PISA 2025 records a deep reading and maths decline—without a simple culprit

the PISA 2025 results: what the primary source confirms

The OECD released PISA 2025 on 8 September, based on about 760,000 15-year-olds in 91 countries and economies. It reported the lowest OECD-average performance recorded in reading and mathematics: [OECD release](https://www.oecd.org/en/about/news/press-releases/2026/09/pisa-2025-students-reading-and-mathematics-performance-declined-sharply-across-the-oecd.html).

PISA is a sample-based international assessment, not a census of every teenager. Its 760,000 participants are used to estimate performance among 15-year-olds in participating systems under a common framework. The scale gives the OECD average weight, yet it does not erase sampling uncertainty, differences among jurisdictions or variation inside schools. Reading the technical design alongside the press release prevents two opposite mistakes: dismissing a broad decline because individual students differ, or treating an international mean as the exact experience of every learner. National and subgroup tables are needed for more specific decisions.

the PISA 2025 results: the dates that define the update

Across the OECD, average reading performance was 28 points lower and mathematics 22 points lower in 2025 than in 2015. This decade comparison is different from a one-cycle change and should be labelled with both endpoints.

PISA points are positions on an assessment scale, not percentage points and not the share of questions answered correctly. The 28-point reading and 22-point mathematics declines compare the OECD average in 2025 with 2015, so both the unit and the decade-long baseline belong in the sentence. Converting those changes into an invented percentage would distort the measure. Policy discussion should also separate a long-term comparison from the movement since the previous cycle. Each answers a different question about trajectory, recovery and how persistent the underlying learning challenge may be.

the PISA 2025 results: how to read the headline number

Science performance was broadly stable compared with 2022, according to the release. That contrast matters because a single narrative of uniform collapse would hide the different pattern measured across the three domains.

The science result resists a single story of universal deterioration. Broad stability against 2022 does not mean every country or skill in science was unchanged; it means the OECD-level pattern differs from the record lows reported for reading and mathematics. That contrast is analytically valuable. It encourages schools and governments to investigate domain-specific conditions rather than attach one explanation to all learning. It also makes the article more accurate: the release describes a serious cross-domain concern, but the direction and comparison period are not identical for each assessed area.

the PISA 2025 results: who may need to respond

One in five students across OECD countries was a low performer in reading, mathematics and science, up from 16% in 2022. The measure identifies a larger overlapping challenge; it does not describe every student or every education system.

The one-in-five figure describes students who fall into the low-performing group in all three domains at once. It is therefore an overlap measure, not three separate rates added together. Comparing it with 16% in 2022 signals a larger group facing compounded difficulty, a problem that can affect access to later study and participation. The statistic still cannot identify an individual child’s needs. Schools need their own diagnostic evidence to distinguish reading fluency, comprehension, mathematical foundations, language support and other barriers before selecting instruction or support.

the PISA 2025 results: what the evidence does not prove

The international average contains substantial variation among countries, schools and learners. A jurisdiction should read its own confidence intervals, trends and subgroup results in the [full OECD report](https://www.oecd.org/en/publications/2026/09/pisa-2025-results-volume-i_5265bfb1/full-report.html) before choosing an intervention.

International league-table language can hide more than it reveals. Country averages are shaped by distributions, participation and uncertainty, while national systems contain regions and student groups with different trajectories. The full OECD volume allows readers to move from the headline average toward those details. A responsible comparison asks whether the same age group, domain, scale and cycle are being used, and it avoids turning small rank differences into claims of clear superiority. For policy, the more useful question is which systems improved or remained resilient under comparable conditions and which practices deserve closer investigation.

the PISA 2025 results: a proportionate planning response

The release discusses digitalisation and changing learning environments, and Euronews highlighted that context. Yet the assessment is not a controlled experiment proving that screens, artificial intelligence or one platform caused the score decline.

Digitalisation rose during the same broad period as falling reading and mathematics results, but temporal movement and statistical association do not by themselves identify a cause. Device access, purpose of use, teaching design, home background, attendance, pandemic disruption and many other variables can interact. PISA can reveal patterns and support hypotheses; it is not a randomised trial assigning students to a single technology. Consequently, claims that artificial intelligence, screens or one application caused the global result exceed this release. Causal language needs a design capable of ruling out plausible alternative explanations.

the PISA 2025 results: questions to ask before repeating the claim

School leaders can respond without pretending the cause is settled: protect sustained reading, diagnose mathematical foundations, support teachers and measure whether chosen changes improve learning. A fashionable explanation is weaker than a local baseline followed by repeat evidence.

A school does not need to wait for one universal culprit before improving practice. It can establish local baselines, protect sustained reading, diagnose number sense and reasoning, support subject teachers, and review whether digital tools strengthen or fragment the intended task. Each intervention should have an owner, a time window and a learner outcome that can be checked. This approach is deliberately narrower than a global reform slogan: it turns international evidence into local questions without pretending the OECD average prescribes the same remedy for every classroom.

the PISA 2025 results: the next signal worth watching

The next responsible step is to inspect country tables, technical notes and future comparable assessments. Policy claims should state whether they refer to an OECD average, a country result, a subgroup or a correlation; those are not interchangeable.

The full report and technical material are the next stop for decisions involving curriculum, funding or accountability. Readers should look for country estimates, uncertainty, subgroup patterns, questionnaire definitions and changes in participation before making a claim. Future PISA cycles can show whether the direction persists, but they will still need comparable methods and careful interpretation. Editorial updates should be tied to a new OECD table, correction or analysis rather than a fresh opinion about the same figures. That preserves the distinction between evidence, explanation and advocacy throughout the public debate.

the PISA 2025 results: a decision-ready reading of the update

A decision-ready education brief should move down an evidence ladder. It begins with the OECD average and decade trend, then checks the participating country, relevant student groups, the school’s own assessment and finally the particular learning task an intervention is meant to improve. Each step narrows the claim and makes action more accountable. Jumping straight from an international mean to a universal ban, purchase or curriculum prescription skips the evidence needed to know whether the proposed remedy matches the local weakness. The value of PISA is to reveal the scale and invite disciplined investigation, not to replace that investigation.

For this edition, the confirmed story is a record-low OECD average in reading and mathematics, a different recent pattern in science and a larger overlapping low-performing group. The unresolved story is causal: the release cannot isolate one technology, teaching practice or social change as the explanation. Future revisions should follow technical analyses, national breakdowns and comparable assessments, clearly labelling whether new work is descriptive or causal. Schools can act now by measuring specific skills and evaluating their response, while resisting the attractive but unsupported promise that one global culprit will produce one global cure.

A useful school response would test one defined change against a local baseline—for example, protected sustained-reading time or targeted support for mathematical foundations—then examine whether the intended learners improve. That evidence cannot repair the international trend by itself, but it can distinguish an effective local intervention from a policy chosen mainly because it fits the loudest explanation of the PISA headline.