PISA results, PISA misgivings

I used AI to trawl through the latest PISA stats. These pertained to 2025 and were published in September 2026. One AI seemed to believe the most recent results were from 2022. Weird. With AI, it really is a matter of caveat emptor! OK, let’s get into it.

The results

ChatGPT Prompt: summarise latest PISA results (2025) in the fields of English, maths and computing.

Yes. One important clarification first: the PISA 2025 results were actually published on 8 September 2026, so they are now the latest PISA results. PISA tests 15-year-olds. Also, PISA does not have a subject called “English”: it tests reading. Computing is represented by a new computational problem-solving assessment. (OECD)

PISA 2025 — the headline picture

Reading / “English”

• OECD reading fell dramatically: −14 points since 2022 and −28 since 2015.

• The UK was much more resilient: 494, essentially unchanged from 2022.

• 80% of UK pupils reached Level 2 or above, compared with 69% across the OECD.

• 9% of UK pupils were high performers (Level 5+), against 6% OECD.

• Particularly worrying internationally is deterioration in evaluating information, connecting information across sources and critical reading—skills especially relevant in the age of AI.

Mathematics

• OECD average: 463, down 9 points since 2022 and 22 points since 2015.

• UK: 488, down just 1 point since 2022, a statistically insignificant change.

• 74% of UK pupils reached Level 2+, compared with 65% OECD.

• 13% of UK pupils were Level 5/6 high performers, versus 8% OECD.

• The UK nevertheless has a sizeable gender gap: boys outperformed girls by 24 points, compared with an OECD average gap of 13.

Computing
This is the really interesting new bit for an ICT/Computing perspective. PISA 2025 introduced computational problem solving rather than simply testing computer knowledge or programming in isolation. It examines how pupils use computational tools and practices to model problems, experiment, program and create digital products.

• OECD average: 500

• UK: 523

• UK therefore performs significantly above the OECD average.

• The highest scores were Macao 572, Singapore 563, B-S-J-Z China 560 and Japan 557.

• The UK was roughly alongside Canada and ahead of countries such as Switzerland, Czechia and Ireland.

• An interesting finding is that computational problem-solving correlates strongly with maths (0.75) and science (0.76), but it isn’t simply another version of those subjects: PISA says it measures a distinct set of skills.

The really significant story

For England/UK education, I’d reduce the results to three messages:

1. The UK is doing considerably better than the OECD average.
The UK is one of the systems in the top ten in all three core PISA subjects: science, mathematics and reading.

2. The international deterioration in reading and maths is much more alarming than the UK figures initially suggest.
The OECD has reached its lowest-ever average performance in reading and mathematics. The UK has largely avoided the dramatic falls seen elsewhere, particularly in reading.

3. “Computing” is no longer just about knowing how to code.
The new PISA assessment is arguably more interesting educationally: it asks whether students can use computational thinking, tools, experimentation, modelling and programming to solve unfamiliar problems and create things. That’s potentially a much better fit with where school Computing is heading.

One other striking PISA finding for anyone working in EdTech: moderate educational use of digital devices is associated with better outcomes, whereas excessive use and digital distraction are associated with worse outcomes. 28% of OECD pupils said classmates were distracted by devices during most or every science lesson. AI use also produced a complicated picture rather than the simple “AI = better learning” story.

PISA misgivings

I made a note of what was wrong with the PISA comparisons in my opinion, and then asked At what the objections to it were. (I’m relieved to report that it came up with the same list!)

Perplexity prompt: What are the misgivings about PISA as a means of measuring and comparing education achievement?

There are deep and well-documented misgivings about using PISA (the Programme for International Student Assessment) as the primary means of measuring and comparing educational achievement across countries. Critics argue that its design, assumptions, and the way its results are used make it a poor basis for judging school systems or driving policy.

Core conceptual criticisms

• Narrow definition of “achievement”: PISA focuses on reading, mathematics and science literacy at age 15, but ignores many valued outcomes of education such as creativity, citizenship, arts, ethics, social skills, and subject-specific knowledge in history, geography, languages, etc

• Unproven claim about “skills for the future”: PISA presents itself as measuring competencies needed for modern life and economies, but scholars note there is little empirical evidence that its items validly capture these broader life skills, or that they are universally applicable across very different societies.

• Western-centric, monolithic view: The test assumes a single set of skills and knowledge is desirable everywhere, which critics say imposes a largely Western, OECD-style vision of education on diverse cultures and systems.

Methodological and statistical concerns

• Cross-sectional, not longitudinal: Each PISA cycle tests a different sample of 15-year-olds; it does not follow cohorts over time. That limits what can be inferred about the impact of specific policies or reforms, because you’re not seeing the same students’ progress.

• Measurement error and unstable rankings: Country rankings often hinge on small differences that sit within the tests’ error margins. When error terms and confidence intervals are properly considered, many “gains” or “drops” in league tables are not statistically meaningful.

• Sampling and representativeness issues: In some high-profile cases (e.g., Shanghai), results have been based on subsets of students that are not representative of the whole population, yet those results are used to hold up entire systems as models.

• Context ignored: PISA does not adequately adjust for differences in family academic resources, socio-economic context, enrolment rates, or the proportion of students actually in school at age 15, all of which strongly influence average scores.

Curriculum and policy distortions

• Curriculum narrowing: Because PISA rankings carry high political stakes, systems often narrow the curriculum toward tested areas, marginalising arts, humanities, and other subjects not measured by PISA.

• Teaching to the test and “PISA industry”: The pressure to climb the rankings encourages test-focused teaching, short-term policy fixes, and a market in PISA-aligned materials and consultancy, rather than deep, sustainable educational improvement.

• Misuse for causal claims: Politicians and commentators frequently treat correlations between PISA scores and economic outcomes as if they were causal, despite the well-known “correlation is not causation” problem and flawed temporal logic in some influential reports.

Governance and conflict-of-interest worries

• OECD as tester and policy interpreter: The OECD both administers the test and interprets the results for policy purposes, creating a potential conflict of interest and reinforcing a particular policy narrative (often aligned with market-oriented reforms).

• Neoliberal policy driver: Critics argue PISA’s ideology promotes competition, standardisation, and accountability regimes that fit a neoliberal agenda rather than a broader educational vision.

Practical comparability problems

• Different systems, different aims: Educational systems value different things (e.g., vocational pathways, local languages, civic education). PISA’s single metric cannot capture these differing goals, yet its league tables imply a universal hierarchy.

• Within-country variation larger than between-country differences: In large, diverse countries (e.g., the US), differences between states or regions can be bigger than differences between nations, making national averages misleading for policy.