Ever since ChatGPT came into existence, AI has reshaped how we study and work. Today, it is one of the most used AI chatbots worldwide. But, can using ChatGPT make us think less? Researchers from MIT recently conducted a study to find answers to this big question. At a time when more people are using AI tools for learning and writing, the team of researchers examined how it is impacting the brain.
The researchers from MIT Media Lab, Wellesley College, and MassArt asked students to write essays, some using ChatGPT, some with the help of Google Search, and some simply using their minds. The researchers tracked their brain activity while they were writing.
The groundbreaking study revealed that those who relied on ChatGPT used their brains less, remembered less, and were less connected to their work. The study shows that although AI tools like ChatGPT can improve efficiency, it may even be making many of us lazy.
What is the research?
The research was led by Nataliya Kosmyna and her team, and the objective was to understand how AI tools, although convenient, could actually be affecting our critical thinking, memory, and learning abilities. As part of the study, the researchers recruited as many as 54 participants from universities in the Boston area. These participants were divided into three groups – LLM group which used ChatGPT exclusively to write essays; Search Engine group that used traditional web search tools without any AI; Brain-only group which wrote essays without any kind of external assistance.
The participants completed three essay-writing sessions using their assigned group and method. In the fourth session, the LLM group wrote without tools (LLM to Brain), and the Brain-only group used ChatGPT for the first time (Brain to LLM). During each of these sessions, the researchers recorded the brain activity of the participants using EEG or electroencephalography to monitor their cognitive engagement and load. Later, the essays were analysed using natural language processing (NLP) and evaluated by human teachers as well as an AI judge.
What are the key findings?
The key findings of the research were divided into three segments: weaker brain engagement with ChatGPT, reduced memory and ownership, and essay quality vs cognitive depth. One of the most striking aspects of the findings was the difference in brain activity across the groups. Candidates from the brain-only group displayed the strongest and most complex neural connectivity. While those using search engines displayed intermediate levels, the ChatGPT group showed the weakest engagement of the brain. Meanwhile, in the fourth session, the LLM to Brain group that has been relying on ChatGPT earlier, reportedly struggled to engage the same brain regions which they once used effectively. Similarly, Brain to LLM candidates showed a spike in neural activity, mostly because they were working with a new tool.
The second most important finding is the reduced memory and ownership. When the researchers asked the participants to quote or summarise parts of their essays, candidates in the LLM group had the most difficulty. As per the study, only a few of the participants were able to recall what they had written just a few minutes ago. Further, their responses also revealed that they felt less ownership of AI-assisted essays compared to those written independently. The researchers have termed this ‘cognitive offloading’, meaning an overreliance on AI that reduces the human brain’s effort in processing and retaining information.
Story continues below this ad
When it came to essay quality and cognitive depth, essays written by using ChatGPT scored higher in terms of structure and grammar, both by human teachers and the AI judge. The research, however, showed that the essays tended to be more homogenous with repetition in terms of common phrases, named entities, patterns that were predictable. On the other hand, essays written by the brain-only group demonstrated diverse vocabulary and critical thought. The researchers found that though the AI-assisted essays appear to be refined, they lacked depth and originality, raising questions about what is valued more in education – superficial quality or genuine cognitive effort.
The biggest takeaway from the research is that LLMs like ChatGPT can offer boundless benefits, but they may lead to reduced mental engagement, weaker memory, and reduced personal connection to one’s work. The researchers paint a cautionary picture about the extensive use of AI in academic settings, especially when the objective is to nurture independent thinking of students.
The above research paper ‘Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task’ was published on June 10, 2025. The team of researchers include Nataliya Kosmyna, Eugene Hauptmann, Ye Tong Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beresnitzky, Iris Braunstein, and Pattie Maes.