OECD PISA 2025: Daily AI Use for Homework Linked to Lower Scores

The OECD's PISA 2025 study found that students using AI tools daily for homework scored an average of 481 in science, compared to 509 for those who never or almost never used them, a gap equivalent to about 1.5 years of schooling. The study covered over 760,000 students from 91 countries, marking the first assessment after generative AI became widely accessible. Weekly use was associated with better performance than daily use, and teacher-guided critical evaluation narrowed the gap.

This research directly impacts the design and use of AI education tools, suggesting that unmonitored use may undermine learning, while moderate, guided use can be beneficial. For teachers, it underscores the importance of integrating critical thinking into classroom practice; for product developers, it highlights the need to design tools that support active learning and teacher oversight rather than merely providing answers. Teachers need to incorporate critical evaluation of AI outputs to enhance learning outcomes. AI education products should avoid encouraging passive answer-seeking and instead support hands-on learning and active thinking.