, NLP, human/AI scoring) and addresses a highly relevant topic regarding AI's cognitive impact. However, the sample size for the crucial Session 4 is quite small (n=18), and the reliance on a single LLM (ChatGPT) limits the generalizability of some findings. The claims regarding "cognitive debt" are significant but based on potentially preliminary data. The stated conflict of interest, while noted as prior to work, still warrants consideration given the company's strong ties to the LLM industry.

Good to know

This is the Starter analysis. Paperzilla Pro fact-checks every citation, researches author backgrounds and funding sources, and uses advanced AI reasoning for more thorough insights.

Explore Pro →

Topic Hierarchy

Domain: Life Sciences

Field: Neuroscience

Subfield: Cognitive Neuroscience

File Information

Original Title: Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task△

Uploaded: February 07, 2026 at 07:24 PM

Privacy: Public