Teenagers who write their own notes understand and remember schoolwork better than those who rely only on generative AI tools such as ChatGPT, according to new research from Cambridge University Press & Assessment and Microsoft Research.
The study examined how large language models affect learning in real classroom conditions. It focused on reading comprehension and memory rather than on essay production or homework completion.
The researchers worked with 405 pupils aged 14 and 15 in seven secondary schools across England. The sample included comprehensive, grammar and private schools.
Students studied two texts from the national History curriculum. One text covered apartheid in South Africa. The other described the Cuban missile crisis.
All students used an AI chatbot based on ChatGPT 3.5 turbo for one of the texts. They then learned the other text using different approaches that involved note-taking.
One group studied the second text only by writing their own notes. The other group combined their own note-taking with use of the AI system.
In the AI conditions, students received a short tutorial on how to use the tool. They then interacted with it in any way they chose.
Three days later, teachers asked the pupils questions about both topics without prior notice. The questions tested how well students understood and retained the information in the texts.
Examples included factual questions such as what happened at the Soweto Youth Uprising in 1976. Other questions required explanation, such as the role of the Soviet Union in the Cuban missile crisis.
After each study session and after the test, pupils also answered questions about their experience of the tasks. They reported whether they enjoyed the work and whether they found it interesting.
The results showed that students who wrote notes performed better on comprehension and memory than those who used the AI tool alone. Outcomes were also stronger when students combined note-taking with AI use than when they relied only on AI.
The findings indicate that the act of writing notes still plays a central role in how teenagers process and retain information. AI chatbots offered other benefits that related more to engagement and exploration of topics.
Many students reported that they enjoyed using the AI system. They also said it made it easier to look up extra information around the core text.
Dr Pia Kreijkes, Senior Researcher at Cambridge University Press & Assessment, led the study. She said the project responds to rapid growth in student use of tools such as ChatGPT and Microsoft Copilot.
“We know that students are using chatbots and other AI tools, including to help them with their schoolwork. However, there has been very limited research on how LLM use influences students’ ability to understand and remember information. Our study shows that students enjoyed using AI chatbots but note taking was more effective for learning outcomes.”
“Our findings can help guide the use of LLMs for learning. In particular, they indicate that students should take notes separately from using LLMs to avoid copying the LLM. They also indicate that students should receive training and guidance on how to use LLMs to support active and constructive learning,” said Kreijkes.
She said teachers could gain insights from how pupils use AI systems during study.
“Teachers could also benefit from their students’ use of LLMs. For example, in the future, teachers may be able to leverage insights from students’ LLM interactions to understand where support is needed and tailor class materials accordingly,” said Kreijkes.
Dr Jake Hofman, Senior Principal Researcher at Microsoft Research, examined how students interacted with the chatbot during the lessons.
“I was struck by how many students used the LLM to try to deepen their understanding - asking about historical context, clarifying unfamiliar references, and exploring the significance of key events. Rather than viewing traditional learning techniques, like note-taking, and newer generative-AI approaches as competing alternatives, we should view them as complementary.”
The experiment appears in the journal Computers & Education. The authors describe it as one of the first randomised classroom studies of large language models in secondary education.
The work highlights a gap between how students use AI tools and what is known about their effects on learning. Many pupils already turn to chatbots outside class for explanations, summaries and homework support.
The authors report that AI tools played a useful role when students used them to ask questions about context, vocabulary and historical significance. The study found that this use worked best when students also created their own written notes.
The research suggests that schools and parents may need to distinguish between enjoyment and engagement on the one hand, and learning outcomes on the other. It points to a need for guidance on when students should step away from an AI interface and work with pen and paper or a digital notebook.
The team plans further work on how different prompts, training and classroom designs affect learning when students use AI alongside traditional study methods. They also intend to explore how teachers might use anonymised chatbot logs in future to shape lesson plans and revision materials.