Artificial Stupidity: AI chatbots are making users less intelligent, researchers say
19 June 2025, 08:35
A new study has warned consistent use of AI chatbots could diminish cognitive skills like memory, language and critical thinking.
Listen to this article
Loading audio...
Researchers at the Massachusetts Institute of Technology (MIT) have found that people who rely on chatbots to craft long-form texts like essays may be damaging their cognitive skills.
The MIT study found that regular users of chatbots like ChatGPT had lower brain activity than those who didn’t rely on the services.
Monitoring 54 people, the study split the cohort into three groups and asked them to write four essays.
One group was able to use ChatGPT to help them write the essays, one group was permitted to use search engines for research, and the other was only able to rely on brainpower.
Read More: OpenAI rolls back ‘annoying’ ChatGPT update, after chatbot overly flatters its users
Read More: Norwegian man calls for fines after ChatGPT ‘hallucinated’ that he’d killed his children
The ChatGPT cohort, permitted to use AI chatbots for three out of four essays, were then asked to write the fourth essay solely on brainpower.
The participants were hooked up to electroencephalogram (EEG) brain activity scans, and were asked questions about their essays while they worked.
The group who used ChatGPT performed notably worse than the other cohorts, and performed worse in all measured neural categories.
The essays produced by the ChatGPT cohort were also deemed to be of worse quality than the other cohorts’, being described as “homogenous”, and featuring “repeating language and themes”.
In their final essay written just on brainpower, the ChatGPT cohort also demonstrated a limited ability to retain learned information, as well as a diminished ability to brainstorm and problem solve.
The study found that brain activity proportionally scaled down with the amount of “external support” received while working.
Those using AI chatbots, which largely take the labour of writing and research out of the writer’s hands, had the weakest levels of brain activity.
Search engine research, on the other hand, didn’t pose much of a negative effect on users, as it still required the user’s own comprehension.
Notably, the ChatGPT cohort displayed significant lower “theta” brainwaves, which are associated with learning and high level cognitive activity in the brain.
The study noted that “human thinking” and “planning” skills were largely offloaded by the use of AI chatbots.
After writing their essays, the participants were then quizzed on what they had written.
When asked to recite any part of their essays, 83% of the ChatGPT cohort failed to recite a single line from any of their essays. This is compared to 10% of those who had written their essays solely on brainpower, or with the assistance of search engine research.
This study comes amid growing concern that over-reliance on AI is weakening human cognition and critical thinking.
A recent study from Microsoft and Carnegie Mellon warned of a cognitive “atrophy” that could happen as a result of AI overuse.
Researchers have said that overuse of AI could lead to a “cognitive debt”, and could pose long-term costs like increased vulnerability to manipulation and decreased creativity.
The study also comes amid growing criticism of AI chatbots' accuracy, as reported "hallucinations" have become more common.
Paired with users' potentially diminished ability to critically detect false information, researchers worry this combination could create a perfect storm of false information.
