Interpreting the results of the PISA 2025 report

I would like to comment on a few aspects of the report relating to technology and education, as educational technology is my field of research and work.

When discussing PISA, it is essential to begin with a methodological clarification that is often overlooked in the analyses carried out. PISA is an assessment conducted by an external body, the OECD, using questionnaires and assessment tests. Specifically, a large proportion of the questions relating to technology (the use of resources and AI) are based on perceived data, rather than direct observation of actual teaching practices in the classroom. These limitations are inherent in any research based on self-reported surveys and are not a criticism of this aspect of PISA per se, but rather methodological characteristics that define its strengths and weaknesses.

PISA is excellent at providing descriptive data that enable international comparisons, identify patterns and trends over time, and serve as a starting point for suggesting more in-depth research questions. However, PISA is insufficient for establishing clear causal relationships between the use of technology and academic outcomes, for explaining the underlying mechanisms of why something works or does not work, or for reaching definitive conclusions about the effectiveness of specific education policies.

With regard to the impact of technology on education, various data points are presented. On the one hand, the capacity of the school’s digital environment to improve teaching and learning is assessed by the headteachers who have provided their views. They are asked about the availability of technology at the school and their opinion on the teaching skills of the staff, amongst other things. Headteachers at Spanish schools express a less favourable view of their schools’ ability to improve teaching and learning through digital resources. Headteachers at state schools perceive this more negatively than those at private schools in Spain.

On the other hand, pupils are asked about the number of hours spent using technology, distinguishing between learning activities and leisure activities. Once again, the data is interesting, but the concept of ‘learning activity’ as self-assessed by a student does not distinguish between fundamental aspects such as the methodological use of the tool. Reading a digital textbook is not the same as programming a robot or developing a telecollaborative project; yet all these tasks can be framed as learning activities.

We know from research in this field that simply knowing the number of hours is not enough to conclude that it is the use of technology that determines performance, as it does not take into account the type of task being carried out or the teacher’s pedagogical approach. We lack key information regarding the methodological context: we do not know the type of lessons the pupils receive, the specific tasks required of them, the pupils’ own digital competence, or the specific technological resource being used.

As for AI, the report attempts to define the type of task to some extent, but even so, the approach is very general and it cannot be claimed that AI contributes to poorer performance. The questions raised by the report regarding artificial intelligence overlook the fact that the determining factor is the teacher’s pedagogical intent and the design of the task. Using AI in the classroom as a guided tool (for example, to check for errors in a model, analyse biases or design a project under supervision) is not the same as relying on it to independently complete homework, where it is often used merely as a shortcut.

For all these reasons, the PISA data are extremely interesting, but must be analysed with caution. Although the report presents findings that suggest certain correlations, the type of question and the methodological approach do not allow for a clear cause-and-effect relationship to be established, particularly in a context such as education, where the didactic aspect—which is one of the key elements of educational research—has not been taken into account.

The fact that the decline is greater amongst families with a higher socio-economic and cultural status in many countries highlights the need to investigate what has changed within this group. Much emphasis is placed on the fact that these households may have more technological devices, but I believe we should also explore the living conditions of this group.

The data also indicate that adolescents are receiving less support at home, and this is consistent with something we already observed in the 2022 wave: the working conditions of university-educated mothers are deteriorating. In other words, an indirect effect of the 2009 crisis has been to make work-life balance more difficult and, consequently, to reduce family support. The crisis did not affect the first PISA cohorts, as those adolescents had already been raised, but today’s adolescents have grown up in the ‘new normal’: upper-middle-class families with less free time.

Until 2015, PISA revealed a divide between northern and southern Spain, reflecting a long-standing trend of educational disparities dating back at least to the 19th century. Since then, a new group has emerged: bilingual regions. It would be worth studying in greater detail why, prior to 2015, several of these regions performed above average (the Basque Country, Navarre and Catalonia) or close to average (the Valencian Community) and have since seen their results decline. The obvious explanation—language—does not account for the results prior to 2015, and the impact of immigration has already been examined; whilst it is a factor in the decline, it is not the sole cause. As this is a politically contentious issue, it requires a calm and well-founded study based on an analysis of the specific circumstances in each region, rather than merely a superficial examination of the PISA data.

There are many and varied reasons behind these results. Furthermore, it must be borne in mind that, in Spain, as in neighbouring countries, decision-making on education policy is shared between central government (25 per cent), the autonomous communities (60 per cent) and schools (15 per cent). Local authorities have no decision-making power, although they do provide resources and specific programmes.

