The brain you outsource: What AI does to students’ thinking, and the order that protects it

By Nandini Bhasin, Student of Class XII, DPS, Pune

Every high schooler knows the moment. It’s 11:40 p.m., the essay is due at midnight, and a chatbot is one tab away. The output is clean, the grammar is flawless, and the deadline is met. The only thing that went missing is the thinking.

AI isn’t the villain here. But the evidence is piling up that when students use it as a replacement for thinking rather than a tool for it, they pay a cognitive price that grades don’t show.

AI use among teens is no longer a niche habit. Pew Research Center found that 26% of U.S. teens ages 13 to 17 used ChatGPT for schoolwork in 2024, double the 13% recorded in 2023. A later Pew survey reported that more than half of teens (54%) used chatbots to help with schoolwork in 2025, and about 1 in 10 said they used them for all or most of it. Teens are also somewhat aware of the tradeoff: the most common reason teens gave for expecting AI to hurt society was overreliance and loss of critical thinking or creativity.

The strongest evidence comes from an experiment at a high school in Turkey, run by researchers at the University of Pennsylvania’s Wharton School. It involved nearly 1,000 high school math students. Students practiced with one of three setups: a standard ChatGPT-style assistant, a tutor version designed to protect learning, or no AI at all.

The results are a plot twist. Students using the standard chatbot improved 48% during practice, but the researchers concluded that students tried to use GPT-4 as a “crutch,” and when that worked, they performed worse on their own. In other words, the practice scores looked great and the closed-book exam told the truth. The study also found that students who used AI were overly optimistic about how much they had learned, even high achievers. That means you can feel like you’re getting smarter while your actual skills stall. The tutor version, which was built to guide students instead of handing over answers, largely reduced these negative effects, so how AI is used matters a lot. The work was later published in the Proceedings of the National Academy of Sciences.

Researchers at the MIT Media Lab put EEG headsets on writers to compare three groups: ChatGPT, search engine, and no tools. The ChatGPT group showed the lowest brain engagement and underperformed on neural, linguistic, and behavioral measures, often resorting to copy-and-paste by the end. In a later session, participants who switched from ChatGPT to writing alone showed weaker neural connectivity, while those who started without AI and then added it did better on recall.

The order mattered: think first, use AI second. The paper was a preprint, not yet peer reviewed, and the authors called the conclusions preliminary. They also asked the media to stop using terms like “dumb” and “brain rot.” Only 18 of the 54 participants completed the final session, and the participants were university students, not high schoolers. Treat it as an early warning, not a verdict.

Creativity is where AI’s effect is trickiest. A study in Science Advances by researchers at UCL and the University of Exeter found that AI-generated ideas made stories rated as more creative and better written, especially for less creative writers. However, those AI-assisted stories were more similar to each other than stories written by humans alone. The authors describe it as a social dilemma: each writer is individually better off, but collectively a narrower range of novel content is produced.

Picture a whole grade turning in essays shaped by the same model. Everyone sounds polished, and nobody sounds like themselves. Original voice is the thing colleges, employers, and audiences actually remember.AI makes work faster, and that’s exactly the problem when the point of school is the struggle.

The goal isn’t to log off. It’s to make sure the brain doing the thinking is still yours.