OECD Report Shows Students Using AI Daily Underperform Non-Users

A broad OECD report analyzing PISA data from over 760,000 students across 91 countries revealed that students who use AI chatbots daily for schoolwork scored 28 points lower on science tests than non-users. This performance gap persists even after controlling for socioeconomic background, representing approximately one and a half years of schooling. As the largest empirical study on generative AI's impact in education since ChatGPT's launch, the findings warn educators and policymakers that passive reliance on AI can bypass crucial cognitive effort. However, the report also shows that controlled, moderate use combined with critical thinking can actually enhance student achievement. Only 14% of students surveyed reported never using AI for schoolwork, with wide country variations ranging from nearly 40% in Japan to just 4% in Vietnam. Notably, students who used AI once or twice a week to assist learning or were trained to evaluate AI-generated outputs performed better than both daily users and complete non-users.