Study finds excessive AI use linked to poorer student performance in academic activities

Students who rely on AI chatbots for academic work may perform worse than those who complete their studies without them, according to a new OECD report.

The report by the Organisation for Economic Co-operation and Development (OECD) examines the growing use of artificial intelligence in education and its potential impact on students’ ability to learn and retain information.

Researchers compared the science test scores of students who did not use AI chatbots with those of 484 students who used the technology for academic tasks. The results showed a notable performance gap: students who stayed away from AI recorded an average score of 509, while those who used chatbots scored 481 on average.

The 28-point difference remained consistent over a period of 18 months, suggesting that the performance gap was not simply a short-term effect.

Similar patterns emerged when students used AI for tasks such as researching unfamiliar subjects or preparing summaries after studying.

Researchers stressed that effective learning is not simply about memorising information. Instead, engaging with challenging material and making the mental effort to understand it can play an important role in developing deeper learning and cognitive skills.

The report does not suggest that AI should be completely excluded from education. Instead, researchers argue that the technology should be used as a temporary support rather than a substitute for students’ own thinking and learning.

They warned that excessive dependence on AI could prevent students from developing the cognitive skills needed to succeed academically.

The findings come as AI chatbots become increasingly common in classrooms and among students, raising fresh questions about how the technology can be integrated into education without undermining independent learning.

Researchers had also conducted similar work in 2022, when AI technology was beginning to gain widespread popularity. The latest findings add to growing concerns over whether convenience provided by AI could come at the cost of deeper learning.