The OECD released the PISA 2025 results on September 8, 2026. The assessment covered more than 760,000 15-year-old students in 91 countries and economies, the largest PISA cohort to date. Science was the focal domain, with reading and mathematics tested as minor domains. The OECD reported that U.S. average reading scores fell 14 points, a decline it equates to roughly six or seven months of learning, matching the 14-point drop in the OECD average.
Across the OECD, average reading scores fell 28 points between 2015 and 2025, about a year and a half of learning, while mathematics scores fell 22 points over the same period. The share of 15-year-olds who are low performers in science, mathematics, and reading rose to one in five, up from 16 percent in 2022.
A break from a stable U.S. trend
The U.S. decline follows a period of stability. In PISA 2022, the U.S. reading average was 504, above the OECD average of 476, and the country outperformed 68 of 80 other education systems in reading, according to NCES. OECD country notes for 2022 described long-term stability in U.S. reading results, which makes the 2025 drop a departure from the earlier pattern.
Despite the drop, the U.S. continued to outperform most OECD nations in reading. Only Singapore, Chinese jurisdictions, and Taiwan scored above the United States on the reading measure. U.S. math and science scores were statistically unchanged from 2022, though the U.S. ranking in math was 27th, below its standing in reading or science, according to Education Week.
The new results also show the widest spread. Matthew Soldner, acting commissioner of the National Center for Education Statistics, indicated that the U.S. recorded the widest reading-performance gap between its highest- and lowest-scoring students of any participating country. He also pointed out that PISA 2025 test-takers were in elementary or middle school when the pandemic began, and suggested that this group may have acquired basic literacy skills before encountering later schooling disruptions, which could be a factor in the reading decline.
Where other countries moved
The United Kingdom was the clearest exception to the global decline. Its students matched their pre-downturn reading performance in 2025, and the OECD listed the UK among the ten highest-performing education systems across all three subjects. At the other end, Iceland and Slovenia posted the steepest reading declines among OECD members, with scores down more than 50 points, according to Deutsche Welle.
Reading improved in only five education systems: Qatar, Brunei Darussalam, Cambodia, the Philippines, and Zambia. Science performance rose in a wider set of countries between 2015 and 2025, including Lithuania, the Slovak Republic, Türkiye, the Dominican Republic, Georgia, Macao (China), Montenegro, Qatar, Thailand, the United Arab Emirates, and Uruguay. The OECD cited those gains as evidence that improvement during a period of broad decline is possible.
What the reading data show
The assessment data point to specific difficulties. Students showed particular problems with longer texts and with evaluating information across multiple texts. Reading fluency, the speed and accuracy with which students read, declined, and the share of students who read rapidly and answered quickly and incorrectly rose, according to Education Week.
Andreas Schleicher, the OECD's director for education and skills, indicated that the organization cannot be certain but suggested that reduced reading for pleasure and growing reliance on digital reading, including rapid scanning of social media content, may be associated with weaker student engagement with complex texts and data. He stated that PISA does not draw causal conclusions connecting these patterns to test results.
AI use and performance
PISA 2025 is the first cycle to ask students about AI use. About half of students said they use AI at least once a week to help them learn, and about one in three said they used it to draft text for writing assignments, according to Education Week. Deutsche Welle, citing the same results, reported that more than 40 percent of students said they used AI to draft texts for writing assignments once or twice a week or every day.
The OECD found a generally negative correlation between students' use of AI to draft writing and their reading scores. Students who did not use AI to draft text outperformed those who did. Among frequent AI users, students who were taught to assess the quality of AI-generated information tended to perform better than those who received no such guidance. The OECD cautions that the data are correlational, not causal.
The pattern extended to science. Students who reported using AI chatbots scored lower in science, on average, than students who did not, according to Deutsche Welle. The same reporting notes that the data themselves cannot explain the reading decline, and the OECD found no direct relationship between how long schools closed during COVID-19 and students' reading performance.
Digital distraction is part of the picture. Education International's review of the results reports that moderate use of digital devices for learning is associated with stronger science performance, while excessive use and distraction are associated with weaker outcomes. More than one in four students said classmates were distracted by digital devices in most or every science lesson, and students surrounded by distracted classmates performed less well. Those are correlational findings.
According to Education International's summary, the OECD reports that countries with the largest reading declines also tended to record the largest drops in mathematics and science, suggesting shared underlying drivers. Increased digitalization, longer screen time, and falling rates of reading for enjoyment coincided with weaker literacy outcomes globally, though the OECD explicitly stops short of causal conclusions. The digital pattern predates AI: an OECD framework document notes that prior PISA data have consistently shown students who report using computers very frequently at school perform worse than other students on most learning outcomes, even after accounting for socio-economic background.
Two different reading tests
The reading results may not translate directly to national assessments. A comparison study commissioned by the National Center for Education Statistics and conducted by the American Institutes for Research found that PISA and NAEP reading assessments have the greatest overlap among shared subjects, especially in cognitive skills. The two tests differ in content: PISA includes web-based and digital texts students encounter in daily life, while NAEP uses more continuous passages, school-based texts, and poetry. The study concludes the two assessments should be read as complementary measures rather than substitutes.