Reading, mathematics, and science scores among 15-year-olds across the industrialized world have fallen to the lowest levels ever recorded by the OECD’s Programme for International Student Assessment, according to results released September 8. The assessment, which tested 760,000 students across 91 countries in 2025, documents a decade-long erosion in foundational academic skills that the OECD links directly to rising digital distraction, declining reading habits, and the emerging effects of artificial intelligence on how students process and produce written work.
Key Takeaways
- Average reading scores across OECD countries fell by 28 points between 2015 and 2025, equivalent to approximately one and a half years of learning; OECD reading scores dropped from 489 to 461. Mathematics scores fell by 22 points over the same period, from 485 to 463.
- One in five 15-year-olds across OECD countries is now classified as a low performer in science, mathematics, and reading simultaneously, up from 16% in 2022.
- Students reported spending an average of 3.4 hours per weekday on digital devices for leisure and another 3 hours for learning; digital distraction was linked to lower science scores in 59 of 82 education systems studied.
- Students who reported not using AI to draft text for writing assignments outperformed those who did; however, among frequent AI users, those who used AI to support learning rather than replace it scored higher than those who used it as a shortcut.
- The U.S. saw reading scores decline while math and science results held essentially flat between 2022 and 2025; 74% of American 15-year-olds reached basic reading proficiency, compared to 69% across the OECD average.
- The Chinese jurisdictions of Beijing, Shanghai, Jiangsu, and Zhejiang, along with Singapore, recorded the highest scores globally across all three domains.
Reading Suffered the Steepest Decline Across OECD Countries
PISA, administered every three years since 2000, has become the most widely cited international benchmark for comparing educational performance across countries. The 2025 cycle tested students in science as its primary focus, with mathematics and reading as secondary domains. A new computational problem-solving assessment was also included for the first time, with results on self-directed learning in digital environments scheduled for release in 2027.
Reading recorded the sharpest drop of any domain. Average reading scores across the 38 OECD member countries fell from 489 in 2015 to 461 in 2025, a 28-point decline that the OECD calculates as equivalent to roughly one and a half years of learning. The decline accelerated between 2022 and 2025, with a 14-point drop in that three-year interval alone, the steepest single-cycle fall in reading since PISA began. Among non-OECD participants, reading scores also fell, declining 16 points over the decade. A 2012 reading high of 501 on the PISA index has now dropped to 466 globally, a trajectory that OECD Secretary-General Mathias Cormann described as directly connected to reduced reading for pleasure and increased screen consumption.
Mathematics followed a parallel downward path. Average OECD math scores fell from 485 in 2015 to 463 in 2025, a 22-point decline equal to just over one year of learning. Science, the focal domain of the 2025 assessment, showed a more modest decline, with average OECD science scores dropping from 489 to 482 over the decade. A handful of countries moved against the trend: Costa Rica, the Slovak Republic, Türkiye, and the United Kingdom all saw science scores increase between 2022 and 2025.
Digital Distraction Is Now Measurable Across 59 Education Systems
PISA 2025 included detailed questions about students’ digital device use, producing the assessment’s most granular data yet on the relationship between screen time and academic performance. Students across OECD countries reported spending an average of 3.4 hours per weekday using digital devices for leisure activities, including social media, gaming, and short-form video consumption. An additional 3 hours per weekday were spent on devices for learning purposes, including schoolwork and research. The combined total of roughly 6.4 hours of daily device use among 15-year-olds represents a substantial share of waking hours outside of sleep.
The OECD found that digital distraction was associated with lower science scores in 59 of 82 education systems for which data was available. More than one in four students reported that their classmates were distracted by digital devices in most or every science lesson. The distraction effect extends beyond the individual user. Students surrounded by distracted classmates performed less well, were less engaged in learning, and felt less connected to their school, regardless of their own device habits. The finding suggests that classroom device policies affect not only the students using the devices but the learning environment for every student in the room.
Cormann addressed the connection between reading decline and digital behavior at a press conference accompanying the release. “In reading, more screen time and less reading for pleasure have gone hand in hand with weaker results,” Cormann said. “We’re also seeing more hasty reading, students rushing through a text and giving a quick wrong answer.” The OECD’s analysis of reading behavior identified a growing share of students who exhibit rapid, surface-level reading patterns, scanning texts quickly and selecting answers before fully processing the content, a behavior researchers labeled “hasty reading” and linked to habitual engagement with short-form digital content.
