Blog

PISA 2025: Sweden's Wake-Up Call and the Skills the Numbers Don't Yet Measure

PISA 2025: Sweden's Wake-Up Call and the Skills the Numbers Don't Yet Measure

How the newest global assessment of 15-year-olds reads for Sweden, what the world's strongest performers do differently, and why the more urgent lesson may lie outside the score tables altogether.

On 8 September 2026, the OECD released the results of PISA 2025- the eighth round of the Programme for International Student Assessment since it began in 2000, and the first to include a new domain built for the world these students actually live in: learning inside digital, AI-saturated environments. Ninety-one countries and economies took part. The headline finding was stark enough to lead news bulletins across the OECD: reading and mathematics performance have fallen to the lowest levels ever recorded by the programme.

This is worth reading carefully rather than reacting to. PISA is not a report card in the everyday sense -it's a large, carefully calibrated comparative study, and its value lies less in any single point estimate than in the pattern across countries, subjects and years. Read that way, the 2025 results say two things at once: something has gone wrong with foundational skills across almost the entire industrialised world, and the countries that are coping best share a small number of identifiable habits. Sweden's results sit inside both stories.

Key takeaways

  • OECD-wide, reading (461) and mathematics (463) scores hit their lowest levels since PISA began; roughly one in five 15-year-olds (20%) is now a low performer across all three core subjects, up from 16% in 2022.
  • Sweden recorded its lowest-ever scores in reading (466, down 21 points) and mathematics (464, down 18 points); science held roughly steady at 485.
  • Sweden's socioeconomic gap remains stark: 49% of disadvantaged students miss basic mathematics proficiency, versus 19% of advantaged students.
  • Singapore and China's benchmarked regions (B-S-J-Z) top the rankings; Estonia and Ireland offer more transferable lessons for a system Sweden's size.
  • PISA's new "Learning in the Digital World" domain signals that operating a digital tool fluently is not the same as being able to judge whether its (or an AI's) output is true.
  • The article closes with 12 recommendations for policymakers, schools and parents.

What Do the PISA 2025 Results Show Globally?

Across the OECD, the average 15-year-old in 2025 scored 461 points in reading, 463 in mathematics, and 482 in science. Reading has now declined for over a decade, falling most steeply since 2012; mathematics held up until 2018 and has fallen sharply since. Science, by contrast, has stayed broadly stable across this cycle.

The distributional story is the more troubling one. Roughly one in five 15-year-old students across the OECD -20%, now fails to reach a basic level of proficiency in all three core domains (science, mathematics and reading), up from 16% just three years ago in 2022. Achievement gaps tied to family background remain wide: socioeconomic status accounts for about 12% of the variation in science scores within countries, and 37% of disadvantaged students are low performers in science compared with 12% of advantaged students. Gender patterns diverged by subject-girls outperformed boys in reading by 30 points on average, boys outperformed girls in mathematics by 13 points, and in science the average gap was negligible but reappeared at the extremes, with girls ahead among low achievers and boys ahead among top performers.

At the top of the table, the same names appear as in previous cycles, with China's four benchmarked provinces/municipalities (reported as "B-S-J-Z") and Singapore trading first place across subjects: B-S-J-Z leads mathematics (612) and science (597); Singapore leads reading (535) and places second in both mathematics (563) and science (560), with Macao close behind in science (541).

How Did Sweden Perform in PISA 2025?

Sweden's numbers moved in the same direction as most of the OECD, but further than most. Fifteen-year-olds scored 466 in reading (down 21 points since 2022) and 464 in mathematics (down 18 points)-both the lowest results Sweden has ever recorded in this assessment. Science held closer to steady at 485, essentially unchanged from 2022 within the bounds of measurement error (some reporting puts the nominal shift at roughly 8 points). All three scores now sit close to the OECD average-not because Sweden improved relative to its peers, but because the OECD average fell alongside it.

Underneath the averages, the proportion of Swedish students failing to reach a basic level of proficiency jumped noticeably: 31% now fall below basic reading proficiency, up from 24% in 2022; 34% fall below basic proficiency in mathematics, up from 27%. Nearly one in five Swedish 15-year-olds (19.4%) is now a low performer across all three subjects at once.

