PISA 2025 Results (Volume I): Future-Ready Students

Key Findings

• — Use of AI for schoolwork is widespread: on average across OECD countries only 14% of students reported never or almost never using AI for any of the schoolwork purposes examined, while almost-daily use remains relatively uncommon
• — Students who do not use AI chatbots for specific schoolwork tasks (summarising texts, preliminary research) outperform users in science on average across OECD countries; the report describes the patterns as complex and dependent on how generative AI is used
• — Weekly users and non-users of AI for the general purpose of helping to learn perform similarly; for task-specific use, moderate users (monthly to twice weekly) outperform both limited and frequent users
• — Students who use AI frequently for learning and who learn at school to assess the quality of AI-generated information achieve slightly higher science performance than non-users or infrequent users, but these learning opportunities are reported more often by socio-economically advantaged students
• — Reading performance fell 28 points across OECD countries between 2015 and 2025, the steepest decline of the three domains; the OECD links reading skills to evaluating information in the AI era (this figure is not about AI use)
• — Moderate use of digital devices for learning at school is associated with higher science performance than either no use or very high use, and device distraction remains widespread

Methodology Notes

PISA 2025 assessment of 15-year-olds (science the major domain), more than 760,000 students in 91 countries and economies; AI items sit in the student questionnaire (Figure I.4.11 and related tables). Cross-sectional associations only. Publication date 2026-09-08 from the OECD landing page. The OECD HTML landing page and DOI resolver block automated access, but the report PDF on the oecd.org content path serves directly; the curator downloaded the 10.3 MB PDF on 2026-09-17, confirmed the citation block and DOI, and read the Chapter 4 summary bullets on AI use. Press reports attributing a 28-point disadvantage to AI use conflate it with the reading decline.

Sources