OECD's PISA finds moderate AI use tied to higher science scores

The OECD's latest PISA assessment arrives at a moment when schools are still wrestling with whether to embrace or restrict AI chatbots. The report itself notes that reading, mathematics and science scores have been dropping since 2003, so AI is not the sole culprit behind declining results. What the data adds is a more granular picture: 86% of students across OECD countries have used AI chatbots for at least one schoolwork purpose, and about one in five use it almost every day. Within that group, the report describes a feedback loop—students who struggle to read critically may find it harder to spot incomplete or incorrect AI answers, while routinely letting AI handle reading, summarizing or reasoning could deprive them of chances to build those very skills.

The report's own framing is careful. It states that the relationships do not necessarily imply a negative impact of AI on science performance, and that moderate, intentional use may be positively related to student performance. The analogy it offers—that learning does not occur through consumption of content but through a productive cognitive struggle with new material—suggests the concern is less about AI itself than about how it is used. The finding that weekly users performed about the same as nonusers after accounting for socioeconomic status reinforces that the difference may lie in purpose and moderation rather than in access alone.

What the outcome ultimately hinges on is whether schools can translate this nuance into practical guidance. If the decline in skills like evaluating information and making connections across multiple sources continues—and the report calls this decline 'most troubling'—the pressure will be on educators to design AI use that supports cognitive struggle rather than replaces it. The test is whether guidance can steer students away from unguided daily reliance without discarding the moderate use linked to stronger performance.