Students who use AI to help them study tend to perform slightly worse in school than those who don’t, according to a recent report by the Organization for Economic Co-operation and Development (OECD). The study, based on data collected in 2025, is the first to examine AI use in education on a global scale, as the technology has become widely available and integrated into daily learning. The report, part of the OECD’s Programme for International Student Assessment (PISA), tested 15-year-old students in science, math, and reading across 91 countries, drawing data from over 760,000 students. After accounting for factors like family income, students who did not use AI generally scored higher in science than those who did. However, the relationship between AI use and academic performance is not straightforward. Students who used AI for specific tasks, such as summarizing texts or drafting essays, tended to perform worse than those who didn’t. In contrast, students who used AI for broader learning purposes—like conducting preliminary research or seeking general help—showed smaller declines in performance. The frequency of AI use also played a role: students who used AI once or twice a year, or daily, tended to perform worse than those who used it weekly or monthly. Interestingly, students who used AI weekly for general learning purposes outperformed non-users, though only slightly. The study also found that students who were regularly asked to assess the reliability of AI-generated information showed better results. This suggests that teaching students to critically evaluate AI tools can help them benefit more from their use. The report’s authors argue that "moderate and intentional use of AI for schoolwork and learning" could lead to positive outcomes, especially when paired with critical thinking skills. The study also highlighted a disparity in AI access among students. AI use was more common among students from more advantaged backgrounds, likely due to better access to devices, internet, and paid AI tools. Internationally, AI use varied significantly: over 95% of Vietnamese students reported using AI tools, compared to just 60% in Japan. Despite this, AI users tended to be more curious about learning, with daily users showing the highest levels of curiosity. However, this curiosity does not always translate into better academic performance, suggesting that the impact of AI on learning is complex and multifaceted.