AI Use in Schools Linked to Worst Student Performance Decline in 25 Years, OECD PISA Study Finds

Global Student Performance Hits 25-Year Low

The OECD's Programme for International Student Assessment has revealed alarming trends in student performance decline across 91 countries, with 760,000 15-year-olds tested in 2025 showing the weakest results in reading, mathematics, and science since global testing began 25 years ago1

1

2

2

. The decline in reading, math, and science scores represents more than just numbers on a test—it signals a fundamental shift in how students engage with learning in an era dominated by AI use in schools and digital distractions in classrooms.

Reading scores across OECD countries dropped 28 points between 2015 and 2025, with 25 points lost between 2018 and 2025 alone2

2

. Since 20 points roughly equals one year of schooling for a 15-year-old, today's students are reading at levels expected of students a year younger. Mathematics fell 22 points over the decade, while science proved more resilient, slipping just 7 points2

2

. The impact of AI on student performance has emerged as a central concern, with the global decline in student reading skills particularly troubling for long-term economic prospects.

The AI Paradox: Moderate Use Shows Promise

The OECD PISA study uncovered a surprising Goldilocks effect in AI chatbot use for schoolwork. Students who used AI moderately—from about once a month to once or twice a week—recorded higher average science scores than both students who rarely used AI and those who used it almost every day1

1

. The advantage was especially clear when students used AI to summarize assigned reading or conduct preliminary research on new topics.

Students who reported using AI for these purposes weekly tended to perform at roughly the same level as AI nonusers after accounting for socio-economic gaps1

1

. However, students who used it almost every day scored about 30 points lower than those who rarely touched it4

4

. Andreas Schleicher, OECD director for education and skills, explained: "You don't get fit by watching sport. If AI makes things easy, it's not great for you in the long run. It gives you better results, but it doesn't enhance learning"4

4

.

The Rise of Hasty Readers and Declining Critical Thinking

One of the most troubling findings centers on the erosion of foundational skills essential for cognitive engagement. The share of hasty readers—students who rush through texts and give quick but incorrect answers—nearly doubled between 2018 and 2025, rising from about 7% to 11%2

2

1

1

. Meanwhile, the share of accurate, fluent readers fell by about 7 percentage points over the same period.

Performance dropped most sharply on tasks requiring critical thinking—evaluating information, making connections across multiple sources, and thinking critically about content1

1

. Schleicher noted that students become "quicker but less accurate readers," no longer excelling at reflection and critical thinking5

5

. This pattern aligns with the Flynn effect reversal, where decades of increasing IQ scores appear to be reversing due to digitalization and screen time3

3

.

Digital Distractions and Social Media Impact

Across OECD countries, 28% of students reported that classmates are distracted by digital devices in most or every science lesson2

2

. Students reported spending about 1.7 hours daily on digital devices for learning at school and 1.1 hours for leisure while at school. AI and social media use showed particularly alarming correlations—students who excessively used social media at school performed about 44 points worse in mathematics than those who barely did4

4

.

Schools implementing banning smartphones on premises recorded lower distraction levels, with nearly half of students across OECD countries now attending such schools, up from 34% in 20222

2

. Schleicher emphasized: "It's the only thing that works. The only reduction of destruction was the cell-phone ban"5

5

.

Surprising Trends Among Advantaged Students

The OECD PISA study revealed an unexpected pattern: socio-economic gaps in performance narrowed between 2022 and 2025, but not due to gains among disadvantaged students. Instead, advantaged students experienced the largest score declines2

2

. In science, disadvantaged students' performance held steady while advantaged students' scores fell 8 points. In reading, disadvantaged students slipped 4 points against a 20-point drop among advantaged students2

2

.

This suggests what Schleicher called "a new phenomenon, with the biggest factor in the overall drop in scores being advantaged students who are performing below what one would expect given their background"5

5

. Socio-economic background nonetheless remains a strong predictor of achievement, accounting for 12% of within-country variation in science performance.

What This Means for Education's Future

The decline in student performance carries significant implications for long-term economic outlook, as most economists believe growth is partly driven by improvements in education levels5

5

. While new AI tools promise to raise productivity among existing workers, the PISA results suggest they could already be lowering productivity levels among future workers.

Education writer Daisy Christodoulou suggests schools might need to become "gymnasia for the mind"—places that allow students to practice and acquire foundational skills no longer directly rewarded in daily life but still vital3

3

. The OECD findings indicate that moderate and intentional use of AI for schoolwork and learning may be positively related to student performance, but excessive or unguided use contributes to distraction and shallower engagement1

1

.

With 86% of students across OECD countries having used AI chatbots for at least one schoolwork purpose, and about one in five using it almost every day1

1

, educators face mounting pressure to establish guidelines that harness AI's benefits while protecting the cognitive struggle essential for genuine learning. As the report notes, "learning does not occur through the consumption of content, but as a productive cognitive struggle of the mind with new material"1

1

.