MIT study finds that ChatGPT is making people dumber; 83% of ChatGPT users can’t recall what they just wrote

A new study out of MIT is raising serious questions about what heavy ChatGPT use might be doing to our brains.

Researchers found that most people who rely on the AI tool for writing are struggling to stay mentally engaged, and many can’t even remember what they wrote minutes later.

The paper, titled “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task”, tracks how the human brain behaves when it leans too heavily on AI. The results? A drop in neural activity, memory gaps, and a loss of original thought.

AI Amnesia: ChatGPT Is Making You Dumber

What the MIT Study Found

MIT researchers studied how using large language models (LLMs) like ChatGPT affects the brain and writing behavior. Participants were split into three groups: one used an LLM, another used a search engine, and the third wrote without tools. Over four sessions, they completed essay tasks while their brain activity was tracked using EEG.

Key findings:

• Brain-only participants had the strongest neural activity and best memory recall.

• LLM users showed the weakest brain connectivity and struggled to remember their own writing.

• When LLM users switched to writing without AI in the final session, their brain engagement dropped even further.

• People who started without tools and then used LLMs showed better cognitive performance than full-time AI users.

• LLM users reported the lowest sense of ownership over their writing.

AI vs. Human Brain: MIT Findings

The experiment followed 54 participants—mainly college students in the Boston area—as they completed four writing sessions. One group used ChatGPT-4o, another used Google Search, and a third wrote without any assistance.

During the sessions, the researchers used EEG headsets to monitor brain activity across 32 regions. The essays were scored by teachers and AI judges. But the real story came from what was happening inside participants’ heads.

By the fourth session, the groups switched roles—ChatGPT users had to go without it, while the “Brain-only” group got to try it for the first time. That twist revealed just how dependent some users had become on AI.

“While LLMs offer immediate convenience, our findings highlight potential cognitive costs. Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels. These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI’s role in learning,” MIT wrote in the paper.

The Warning Signs

The numbers are hard to ignore:

• Brain activity fell off a cliff. ChatGPT users showed significantly less neural connectivity, especially in alpha and theta waves—signals tied to memory, creativity, and deep thinking. In fact, the number of active neural connections dropped from 79 to 42.

• Memory all but vanished. A staggering 83% of ChatGPT users couldn’t recall what they wrote minutes earlier. In contrast, just 10% of the non-AI and Google Search users had similar recall issues.

• Writing lost its spark. Teachers called the AI-assisted essays “soulless” and “bland.” Many users copied and pasted the AI’s output without editing, resulting in flat, generic writing.

• AI dependence came at a cost. When the ChatGPT group had to write without it, their brains showed signs of cognitive fatigue. They struggled to engage and had weaker performance overall.

• People felt disconnected from their own work. Users who relied on ChatGPT reported lower satisfaction with what they produced and felt less ownership over their writing.

What’s Going On?

The research team says this is about more than convenience—it’s about offloading mental effort. When people start a task with AI from the get-go, they tend to stop thinking deeply. Planning, analyzing, revising—those steps fade away.

Lead researcher Nataliya Kosmyna summed it up bluntly: “The task was executed, and you could say that it was efficient and convenient. But you basically didn’t integrate any of it into your memory networks.”

Credit: MIT Research Lab

The EEG data backs that up. ChatGPT users showed lower activity in brain areas involved in decision-making and long-term memory. Meanwhile, the Brain-only group lit up those regions, suggesting deeper cognitive involvement.

Interestingly, the Search group sat in the middle. While search engines still offer outside help, they require users to think more, formulating questions, scanning results, and evaluating sources. ChatGPT, on the other hand, gives a clean output with minimal effort.

What This Means for Schools and Workplaces

The implications go beyond academia. Students who rely too heavily on AI might see their writing skills deteriorate. And in the workplace, teams banking on AI for reports or communication could be short-circuiting their own creativity.

One startup founder put it bluntly: “AI isn’t making us more productive, it’s making us cognitively bankrupt.”

That’s not hyperbole. The study found that over just four months, ChatGPT users saw consistent dips in memory, engagement, and originality. And with a recent UK survey showing 88% of students are now using AI for assessments, the trend is clearly growing.

It’s Not All Bad—When Used the Right Way

To be clear, the study doesn’t say ChatGPT is inherently harmful. When the Brain-only participants got to use it in the fourth session, their brain activity spiked. That group already had a strong foundation—they used AI to build on their own ideas, not replace them.

Even ChatGPT itself responded with some self-awareness. Asked about the study, it replied: “This study doesn’t say ChatGPT is inherently harmful—rather, it warns against overreliance without reflection or effort.”

Used strategically, the tool can boost productivity and creativity. But when people hand over the entire process from the start, they risk losing the mental muscles needed to think independently.

The Study’s Limits—and What Comes Next

The sample size was small—just 54 people from the same metro area—and focused only on essay writing. It didn’t explore how AI affects other tasks like coding or creative arts. And while the paper hasn’t yet been peer-reviewed, its methodology was grounded in real brain data, not just surveys.

Kosmyna’s team is already running follow-ups, including a new study on coding tasks. Early signals suggest the cognitive effects might be even more pronounced there.

They’re also exploring ways to track how different usage patterns—like brainstorming with AI vs. copying its output—impact long-term thinking.

Bottom Line

The takeaway isn’t “Don’t use ChatGPT.” It’s: don’t let it think for you.

Start with your own ideas. Use AI as a sidekick, not a crutch. Because the cost of convenience might be sharper than we think.