Socioeconomic background continues to do heavy lifting: 49% of disadvantaged Swedish students fail to reach basic mathematics proficiency, versus 19% of advantaged students — a 30-point gap in the share of students below the floor. One genuine oddity in this cycle, also visible elsewhere in the OECD, is that advantaged Swedish students' scores fell by more than disadvantaged students' scores did. That narrows the gap in points terms even as the gap in who clears a basic bar stays enormous — a reminder that a shrinking average gap and a shrinking real disparity are not the same thing, and that this cycle's numbers deserve a second look once more data is available rather than a single tidy headline.

Within the Nordic region, the comparison is instructive: Finland still scores above Sweden in reading and science, Denmark is essentially tied with Sweden in reading and mathematics, and Norway and Iceland now score below Sweden across all three domains. Christina Månberg, Skolverket's acting director-general, put it plainly: "No single factor explains the decline; each nation must analyse causes at student, school and societal levels." Skolverket

Is Screen Time to Blame for Sweden's PISA Decline?

Swedish commentary has moved quickly to a single explanation: screens. It's not an unreasonable instinct, and the timeline is genuinely striking. Sweden mandated digital tools in its national preschool curriculum in 2019; by 2021, smartphones were near-ubiquitous in classrooms and, by teachers' own account, disruptive- roughly nine in ten Swedish teachers reported in prior surveys that smartphones harmed learning, stamina and attention, and nearly four in ten students reported being distracted by devices during mathematics lessons. By late 2022, Sweden's own schools minister, Lotta Edholm, was describing the country's digitalisation push as "an experiment" that "wasn't scientifically based and that harmed children's learning." Worldcrunch / After Babel. Sweden issued its first national screen-time guidelines in 2024, began restricting phones in classrooms, redirected funding toward printed textbooks and school libraries, and will complete a nationwide school phone ban in 2026.

A scientifically careful reading has to hold two things together, though. First, the 15-year-olds tested for PISA 2025 moved through their primary and lower-secondary years during exactly the period of heaviest, least-regulated digitalisation (roughly 2019–2023) -so this cohort's results plausibly reflect policy Sweden has already reversed, rather than the policy in place today. Second, Sweden's own education agency is right to resist a single-cause story: declining leisure reading, shortened attention spans, chronic teacher shortages, and shifting attitudes toward schooling more broadly are all cited by Swedish researchers as contributing factors, and similar declines, including among historically top-performing systems like Singapore, Japan and South Korea- occurred even where classroom technology policy looked very different. (The Local Sweden) OECD education director Andreas Schleicher's framing is worth keeping close: "reading is fundamental to how well pupils perform in other subjects" which suggests that whatever combination of causes is at work, rebuilding sustained, attentive reading is close to the root of the fix, not just one item on a list.

What Can Sweden Learn From Singapore, Estonia and Ireland?

No country's system transplants cleanly onto another's, and the differences in scale, culture and starting conditions between Sweden and, say, Singapore are real. But three consistent patterns separate this cycle's stronger performers from the pack, and each maps onto a specific, actionable gap in the Swedish debate.

From Singapore- coherence over quantity. Singapore's system is built around alignment: the curriculum, the way teachers are trained to teach it, and the way students are assessed all point in the same direction, and reforms since the 2000s have deliberately traded curriculum breadth for depth ("Teach Less, Learn More"). Its widely exported mathematics pedagogy moves students explicitly from concrete objects to pictorial representations to abstract symbols before expecting fluency — the opposite of introducing abstraction and hoping engagement follows. Teachers receive extensive initial training and ongoing, structured professional development as a career-long expectation, not a one-off. For Sweden's current debate — several parties are proposing better initial and continuing training specifically in reading instruction — this is the underlying principle worth generalising: pick a small number of things, teach them well, and invest continuously in the adults doing the teaching.

