ChatGPT Can Speed Up Studying, New Research Suggests, Yet Long-Term Memory May Pay the Price

A new study suggests students who rely heavily on ChatGPT may finish assignments faster, but remember less weeks later.

The findings add to growing evidence that AI tools can encourage cognitive offloading, shifting effort away from the mental work that supports durable learning.

The research, led by AI specialist André Barcaui at the Federal University of Rio de Janeiro, followed 120 university students completing a short course assignment about artificial intelligence. Half could use ChatGPT for research, explanations and structure, while the rest used conventional study methods.

Participants were later given an unexpected test 45 days after completing their presentations. Students who had used ChatGPT averaged 5.75 out of 10, compared with 6.85 for those who studied without the chatbot, a gap the paper attributes to reduced cognitive effort during learning.

Faster work, weaker recall

Time savings were clear in the experiment: the ChatGPT group spent about 3.2 hours on the task, versus 5.8 hours for the non-AI group.

But the stronger long-term scores clustered more tightly among traditional learners, while results for chatbot users varied more widely.

Barcaui wrote that “unrestricted ChatGPT use impaired long-term retention,” likely because it removed some of the productive struggle that strengthens memory. The study frames this in line with desirable difficulties, the well-established idea that learning can improve when it feels harder.

How this fits earlier findings?

The results echo earlier research on the Google effect, described in 2011 by psychologist Betsy Sparrow and colleagues, showing people are less likely to remember information when they expect to be able to look it up later.

AI assistants can amplify that dynamic by delivering polished summaries that reduce active processing.

The author argues the goal should not be banning chatbots, but designing teaching that keeps students cognitively engaged while using AI for support. The paper calls for classroom strategies that preserve critical thinking, verification and effortful recall, rather than turning ChatGPT into a default shortcut.

The study is limited by its single setting, moderate sample size and focus on one type of task, and it does not prove the same effect will appear in every subject or age group. Still, as schools and workplaces adopt generative AI, it adds timely evidence that how people use these tools may matter as much as whether they use them.