The Organisation for Economic Co-operation and Development (OECD) released the 2025 Programme for International Student Assessment (PISA) results on September 8. Among the 91 countries and regions that took part, students from mainland China's Beijing, Shanghai, Jiangsu, and Zhejiang provinces—collectively referred to as "BJ-SH-JS-ZJ"—delivered an outstanding performance: they ranked first in science with 597 points and in mathematics with 612 points, while their reading score of 527 placed them second only to Singapore.
The OECD has conducted the PISA assessment every three years since 2000, aiming to measure the proficiency of 15-year-old students who are nearing the end of their compulsory education in science, reading, and mathematics. Unlike tests that focus solely on curriculum knowledge, PISA places greater emphasis on students' ability to apply what they have learned to solve real-world problems. The PISA 2025 results represent a key milestone in the ninth cycle of this international evaluation.
Where the top performers stand
According to the report, the BJ-SH-JS-ZJ region, along with Singapore, demonstrated the most impressive performance across all three core domains of science, reading, and mathematics in PISA 2025. The Chinese region secured the top spot in both science and mathematics, while finishing just behind Singapore in reading. This outcome once again underscores the strong international competitiveness of basic education in the BJ-SH-JS-ZJ region, particularly in terms of students' academic foundations, knowledge mastery, and comprehensive application skills.
A new frontier: computational problem-solving
In addition to the traditional trio of science, reading, and mathematics, PISA 2025 introduced an entirely new assessment dimension for the first time: computational problem-solving. This test is designed to evaluate students' abilities to use modelling and programming tools, conduct experiments, and build, design, and develop digital products. Students from Macau, China, topped this new domain with a score of 572, followed by Singapore at 563 and the BJ-SH-JS-ZJ region at 560. The average score for students from OECD member countries in this test was 500.
The inclusion of computational problem-solving reflects PISA's ongoing adaptation to the demands of a digital society. The future workforce requires not just individuals who possess existing knowledge, but also learners who can understand complex systems, leverage digital tools, analyse problems, and generate creative solutions. Through this new assessment area, PISA sends a clear signal: in an era of rapidly advancing artificial intelligence and digital technologies, education systems must place greater emphasis on cultivating students' capacity to tackle complex problems, while also ensuring that digital competence is not simplistically equated with device proficiency or coding skills alone.
What PISA indicates about AI's entry into education
The integration of artificial intelligence into education is one of the most closely watched new contexts of PISA 2025. The report reveals that, on average, 46% of students in OECD member countries use AI chatbots at least once a week to support their learning. However, the relationship between AI usage and academic performance is not a simple positive or negative one—it largely depends on how students utilise the technology.
Data shows that students who use AI to complete specific school tasks such as summarising texts, writing assignments, or conducting research tend to score lower in science, on average, than those who do not. The performance gap is approximately 20 points, which is equivalent to roughly one year of schooling. In contrast, students who frequently use AI for general purposes and who also receive opportunities to develop AI literacy at school perform no worse than non-users in science, with some even showing slightly better results.
This finding suggests that artificial intelligence does not automatically lead to improved learning outcomes. The OECD argues that the key lies in whether students truly understand AI, use it responsibly, and treat it as a tool that supports learning rather than a substitute for their own thinking. It is also noteworthy that students who have access to AI literacy learning opportunities are more likely to come from socioeconomically advantaged families, which could potentially widen the existing digital divide and create a new "AI gap." Ensuring that students from all backgrounds acquire the necessary digital competencies and AI literacy is emerging as a pressing challenge for educational equity.
Beyond the new issues raised by AI usage, PISA 2025 also highlights a broader international trend that warrants attention: student performance worldwide continues to face significant pressure. The average performance of OECD member students in science, reading, and mathematics is at its lowest level ever recorded. Between 2015 and 2025, the average science score among OECD members declined, while the average mathematics score fell by 22 points. Many non-OECD countries and economies also experienced declines in mathematics performance. The stagnation or even continued decline in learning outcomes has become a common challenge confronting numerous education systems.