From Estonia - autonomy paired with trust, and technology as a means, not a goal. Estonia has been Europe's most consistent PISA performer for a decade, built on a national curriculum that leaves real day-to-day autonomy to schools and teachers, a comprehensive system with no early academic tracking, and a highly digitalised society that nonetheless treats technology as an instrument selected for a pedagogical purpose rather than a policy goal in itself. Analysts studying Estonia's results also flag a real vulnerability worth Sweden's attention: even a strong system erodes if teacher shortages, heavy workloads and low relative pay aren't addressed- a warning as relevant in Stockholm as in Tallinn. (European School Education Platform)

From Ireland- equity as an engineering problem, not just a value. Ireland now ranks second only to Singapore in reading and has measurably narrowed the mathematics gap between advantaged and disadvantaged students, through sustained, targeted funding and support aimed specifically at schools serving disadvantaged communities, alongside a long-running national literacy and numeracy strategy. Given that Sweden's own socioeconomic gap in basic proficiency (49% versus 19%) is one of the starkest figures in its 2025 results, Ireland's experience suggests that gap is not fixed — it responds to deliberately targeted, sustained investment rather than system-wide funding increases alone.

The common thread across all three: none of these systems got there through a single lever — a curriculum tweak, a technology ban, or a funding boost — but through several reinforcing choices held steady for years. Sweden's declines took roughly a decade to accumulate; its recovery is unlikely to be faster.

Why Aren't Reading and Math Scores Enough Anymore?

Here is the harder point, and arguably the more important one. Even a full recovery in reading and mathematics scores would leave an open question: are today's 15-year-olds being equipped for a world where a lot of factual and computational work is done with AI tools, not instead of them?

PISA itself has been moving, cautiously, toward this question. The 2022 cycle introduced an optional Creative Thinking domain, assessing students' ability to generate, evaluate and refine diverse ideas rather than converge on one correct answer. PISA 2025 goes further with a new innovative domain, "Learning in the Digital World," designed to measure how well students can self-regulate, apply computational and scientific-inquiry thinking, and build knowledge iteratively while working with digital tools — in effect, whether students can learn well in an environment saturated with information of wildly uneven reliability, including content generated by AI. Early commentary on this new domain has already surfaced a pointed tension: in at least one country studied so far, students posted a strong result in the digital-learning domain while still losing ground in the foundational reading and mathematics skills that make digital tools useful in the first place. (Karmactive) A student who can operate a digital tool fluently but cannot judge whether its output is true has not solved the problem this domain is trying to name.

That tension is the real argument for reframing what schools are for. Literacy and numeracy remain non-negotiable — Schleicher's point about reading underpinning everything else is well evidenced, and no amount of "21st-century skills" rhetoric substitutes for a child who can read a paragraph and do arithmetic. But treating creative problem-solving, critical thinking, source evaluation and the disciplined habit of checking AI-generated output as enrichment activities- nice if there's time left over- is no longer a defensible ordering of priorities. A generation that will do a large share of its future thinking alongside generative AI needs, as a core and assessed part of its education rather than an add-on: the instinct to ask where a claim comes from before repeating it; enough working knowledge of a subject to notice when a plausible-sounding AI answer is subtly wrong; the confidence to generate an original approach to a problem rather than accept the first output offered; and enough practice with ambiguity and open-ended tasks that "there's one right answer, look it up" stops being the default mental model. These are teachable, assessable skills -PISA's own two newest domains are proof that the field agrees they can be measured- and the Swedish debate, currently focused almost entirely on reading instruction and screen policy, has room to make them explicit curriculum goals rather than hoped-for side effects.

Recommendations: How Should Sweden Respond to PISA 2025?

For policymakers and school systems

  1. Treat the causes of the decline as plural and interacting - reading habits, attention, teacher supply, and technology policy together, rather than picking one explanation to legislate against; Skolverket's own caution on this point is well founded.
  2. Fund and staff Ireland-style targeted equity interventions for schools serving disadvantaged communities, since Sweden's socioeconomic gap in basic proficiency is now one of its starkest and most persistent results.
  3. Follow through on teacher supply and working-conditions reforms with the same urgency as screen-time policy, Estonia's own experts warn that even a strong, autonomous school system degrades under teacher shortages and excessive workload.
  4. Give schools genuine day-to-day pedagogical autonomy within a clear national curriculum, rather than managing instruction through detailed central mandates on tools or methods ,the Estonian and Singaporean systems differ sharply in how centralised they are, but both pair a clear framework with real classroom-level trust.
  5. Fold "Learning in the Digital World" and creative-thinking style competencies into the core curriculum and assessment, not an optional module, commit to teaching source evaluation, AI-output verification and open-ended problem-solving as explicitly as reading and arithmetic.