The decline in reading skills is often attributed to screen time and, more recently, to artificial intelligence (AI). This may well be true. However, what is most worrying is the reading standard amongst the adult population, which fails to set a good example or serve as a role model for children and young people. It is important to remember that reading is a skill that develops over the long term, beginning at a very early age; therefore, the causes cannot be attributed solely to what has been done in the last three to four years, but rather we must look at how reading is encouraged and developed in early childhood education and, of course, in primary education.

As for mathematics and science, which are instrumental subjects in terms of the world of work – again, according to the OECD – we must look at how many and what types of pupils will choose these subjects at sixth form and university level. With regard to basic education for all, it is worrying that there are no firm policies in this area; just as with reading, these skills are not learnt and consolidated in the last three to four years, but scientific and mathematical thinking can be developed from a very early age.

An inclusive school is not an obstacle, but a challenge, one that is consistent with democratic and inclusive societies. To suggest otherwise would be to go against a democratic and participatory model of society and schooling.

For its part, the LOMLOE has nothing to do with outcomes, since its implementation will only extend over a few years; moreover, I would emphasise that educational policies are far broader in scope than anything a single law can define. Its implementation falls to the regional governments and, ultimately, to the schools themselves. In this regard, we need to address two issues without delay: the lack of resources and the decline in public and administrative confidence in the teaching profession. First and foremost, we must engage the political class so that, across the government, coalition partners and the opposition, they can unite in their commitment.

The problem is a fundamental one, but we must indeed demand [of those responsible for education policy in the national government and the autonomous communities] that they lead the implementation of coherent, robust and sustainable policy programmes in the medium and long term.

And one final point: PISA does not measure the quality of the education system; it merely measures those three skills. The OECD, as an economic organisation, overlooks many other skills taught in schools: in addition to physical education, the humanities and the arts, these include values, social skills and the ability to navigate the world.

The latest PISA results have already sparked numerous reactions, with people attributing the decline to reading habits, a lack of discipline, the LOMLOE or, of course, screens and AI. Simplifying matters in this way may make for a snappy headline, but not if we wish to engage in a serious and measured debate. Reducing everything to a single figure obscures what is essential: embracing the complexity of an issue as multifaceted as education.

Before discussing the results, it is worth asking who conducts this test and for what purpose. PISA is designed by the OECD, which is an economic organisation, not an educational one. Its objective, from a neoliberal perspective, is growth and competitiveness, and its view of schools is in line with this, as it assesses pupils as future workers and countries as economies competing against one another, channelling that competition into the familiar rankings. Preparing pupils for the labour market is one of the functions of schools, according to the tradition of the sociology of education, but it is by no means the only one, even though we have been treating it as such for years.

We must also be cautious about what the test measures and how. PISA shows that teenagers read less well, and this is true in terms of what it assesses. However, the 2025 Reading Habits Barometer states that the 14–24 age group is the most avid of readers in Spain: 76.9 per cent read books in their spare time, the highest percentage across all age groups and six percentage points higher than in 2017. In other words, they do read, and quite a lot. What has changed is not the habit itself, but the ability to sustain attention on long and complex texts as measured by the test – which is only one aspect of what reading entails. It is worth bearing this in mind because the narrative that young people do not read fuels moral panic.

These results will be used by certain quarters to call for a return to content-based schooling. But they are using an exam as evidence that does not actually measure what they claim: PISA assesses whether pupils apply what they know to a real-life situation, not whether they have memorised it. In fact, the competence-based approach that is currently being criticised originated largely from the OECD itself, which points out that the systems that fare best are not those that prioritise discipline or content, but those that cover fewer subjects, yet in greater depth. There is also no shortage of voices attributing the results to Spanish education legislation, but the figures had already begun to ‘deteriorate’ before its implementation, and the decline is widespread across the OECD.

Other reactions unanimously blame mobile phones or AI. Yet those who have seen the sharpest decline in Spain are pupils from better-off families: they have lost 33 points in reading, compared with 15 for the most disadvantaged. The OECD itself admits that social gaps have narrowed because those at the top are falling behind, not because those at the bottom are catching up. Neither mobile phones nor AI can explain this. Nor does this anti-technology narrative explain why Estonia – one of the most digitised systems in the world – is the OECD country with the best results in science after Japan, or why Japan and South Korea, with their hyper-connected teenagers, top almost every table. The decisive factor is not how much technology is used, but for what purpose and with what support.

Finally, the headlines won’t highlight it, but there are variables where Spain ranks among the top, whilst high-performing countries lag far behind: a sense of belonging to the school community, social cohesion, very low levels of bullying, perseverance and curiosity. This is what the rankings do not capture, and it is a large part of what a school does every day. Without a doubt, it should also be a priority objective of education.