ChatGPT has become part of the daily lives of millions of people for writing, searching for information, or saving time at work. But by delegating certain tasks to artificial intelligence, could our brains end up working less? A four-month study conducted at MIT provides an initial answer.
In just a few years, ChatGPT has gone from a technological curiosity to an everyday assistant. From writing an email and summarizing a document to finding a recipe idea, preparing a presentation, or simply asking a question, the opportunities to use it are numerous. The figures clearly illustrate the scale of this phenomenon. By 2026, ChatGPT is projected to have approximately 900 million active users each week and process several billion requests daily. Its uses extend far beyond the professional sphere. According to data from Incremys , a large portion of interactions now concern personal needs, such as information retrieval, writing, and practical advice.
This growing adoption is easily explained. The tool allows users to obtain results in seconds that would sometimes require several minutes, or even hours, of research or work. However, this time saving has a potential downside, which researchers at the Massachusetts Institute of Technology (MIT) wanted to explore. What happens when we entrust part of our thinking to artificial intelligence? For four months, 54 participants were asked to write several essays under different conditions. One group could use a generative AI tool, a second had access to a search engine, and the third had to work solely with their own knowledge.
With ChatGPT, the brain would be less strained during certain tasks
The rest continues on the next page
Their brain activity was studied using electroencephalography (EEG). The goal was therefore not simply to compare the quality of the resulting texts, but to observe how the brain was engaged during the writing process. And the initial findings are already raising some interesting questions. As the tool takes over part of the intellectual work, the user seems less engaged in the task. The differences observed between the three groups are particularly interesting. Participants who wrote their texts without any digital assistance showed the greatest brain engagement. Those who used a search engine fell somewhere in between, while the users of artificial intelligence displayed the lowest levels of brain activity.
In other words, the more the tool handled idea generation, formulation, and text structuring, the less certain brain connections seemed to be activated during the exercise. This obviously doesn’t mean that ChatGPT automatically “damages” the brain or that all use leads to cognitive decline. The study involved a limited number of participants, and its results still need to be confirmed by larger research. Nevertheless, it highlights a fairly intuitive phenomenon: when some of the intellectual effort is outsourced, our brain has less work to do. The researchers also observed a difference in memory and the assimilation of the produced content. Some participants who used the AI had more difficulty remembering precisely what they had just written.
Do you use ChatGPT regularly? An MIT study reveals what happens to your brain after just 4 months.
After several months of use, the effect could persist even without AI.
Against all expectations, some participants even struggled to quote a sentence from their own essay shortly after the exercise. The authors then suggest the idea of a “cognitive debt .” The principle is simple. Using a tool to temporarily lighten a task can be very effective, but systematically delegating to it the steps that require searching, organizing, formulating, and memorizing could reduce cognitive investment. It is precisely because finding the right words, hesitating, correcting a sentence, or constructing one’s own reasoning requires effort that these operations contribute to the assimilation of information. One of the most surprising observations emerged when the researchers modified the experimental conditions.
After several sessions conducted with the help of artificial intelligence, some participants had to repeat the exercise without ChatGPT. One might have imagined that their level of brain engagement would be immediately similar to that of people accustomed to working alone. However, this was not observed. Their brain activity remained generally lower than that of the group that had never benefited from AI assistance. This difference suggests that how we become accustomed to performing a task could be as important as the occasional use of the tool itself. Should we therefore shut down ChatGPT and never ask it a question again? The researchers don’t go that far. These results remain preliminary and concern a limited sample. Above all, they do not allow us to conclude that four months of use causes lasting brain damage. Rather, they invite us to reflect on how we use artificial intelligence.