AI Use Presents a Complex and Uneven Effect on Student Performance
For the first time, PISA 2025 collected data on how students use artificial intelligence tools in their schoolwork. The headline finding was straightforward: students who reported not using AI to draft text for writing assignments outperformed those who did. The result is consistent with educator concerns that AI tools, when used to generate rather than support written work, allow students to bypass the cognitive processes that build comprehension and analytical skill.
The data underneath that headline, however, was more nuanced. Among students who did use AI frequently, those who deployed it as a learning aid, using AI to help them understand material, check reasoning, or explore concepts, scored higher than those who used it primarily to produce finished work. The distinction matters for the policy debate around AI in education. The OECD did not recommend blanket bans on AI in schools but warned that the purpose and moderation of use determined whether the effect was positive or negative. Similarly, the OECD stopped short of endorsing school smartphone bans, though the data showed that academic performance declined when daily device use exceeded five hours.
The teacher shortage adds another dimension to the digital and AI story. Across OECD countries, approximately four out of ten students attended schools where principals reported that a shortage of teachers hindered instruction to some extent or significantly. A similar proportion reported that a lack of assisting staff impaired learning. Teacher shortages were more frequently reported in socioeconomically disadvantaged schools than in advantaged ones, and in public schools more often than in private schools, compounding the equity implications of both digital distraction and uneven access to guided AI instruction.
U.S. Results Show a Paradox: Relative Gains Amid Absolute Decline
The United States presents an unusual case within the PISA 2025 data. American 15-year-olds saw reading scores decline between 2022 and 2025, consistent with the global trend. However, math and science scores held essentially flat over the same period. The flat performance, in a context where most other industrialized nations recorded declines, caused the U.S. to move up in international rankings, not because American students improved, but because other countries fell faster.
In the United States, 4,625 students across 144 schools completed the 2025 assessment, representing approximately 3.58 million 15-year-olds. Seventy-four percent of American students reached basic proficiency in reading, compared to 69% across the OECD average. In mathematics, U.S. proficiency was 65%, matching the OECD average. In science, American performance was roughly comparable to that of students in the United Kingdom, Canada, New Zealand, Australia, and Finland. Eight percent of U.S. students qualified as top performers in mathematics.
The U.S. data carried a notable caveat. More than 40% of American students who took the PISA assessment did not complete an accompanying student questionnaire, limiting the ability to break out results by race, family income, or other demographic categories that typically drive analysis of domestic achievement gaps. The U.S. Education Department acknowledged a “month-long pause in PISA activity in spring 2025” that resulted in some questionnaires not being administered. The data gap means that the most policy-relevant dimension of U.S. performance, the size and trajectory of equity gaps between student subgroups, cannot be fully assessed from the 2025 cycle.
The Results Carry Workforce and Economic Implications That Extend Beyond the Classroom
PISA scores have been repeatedly linked in economic research to long-term workforce productivity and national economic growth. The OECD’s own analysis frames foundational literacy as a prerequisite for productive engagement with AI tools in the workplace, meaning that the same students who are underperforming on reading assessments today will enter a labor market that demands higher, not lower, levels of text comprehension, critical reasoning, and analytical writing.
The 20% of OECD 15-year-olds now classified as low performers across all three domains represent a cohort that will age into the workforce over the next decade. The OECD’s Andreas Schleicher, director for education and skills, described reading as “the mother of all skills” and warned that the decline is not a temporary disruption but a structural shift that education systems have not yet adapted to address. The results are expected to influence policy debates in multiple countries, including ongoing discussions in the United States about school device policies, AI regulation in education, and the funding of teacher preparation and retention programs.
FAQs
What Is the OECD PISA Assessment?
The Programme for International Student Assessment is a triennial evaluation conducted by the Organisation for Economic Co-operation and Development that measures the reading, mathematics, and science skills of 15-year-olds across participating countries. PISA 2025 tested 760,000 students across 91 countries and economies, representing approximately 33 million 15-year-olds worldwide. The assessment has been conducted since 2000 and is considered the leading international benchmark for comparing educational outcomes.
How Did U.S. Students Perform on PISA 2025?
U.S. reading scores declined between 2022 and 2025, while math and science scores held essentially flat. Seventy-four percent of American 15-year-olds reached basic proficiency in reading, above the 69% OECD average. In math, U.S. proficiency was 65%, matching the OECD average. The U.S. moved up in international rankings primarily because other countries declined more sharply.
Did PISA 2025 Recommend Banning Smartphones in Schools?
The OECD stopped short of recommending smartphone bans. The assessment found that digital distraction was associated with lower scores in 59 of 82 education systems and that performance declined when device use exceeded five hours per day. However, the OECD concluded that moderate, purposeful device use could support learning, and that the critical variable was how and why devices were used rather than whether they were present.
