Every few years, the global education community goes through a familiar “results day”. On September 8, the Organisation for Economic Co-operation and Development (OECD) released the latest round of PISA 2025 results. More than 760,000 15-year-old students from 91 countries and economies took part in the assessment, making it the largest PISA to date.
For Australia, the results do not look particularly good. Australian students recorded an average score of 509 in science, 491 in reading and 478 in mathematics. Reading fell another 7 points from 2022, while mathematics fell by about 9 points, with both reaching their lowest levels since Australia first participated in PISA. Looking further back, Australia’s reading score has fallen from 528 in 2000 to 491, while mathematics has fallen from 524 in 2003 to 478. It is therefore natural that “Australian education is continuing to decline” would become a news headline.
But another set of figures is equally true: Australia’s scores in all three subjects remain above the OECD average. In 2025, the OECD averages were 482 for science, 461 for reading and 463 for mathematics. In Australia, 82% of students reached the basic proficiency level in science, 71% reached the basic level in mathematics and 77% reached the basic level in reading, all above the OECD averages.
So what exactly does PISA tell us about Australian education: that it is “very poor”, or that it is “actually not bad”? The answer may depend on what exactly we want to prove with this report card.
International Education Comparisons at the Political Level
Today, we are accustomed to seeing “global education rankings”, but putting students from different countries on the same test paper and comparing them has only a history of several decades.
In the late 1950s, the International Association for the Evaluation of Educational Achievement (IEA) began exploring the possibility of cross-national comparisons in education. In 1960, the IEA conducted a pilot study in 12 countries, testing the abilities of 13-year-old students in mathematics, reading, geography, science and other subjects. In 1964, it conducted the first International Mathematics Study, or FIMS.
The timing was not accidental. In 1957, the Soviet Union successfully launched the world’s first artificial satellite. For the United States at the time, this was not simply a setback in the space race. It was quickly understood as an education crisis. If the Soviet Union could be the first to put a satellite into space, did this mean that the United States had already fallen behind in its ability to train scientists, mathematicians and engineers?
The following year, the US Congress passed the National Defense Education Act, significantly increasing investment in mathematics, science, foreign languages and the development of talent in higher education. From then on, education became more directly connected to national security, technological competition and economic strength. Therefore, early international education comparisons cannot simply be understood as asking, “Which country’s children are smarter?” The larger question was actually whether our education systems could produce the people our countries would need in the future.
From 1980 to 1982, the IEA conducted the Second International Mathematics Study, or SIMS. In 1995, this developed into the larger Third International Mathematics and Science Study, later known as TIMSS, which has continued to this day. TIMSS still focuses on comparing students’ learning outcomes in mathematics and science, and is relatively closely aligned with the content taught in schools.
The Way We Assess Education Has Changed
In 2000, the OECD conducted PISA for the first time. Its biggest difference from TIMSS was not simply the addition of reading, but that it asked a different question.
TIMSS is closer to asking, “Have students learned the mathematics and science that schools are supposed to teach?” PISA, by contrast, deliberately does not fully follow the curriculum of any particular country. Instead, it tests whether 15-year-old students can apply what they have learned to unfamiliar and real-world situations, such as reading charts, evaluating information, solving problems and making inferences. In other words, PISA is not simply asking about the content of textbooks, but about students’ ability to apply knowledge.
This also reflects another shift behind international education comparisons. During the Cold War, countries were concerned about whether they had enough scientists to keep up with their rivals. In the era of globalisation, countries became more concerned about whether their young people could compete in a knowledge-based economy.
The scale of PISA therefore expanded rapidly. The first round involved 43 countries and economies. By 2022, there were 81, and in 2025 the number rose to 91. What began largely as an international comparison involving wealthy industrialised countries gradually became a global map of education. Once the whole world began using the same measuring stick, that measuring stick itself naturally acquired unprecedented political power.
Does Every Country Have an Education Crisis?
After PISA results are released, politicians, education organisations and the media often immediately ask: Where did we rank? Did we rise or fall?
One of the most famous examples is Germany. When the first PISA results were released in 2000, Germany’s results were below domestic expectations, creating what later became known as the “PISA shock”. This triggered years of education reform, including the development of national education standards, stronger support for disadvantaged and migrant students, and the establishment of more systematic education monitoring. This shows that PISA’s function has long gone beyond research.
A fall in rankings can be used by an opposition party to demonstrate government incompetence, by teacher organisations to demand more resources, or by a government to demonstrate that a particular reform is urgently needed.
Australia this year is a good example. The day after PISA was released, Education Minister Jason Clare had already linked the results to the government’s ongoing efforts in reading, phonics, mathematics curricula and school reform, and used improvements in England’s recent results as evidence supporting the direction of reform. This does not mean that these reforms lack a basis, but it reminds us that international rankings are never interpreted in a political vacuum.
The same Australian results can be presented as “reading and mathematics have reached historic lows”, or as “all three subjects remain above the OECD average”. Neither statement is wrong. The former emphasises the downward trend over two decades, while the latter emphasises Australia’s relative position in international comparisons.
Are East Asians Better at Taking Tests?
Whenever PISA results are released, another familiar discussion concerns East Asian education.
In 2025, the four participating Chinese regions and Singapore once again ranked near the top, while Japan, South Korea, Taiwan, Macao and Hong Kong also ranked near the top in several areas. Many people therefore explain this by saying that East Asia has been influenced by Confucian culture, places greater importance on study and examinations, and that students study more, attend more tutoring and are more accustomed to high-pressure examinations, giving them a natural advantage in large-scale international assessments. By contrast, Western countries such as Australia place greater emphasis on student autonomy, creativity and a freer form of education, putting them at a disadvantage in standardised tests.
