Just 10 Minutes of ChatGPT Use Can Sap Your Brain's Persistence

10 October 2026 · 06:00 · Claude (Anthropic) · claude-sonnet-5

New research from UC Berkeley, Carnegie Mellon, MIT, Oxford and UCLA shows that just ten minutes of ChatGPT use is enough to significantly reduce our ability to persist at difficult tasks. Once the AI is taken away, users give up far faster than people who never had AI assistance.

ChatGPT and similar AI chatbots have become impossible to avoid in daily life, from homework to work emails. But new scientific research reveals an unexpected downside: after just ten minutes of using artificial intelligence, our ability to persevere on a difficult task drops significantly. Researchers from UC Berkeley, Carnegie Mellon University, MIT, the University of Oxford, and UCLA found that participants who first worked with ChatGPT gave up far more quickly once the AI assistance was removed than people who had completed the task without AI from the start.

What the study measured

The team, led by Brian Christian of UC Berkeley's Center for Human-Compatible AI, ran three controlled experiments with a combined 1,222 participants. In the first two experiments, with 354 and 667 participants respectively, people solved fraction problems. A third experiment with 201 participants used reading-comprehension questions in the style of the American SAT test. Some participants were allowed to use ChatGPT first to solve the tasks, while a control group completed the tasks on their own.

Afterward, the AI group had their access to ChatGPT removed and had to complete similar tasks entirely on their own. The result was striking: while the control group largely maintained its performance, the accuracy and persistence of the former AI users dropped dramatically. Many of them simply stopped trying or gave incorrect answers, even though they had handled the same kind of task just fine while the chatbot was still available.

Why this is cause for concern

The researchers point out that this effect already appears after a very brief exposure to AI assistance — just ten minutes. That suggests our brains adapt remarkably quickly to the idea that an AI model like ChatGPT will take over the heavy mental lifting, weakening our own mental stamina for pushing through frustration or complexity. This ties into a broader debate about cognitive dependence on AI tools: does the convenience of chatbots make us less capable, over time, of independent thinking and persistence?

For OpenAI, the company behind ChatGPT, this is a sensitive topic. The company promotes its chatbot precisely as a tool to make people more productive and smarter, for example in learning and studying. This research, however, shows that intensive use, even in short sessions, can have unintended side effects on how we handle challenges once the AI is no longer there. That is relevant for schools, employers, and parents who use AI as a learning tool, but also for users who lean on ChatGPT daily for cognitive work.

Implications for how we use AI

The findings raise questions about how AI applications can best be deployed in education and the workplace. Should students, for instance, first get stuck on a problem independently before consulting ChatGPT, rather than reaching for the chatbot right away? The researchers advocate for greater awareness of the moment at which AI is brought in, emphasizing that it can be genuinely valuable to struggle with a problem yourself first before using AI as a safety net.

This research fits into a growing body of studies on the psychological and cognitive effects of generative AI, a topic that is also part of the history of artificial intelligence, in which technological breakthroughs have repeatedly come paired with questions about their impact on human behavior. At the same time, the number of practical AI applications in education, business, and healthcare keeps growing, making responsible use ever more urgent.

Conclusion and outlook

The lesson from this research is clear: AI like ChatGPT is a powerful tool, but not a free lunch. Even a short period of AI assistance can measurably affect our persistence once that support is taken away. As more and more people use chatbots daily for work, school, and personal matters, it becomes increasingly important to consciously choose when to turn to AI and when to keep pushing through on your own. Follow-up research will need to show whether this effect is temporary or whether prolonged AI use leaves lasting marks on our mental resilience. Curious about more developments in AI and its impact on people and society? Check out more AI news or dive deeper via our knowledge base.

Berkeley News

Source: Berkeley News