AI in the Classroom: OECD Finds Mixed Results on Student Performance

However, combining AI usage with training to evaluate its effectiveness can lead to improved test scores.

According to a global OECD educational report, students who utilize AI for studying tend to achieve lower grades than their non-using peers. The reality is more nuanced, as specific AI applications can provide a slight advantage, particularly when students are taught to critically evaluate the performance of these tools.

The OECD’s Programme for International Student Assessment (PISA) collects data from nations worldwide on a recurring basis. The 2025 study, based on data gathered that year, marks the first assessment since AI adoption became widespread. It evaluates 15-year-olds in science, mathematics, and reading, encompassing a sample of more than 760,000 students across 91 countries.

After accounting for socioeconomic status, students who abstain from AI generally outperformed frequent users in science. The severity of this decline varies based on specific applications; students who employed AI for summarizing texts or drafting assignments tended to perform worse, whereas those who used it for preliminary research or broader learning assistance experienced a smaller performance dip.

The frequency of usage also plays a significant role, with students using AI only once or twice a year or daily showing the poorest results, while those utilizing it monthly or weekly achieved better outcomes.

In certain instances, AI can be a significant asset. Students who reported using AI weekly for general learning assistance outperformed non-users, albeit marginally. This advantage becomes more pronounced among students who are regularly tasked with evaluating the quality of AI-generated content; with this critical training, all regular AI users for learning assistance outpace their non-using counterparts. The study authors interpret this to indicate that "moderate and intentional use of AI for schoolwork and learning" may yield positive educational outcomes.

"In the same way that we do not become fit by watching sports but by doing sports, learning does not occur through the consumption of content, but as a productive cognitive struggle of the mind with new material," wrote OECD Director of Learning and Skills Andreas Schleicher. "Where technology enables or enhances that cognitive struggle, students will advance. Where technology short-circuits the productive struggle of learning, it will undercut students' development."

The study also examined the distribution of AI usage among learners, finding that its use is more prevalent among students from relatively advantaged backgrounds. This disparity likely stems from better access to hardware, stable internet connections, and paid AI services. Internationally, the adoption rates vary significantly, with over 95 percent of Vietnamese students reporting AI usage compared to just 60 percent in Japan.

Despite the mixed academic results, AI users possess a distinct advantage: the study reveals that curiosity levels correlate with AI usage, peaking among daily users. However, this enthusiasm is not yet consistently translating into superior grades.