Are Your Cognitive Skills Affected by ChatGPT Influence?

Key Takeaways:

Brain activity declines with AI assistance: EEG data showed that students using ChatGPT engaged fewer neural networks related to memory and attention, compared to those working unaided.
Poor Memory and Quoting Accuracy: Most ChatGPT users had difficulty quoting from their own essays shortly after writing them, while students in other groups performed significantly better.
Essays lacked originality and ownership: Essays generated with ChatGPT had a similar structure and tone. Students also reported feeling less connected to their work.
Researchers urge cautious integration of AI in education: The study recommends pairing AI tools with teaching methods that promote cognitive engagement, critical thinking, and memory retention.

A new study from researchers at the Massachusetts Institute of Technology (MIT) warns that using generative AI tools, like ChatGPT, to write essays may reduce memory, weaken brain connectivity, and impair deeper learning, even if those essays receive high grades.

The four-month study, conducted with 54 students from Boston-area universities, found that participants who relied on AI assistance demonstrated lower neural engagement and a weaker ability to recall or feel ownership over their written work.

“While the benefits were initially apparent,” the researchers wrote, “the LLM group’s participants performed worse than their counterparts in the Brain-only group at all levels: neural, linguistic, scoring.”

Study Setup: Essays, AI, and EEG Scans

Participants were divided into three groups: one using ChatGPT-4, another using only search engines such as Google, and a third writing with no tools at all. Each wrote three essays over three sessions. A fourth session reversed the roles, where ChatGPT users went unaided, while Brain-only participants were given access to the AI.

Throughout the sessions, participants wore EEG headsets that recorded real-time brain activity, while researchers analyzed the essays with natural language processing tools and interviewed students after each session.

The essays were scored by both human teachers and an AI model built for evaluation.

Neural Engagement Drops With AI Use

The study found significant differences in brain activity among the groups. Students using no tools exhibited the most extensive and interconnected neural patterns, while those using ChatGPT showed the weakest.

“Brain connectivity systematically scaled down with the amount of external support,” the study stated. “The Brain‑only group exhibited the strongest, widest‑ranging networks, the Search Engine group showed intermediate engagement, and LLM assistance elicited the weakest overall coupling.”

Using ChatGPT May Lead to Shallow Thinking

Students who used ChatGPT were found to rely more on surface-level information and showed lower originality. Their essays scored well, but exhibited less linguistic diversity and creativity.

“Though scored high by both AI judge and human teachers,” the paper noted, “their essays stood out less in terms of the distance of NER/n-gram usage compared to other sessions.”

This pattern suggests a tendency to offload thinking to the AI.

Additionally, many essays followed predictable patterns, and the most common choice of words aligned closely with ChatGPT’s default phrasing.

AI-Written Essays All Sounded the Same

The study’s Natural Language Processing (NLP) analysis showed that essays written with ChatGPT shared similar structural and topical elements.

Participants writing without AI engaged more freely with the essay prompts, showing greater variation in phrasing and approach.

In contrast, the LLM essays clustered tightly in semantic space, echoing the model’s training tendencies.

“We discovered a consistent homogeneity across the Named Entities Recognition (NERs), n-grams, [and] ontology of topics within each group,” the researchers wrote. This effect was strongest in the LLM group.

AI Users Could Not Remember Their Own Writing

Students who used ChatGPT to write essays showed clear signs of weaker memory and quoting ability compared to those who worked unaided or used search engines.

In the first session, most AI users struggled to recall lines from their own work, and none were able to quote accurately. This difficulty persisted through later sessions, with several still unable to quote correctly by the end.

Impact of AI on Deep Memory Encoding

Researchers linked this memory gap to reduced activity in specific brainwaves, especially theta (4–8 Hz) and alpha (8–12 Hz), which are essential for memory consolidation and semantic encoding. These brainwaves are typically strongest when individuals actively generate and structure content themselves.

In the AI-supported group, this activity was significantly weaker, suggesting that participants were reading and selecting tool-generated suggestions without deeply processing them.

This lack of engagement likely prevented the content from being stored in long-term memory networks.

In contrast, students in the Brain-only group showed much stronger theta and alpha connectivity. Their essays reflected higher mental effort and deeper integration of ideas, supported by active parietal-frontal and temporal-frontal pathways associated with attention and memory processing.

Students Felt ChatGPT Wrote the Essay, Not Them

Participants using ChatGPT were less likely to feel the essay was their own. Some said it felt more like “ChatGPT’s essay than mine.”

In contrast, students in the Brain-only group overwhelmingly reported full ownership. One Brain-only participant said they felt “personally connected” to their arguments, while an LLM user described their writing as “copy-pasted, with edits.”

The study also tracked essay satisfaction. All participants in the Search group said they were satisfied with their writing. That number dropped slightly in the LLM and Brain-only groups.

Session 4: Group Reversals Expose AI’s Limits

The fourth session allowed researchers to study cognitive switching. Participants who started without tools and then used ChatGPT (Brain-to-LLM) showed improved scores and maintained strong neural engagement. Those who switched from ChatGPT to unaided writing (LLM-to-Brain) showed significantly reduced engagement.

“Connectivity was significantly lower than the peaks observed in Sessions 2, 3… but remained above Session 1,” the study noted.

The authors interpreted this as a sign of cognitive efficiency adaptation, but also potential long-term impact from prolonged reliance on AI.

Emotional Responses: Less Fulfillment, More Doubt

In interviews, students voiced mixed feelings about using AI. While many in the LLM group appreciated the speed and ease, others felt disconnected from their essays.

One said, “I would rather use the Internet over ChatGPT as I can read other people’s ideas on this topic.”

Several noted that ChatGPT’s tone was “robotic” and “dry.” One participant added that using the AI gave them “analysis-paralysis” and made it harder to commit to ideas.

What This Means for Schools and Learning

The researchers did not directly say that AI should be banned in education. However, they warned against using it without limits. They suggested that AI should help students, not replace their thinking.

“We hope this study serves as a preliminary guide to understanding the cognitive and practical impacts of AI on learning environments,” the authors wrote.

The team advised schools to teach students how to use AI in smart and limited ways. They suggested starting with unaided learning, then using AI as a support, not the main tool.

As AI tools like ChatGPT become more common in schools, the MIT study reminds educators to think carefully about how, when, and why they are used.

In Conclusion

The study highlights a clear cognitive divide between students who used AI, in this case, ChatGPT, and those who did not. While AI tools helped improve writing speed and surface quality, it was linked to weaker memory, reduced brain activity, and a lower sense of ownership over the work.

Students who wrote without assistance showed stronger mental engagement and better recall, supported by higher activity in brain areas related to learning and attention.

The findings suggest that while AI can support creative tasks that include writing. However, relying on it too heavily may limit deeper learning.

Researchers recommend using AI tools with care, ensuring they complement practices that support critical thinking and retention, rather than replace the cognitive effort essential to learning.