Students who use AI generally score worse at school

Data from the Organisation for Economic Co-operation and Development indicates that students who rely on artificial intelligence for study purposes generally achieve lower scores than their peers who do not use such tools. The 2025 Programme for International Student Assessment reveals a clear inverse relationship between frequent AI dependence and performance in science, mathematics, and reading across participating nations. While the overall trend points to reduced academic outcomes, the report notes a specific exception where learners taught to critically evaluate AI responses show a slight improvement in results. This nuance suggests that the tool itself is not the sole determinant of success, but rather the manner in which a student interacts with it. The findings challenge the assumption that immediate access to answers equates to better learning, highlighting instead the risks of reduced cognitive effort during the study process.

The study matters because it provides the first major international benchmark for education after AI became ubiquitous in classrooms. Educators now have empirical evidence to guide policies on technology integration rather than relying on speculation. Without strict oversight, widespread adoption could systematically lower achievement standards if schools fail to teach critical verification skills.