The latest PISA results carry a stark warning about learning outcomes across much of the world: reading and mathematics performance has fallen to the lowest OECD-average levels the assessment has recorded, and one in five 15-year-olds across the OECD now fails to reach baseline proficiency in all three subjects.
India has no score in this cycle. It did not sit PISA 2025.
That makes this round less useful for asking where India stands on a single global scoreboard, and more useful for a different question: what lessons are relevant to an Indian education system still building its own responses through NEP 2020, NIPUN Bharat and NCERT’s PARAKH framework.
The Numbers
The OECD’s Programme for International Student Assessment tested around 760,000 15-year-olds across 91 countries and economies in 2025.
| Metric | 2015 → 2025 | 2022 → 2025 |
|---|---|---|
| Reading, OECD average | −28 points | −14 points |
| Mathematics, OECD average | −22 points | −9 points |
| Science, OECD average | gradual decline since ~2009 peak | −3 points (not significant) |
One in five 15-year-olds across the OECD now fails to reach baseline proficiency in all three subjects, up from 16% in 2022.
Five findings for Deans, Principals, School Founders and Teachers
This is the part of PISA that rarely makes headlines but is the most operationally useful for anyone running a school, a university department or a teacher-training programme.
1. AI use for schoolwork shows a non-linear relationship with performance. PISA’s first-ever question on AI chatbot use for schoolwork found that non-users outscore users on average, but among users, moderate use beats both occasional and heavy use. More relevant for policy: opportunities to build structured AI literacy alongside AI access are concentrated among already-advantaged students. A school or system rolling out AI tools without a parallel AI-literacy curriculum is, per this data, more likely to widen an existing gap than close one, a direct caution for any Indian classroom-AI mandate that treats device access as equivalent to capability.
2. Growth mindset and classroom engagement track together, and both skew toward advantage. Nearly four in five academically resilient students, those who perform well despite a disadvantaged background, report a growth mindset (a belief that intelligence can be developed). Separately, students with a growth mindset are about three times more likely to perceive opportunities for social mobility than those with a fixed mindset. The report treats both as correlations, not proven intervention effects. Curiosity, perseverance, help-seeking and sense of belonging show a similar skew toward advantaged students in most systems, not just toward high scorers, meaning the engagement gap opens before the score gap does. Gender cuts differently: girls report more curiosity and help-seeking, boys report a stronger sense of belonging.3. Bullying and school climate: cyberbullying is rare but is associated with the largest performance gap. Twenty percent of OECD students report being bullied at least a few times a month, up since 2022 but still below 2018. Cyberbullying itself is comparatively rare, at 3%, yet victims scored 59 points lower in science than their peers on average, and 48 points lower after accounting for socio-economic profile, the largest gap tied to any climate factor measured. Relational bullying (rumour-spreading, exclusion) rose faster than physical or verbal bullying over the last three years. That argues for anti-bullying attention shifting toward digital and social exclusion, not just visible incidents.
4. Staffing shortages are easing on paper, but not where it matters most. Principals reported fewer teacher shortages in 2025 than in 2022, and student-to-teacher ratios improved in 36 systems, though improved ratios can also reflect falling enrolment rather than more hiring. Principals in most systems remain more concerned about staffing than in 2018, and worry about the adequacy and training of assisting staff (aides, counsellors, special-needs support) actually increased. Shortages stay consistently worse in disadvantaged and public schools than in advantaged and private ones.
5. Teaching practice and disciplinary climate show links with more stable performance. Disciplinary climate in science lessons improved since 2015 in 37 systems, most sharply in Korea and Croatia, and systems with bigger gains tended to show steadier or improving science scores, an association, not a demonstrated cause. Separately, teacher support has risen since 2015 in about half of systems (72% to 75% of OECD students report their teacher helps them regularly), but fewer students in 2025 report their teacher “continues teaching until they understand,” and systems that held onto that specific practice showed more stable scores than those where it fell away. Of the findings in the report, this comes closest to identifying a teacher-level practice that schools could potentially strengthen and that has a measurable association with outcomes.
The reading decline, more specifically
The drop concentrates in longer texts and higher-order tasks (evaluating claims, cross-referencing sources); single-fact retrieval held up better. The share of “hasty readers” (fast but wrong) nearly doubled, from 7% in 2018 to 11% in 2025, while accurate, fluent readers fell about seven percentage points. Self-reported test effort also declined. The report’s own read leans toward flagging sustained attention, not a collapse in underlying literacy, a distinction with direct bearing on how Indian foundational-literacy programmes define and test “reading.”
