AI Usage Among Students Linked to Lower Academic Performance
Students leveraging AI for their studies often see lower academic performance compared to their peers who don’t use these tools, according to findings from a global educational report. However, the dynamics are more nuanced, as certain AI applications can provide a slight advantage, particularly for those trained to critically evaluate the effectiveness of these tools.
The OECD’s Programme for International Student Assessment (PISA) gathers data from various countries every few years. This year’s report, based on data collected in 2025, marks the first comprehensive analysis since AI became widely adopted. It assesses 15-year-old students in subjects like science, math, and reading, drawing from over 760,000 students across 91 countries.
When accounting for socioeconomic factors, students who refrain from using AI generally outperformed their AI-using counterparts in science. The extent of this difference varies: students employing AI for specific tasks, such as summarizing readings or drafting assignments, tended to perform worse, while those using AI for preliminary research or general learning assistance experienced a smaller decline in performance.
The frequency of AI use also plays a significant role. Students who engaged with AI only once or twice a year, as well as those who used it daily, often had the lowest performance, whereas those who utilized AI on a monthly or weekly basis fared better.
In some instances, AI can be beneficial. Students who reported using AI weekly for general learning assistance outperformed non-users, albeit marginally. This gap widened among students who frequently evaluated the quality of AI-generated content; with this additional training, regular AI users for learning purposes surpassed non-users. The study’s authors suggest that “moderate and intentional use of AI for schoolwork and learning” may yield positive outcomes.
OECD director of learning and skills, Andreas Schleicher, emphasized that just as physical fitness requires active participation rather than mere observation, effective learning demands engagement with new material. He noted that technology can enhance this cognitive engagement, but if it bypasses the necessary struggle of learning, it may hinder student development.
The PISA study also examined the prevalence of AI use among students, revealing that it is more common among those from relatively privileged backgrounds, likely due to better access to hardware, reliable internet, and paid AI services. Internationally, AI usage varied significantly, with over 95 percent of Vietnamese students reporting use of AI tools, compared to just 60 percent in Japan.
Interestingly, AI users tend to exhibit higher levels of curiosity, peaking among daily users. However, this curiosity has not consistently translated into improved academic performance.