For schools and teachers

  1. Reintroduce sustained, uninterrupted reading time, for pleasure, not only instruction, since leisure reading decline is one of the most consistently cited factors behind falling scores across the Nordics.
  2. Use digital tools deliberately and sparingly, chosen for a specific pedagogical purpose (as in Estonia) rather than deployed by default; the evidence from Sweden's own 2022–2025 experience is that undirected device use in the classroom correlates with worse, not better, outcomes.
  3. Build in regular practice with genuinely open-ended problems, tasks with more than one reasonable path or answer, so that creative problem-solving becomes a habit of mind, not an occasional enrichment exercise.
  4. Make "check the source" and "verify the AI's answer" a routine, explicit step in assignments that involve any digital research or AI assistance, rather than an informal expectation.

For parents and guardians

  1. Protect daily space for reading and unstructured, screen-free problem-solving at home; Sweden's own 2024 national screen-time guidelines (for example, one to two hours daily for ages 6–12) are a reasonable starting benchmark if a household has none.
  2. When a child uses AI tools for homework, ask them to explain why the answer is right and how they'd check it , the habit of interrogating an answer, not just producing one, is the transferable skill.
  3. Treat curiosity and persistence with hard, ambiguous problems as achievements worth praising in their own right, alongside correct answers , the instinct to keep worrying at an open problem is exactly what standardised tests struggle to capture, and exactly what the emerging PISA domains are trying to measure.

The Bottom Line

PISA 2025 is a genuine warning for Sweden: two historic lows, a widening gap in who reaches even a basic standard, and a persistent socioeconomic divide that this cycle's numbers flatter slightly without closing. The screen-time reversal already under way is a reasonable response to real evidence, but it is one piece of a longer, decade-scale project that better-performing systems have pursued through teacher quality, targeted equity spending, and pedagogical coherence held steady over years, not one electoral cycle. And underneath the recovery project sits a second, less discussed one: making sure that by the time these reforms show up in a future PISA cycle, Swedish 15-year-olds haven't just relearned how to read and calculate, but have also learned to think originally, question confidently, and verify carefully in a world that will hand them answers, human and artificial alike, far faster than it hands them the truth.

Frequently Asked Questions About PISA 2025 and Sweden

What was Sweden's PISA 2025 score? Sweden scored 466 in reading, 464 in mathematics, and 485 in science. The reading and mathematics scores are the lowest Sweden has recorded in PISA's history, down 21 and 18 points respectively since 2022; science was essentially unchanged.

Did all OECD countries decline in PISA 2025? Most did. Reading and mathematics fell to their lowest-ever OECD averages (461 and 463), and countries as different as Sweden, Norway, Japan and South Korea all recorded lower scores than in 2022. Science was the exception, staying broadly stable across the OECD.

Which country topped PISA 2025? China's four benchmarked provinces and municipalities, reported as "B-S-J-Z," led mathematics (612) and science (597). Singapore led reading (535) and placed second in mathematics and science.

Is screen time the cause of Sweden's declining PISA scores? It's one plausible contributing factor, not a proven single cause. Sweden heavily digitalised its classrooms from 2019, then reversed course from 2022 onward after officials and teachers linked device use to falling attention and performance. But Skolverket and outside researchers point to several factors together — declining leisure reading, teacher shortages, and shifting attitudes toward school — and note that some countries with different technology policies saw similar declines.

What is the new "Learning in the Digital World" domain in PISA 2025? It's a new assessment area introduced in the 2025 cycle that measures how well students self-regulate, apply computational thinking, and build knowledge while working with digital tools — in effect, whether they can learn reliably in an information environment that includes AI-generated content of uneven accuracy.

What can Sweden do to improve future PISA results? Evidence from higher-performing systems points to a combination: sustained investment in teacher training and working conditions, targeted (not just system-wide) funding for disadvantaged schools, protected time for reading, deliberate rather than default use of classroom technology, and explicit curriculum space for critical thinking and source evaluation alongside literacy and numeracy. See the full recommendations above.


Sources


Webbyrå företag