This explanation is not entirely wrong. A number of Asian education systems, including Hong Kong, Singapore, Japan and Taiwan, have indeed had high-stakes public examinations for a long time. The OECD has also pointed out that high-stakes central examinations can affect school teaching and student behaviour. But attributing all of East Asia’s high scores to examination culture is also an oversimplification.
First, PISA itself is not a typical curriculum examination. It deliberately tests the ability to apply knowledge to new situations. In fact, when the OECD compared PISA and TIMSS, it found that countries’ relative performance in the two assessments was not completely the same. Some of the differences came from the content of the assessments, as well as the fact that PISA samples by age while TIMSS samples by grade.
Second, Singapore’s high scores cannot simply be explained by saying that it only selects elite students to sit the test. PISA requires participating jurisdictions to select a representative sample of 15-year-old students enrolled in school, rather than choosing only students from elite schools or top performers. Singapore’s small system, highly centralised policies, and the strong coordination between its curriculum, teacher training and assessment may all be factors more worthy of study when understanding its results than simply saying that “its students are better at taking tests”.
The same applies to Hong Kong. The gradual move from “elite education” towards a more universal education system is indeed part of Hong Kong’s education history, but it cannot therefore be directly claimed that the recent decline in PISA scores is “because elite students were taught in the past, while everyone goes to school now”. PISA itself is based on the population of 15-year-old students enrolled in school, and Hong Kong had already entered the era of universal basic education when it participated in PISA.
Population structure, curriculum reforms, learning time, examination systems, family resources, student motivation and even sampling methods may all affect the final scores. This shows that the illusion most easily created by international rankings is to compress a complex education system into a single number.
AI’s Fault
When PISA was released in 2022, COVID-19 was an unavoidable part of the background. Students in many countries experienced school closures and remote learning, while mathematics and reading scores across OECD countries also experienced historically unusual declines. But at the time, the OECD had already warned that declines in many countries had actually begun before the pandemic.
By 2026, a new explanation has emerged: AI. This PISA cycle is the first full-scale international assessment conducted after generative AI such as ChatGPT became widespread, so the OECD asked students in detail for the first time about how they use AI.
In Australia, 50% of 15-year-old students said they used AI chatbots to help with learning at least once a week, above the OECD average of 46%. Thirty-five per cent used AI to conduct initial research on new topics, 33% used it to summarise assigned reading, and 27% used it to draft homework. Only about 15% of Australian students said they almost never used AI for the learning tasks covered by PISA.
On the surface, the data seems to provide an easy story: “Students rely more on ChatGPT, so they are becoming worse at reading.” But PISA itself did not draw such a simple conclusion.
The data shows that students who used AI to summarise, conduct initial research or write tended to have lower average scores than those who did not use it at all. But the OECD specifically emphasised that this is a correlation and does not mean that AI itself caused lower scores. Students who already have greater learning difficulties may be more reliant on AI, or the key issue may not be “whether it is used” but “how it is used”.
Australia’s data reflects this point particularly well. Some 68.3% of Australian students said they had learned in class how to assess the quality of AI-generated information, above the OECD average of 62.6%. Among students who frequently use AI, those who had received this kind of AI literacy education generally performed better in science.
So the real question may not be “Should students use AI?” but whether students are treating AI as an answer machine that replaces thinking, or as a tool that helps them think.
PISA in the Age of AI
This is also the most noteworthy aspect of this year’s PISA. The OECD found that some of the reading abilities that have declined most significantly in recent years are precisely those that are most needed in the age of AI: comparing different sources of information, evaluating evidence, distinguishing facts from opinions, and making inferences when information is contradictory. Compared with 2018, the proportion of “hasty readers” who gave incorrect answers after reading quickly had almost doubled in 2025.
This makes something somewhat ironic.
When ChatGPT can write an article in seconds, and Google and AI can provide answers immediately, we might think that memorising knowledge and reading long texts are becoming less important. But the easier it becomes for AI to generate answers, the more people need to know: Is this answer trustworthy? What has it left out? Where is the evidence? When two claims contradict each other, which one should I believe?
These are precisely the abilities that PISA has been trying to measure for years.
PISA 2025 also introduced “Learning in the Digital World” for the first time, testing whether students can learn independently in digital environments, use computational tools, break down problems and adjust their problem-solving strategies. Australia scored 532 on the currently released computational problem-solving indicator, above the OECD score of 500, and performed relatively better on this measure than on its three traditional subjects. A complete analysis will not be released until 2027.
Therefore, PISA 2026 may have raised a question that is more important than rankings.
More than 60 years ago, international education comparisons emerged because Sputnik made countries concerned about whether their students could be trained to become the next generation of scientists and engineers. Today, the threat is no longer a satellite suddenly launched into space, but an artificial intelligence technology whose capabilities are changing almost every few months.
But the core question facing education has not really changed. We are still asking: Is what schools are teaching today enough to prepare the next generation for a completely different world?
So rather than simply asking whether Australia ranks 10th or 16th, or how many more points it has fallen, perhaps the more important questions are: Why have students’ reading and mathematics abilities been declining for two decades? Why have some education systems managed to improve against the trend? What types of technology use genuinely promote learning, and what types simply outsource thinking?