What other developing education systems tell us for the Indian debate
India has no PISA 2025 data point, and nothing here should be read as an estimate of where India would score. A set of lower-middle-income systems, chosen for population scale and development stage rather than any formal matching to India’s economic profile, did sit the test, and their results offer a reference point for the kind of pace and pattern of change a large developing school system can show over three years.
| Country (illustrative reference, not a matched comparator) | Science | Reading | Mathematics | Change 2022→25 (sci/read/math) | Low performers, all 3 subjects (%) |
|---|---|---|---|---|---|
| Viet Nam | 457 | 392 | 443 | not trended* | — |
| Jordan | 407 | 363 | 388 | +32 / +20 / +26 | 48.1 |
| Indonesia | 389 | 365 | 364 | +6 / +7 / −2 | 60.9 |
| El Salvador | 385 | 366 | 346 | +12 / +2 / +3 | 61.3 |
| Cambodia | 382 | 347 | 366 | +35 / +18 / +29 | 61.7 |
| Philippines | 373 | 367 | 371 | +17 / +20 / +16 | 64.3 |
| Morocco | 358 | 321 | 355 | −7 / −18 / −10 | 73.2 |
| OECD average | 482 | 461 | 463 | −3 / −14 / −9 | 19.7 |
*Vietnam changed testing mode this cycle; the OECD flags its trend data as not comparable.
This is an illustrative peer group, not a matched comparator. India differs from every system listed in population, schooling structure, language diversity and sampling history, so these scores provide context rather than a proxy for where India might perform. Two of these systems, Cambodia and the Philippines, along with Jordan, are improving faster than most OECD systems are declining, which the report attributes partly to expanding school access, a rough external reference point for NIPUN Bharat’s own foundational literacy and numeracy targets.
Money vs. outcomes
National wealth explains only about half the cross-country variation in scores. Luxembourg, the top spender at USD 288,521 per student, scored below Ireland, France and Italy, each spending under half that.
Equity: narrowed, but not for the reason it sounds
Socio-economic gaps narrowed between 2022 and 2025 largely because advantaged students’ scores fell, not because disadvantaged students’ scores rose. The OECD classifies a system as “high fairness” when socio-economic status accounts for under about 12% of the variance in science scores; B-S-J-Z (China), Canada, Estonia, Ireland, Japan, Korea, Lithuania and Macao cleared that bar while also posting strong overall scores, a combination the report treats as rare.
For context: who’s at the top
| System | Science | Reading | Mathematics |
|---|---|---|---|
| B-S-J-Z (China) | 597 | 527 | 612 |
| Singapore | 560 | 535 | 563 |
| Macao (China) | 541 | 501 | 549 |
| Chinese Taipei | 540 | 508 | 546 |
| Japan | 538 | 503 | 525 |
| Korea | 526 | 501 | 522 |
| OECD average | 482 | 461 | 463 |
Steepest three-year declines: Guatemala, Hong Kong and Israel led in science; Israel, Latvia, Malta, Serbia and Slovenia in maths; Latvia, Israel, Guatemala and Malta in reading, several exceeding 20 to 30 points in a single subject, more than a typical year of learning gain for a 15-year-old. The OECD counts eight systems that improved in all three subjects; it names Türkiye and the UK as established performers extending already-strong results, and Cambodia, El Salvador, Jordan and the Philippines as lower-base systems catching up, in part through expanding school access.
India: the absent data point
India appears nowhere in the 377-page Volume I report. India’s 2009 participation was limited to Tamil Nadu and Himachal Pradesh, both of which recorded very low scores by international standards, not the country as a whole. Chandigarh sat the 2022 cycle separately, as a standalone subnational entry, not as a national return. India did not participate in 2025.
That leaves NEP 2020’s foundational-literacy push, NIPUN Bharat’s targets and NCERT’s PARAKH framework without a current PISA benchmark covering the full 15-year-old population across these three domains, not without any external benchmark at all; India continues to run its own national and state-level assessments (NAS, ASER) in parallel. The gap is specifically the absence of an internationally comparable data point at the scale and age group PISA measures.
Source: OECD (2026), PISA 2025 Results.
