AI in education is spreading faster than Europe’s teachers are being prepared for it, according to two reports published by the European Commission on 22 September 2026. The headline figure is blunt: only 29.7 per cent of EU teachers surveyed in the OECD’s TALIS 2024 study had taken part in professional development on the use of artificial intelligence for teaching and learning. Students, meanwhile, are already using the tools, with Eurostat data cited by the Commission showing that four in ten young people aged 16 to 24 used generative AI for formal education in 2025.
The first study, titled Friend or foe? Evidence from the use of generative artificial intelligence in learning and teaching, reviews research on students, teachers, parents and school systems. Its central finding is more nuanced than either enthusiasts or sceptics tend to admit. Generative AI can improve performance on tasks, but better performance does not automatically mean better learning. Students benefit when the tools support their own thinking and they engage actively with the content, and purpose-built educational tools tend to produce better outcomes than general-purpose chatbots. The risk, the paper warns, is over-reliance, with pupils handing off exactly the tasks that build knowledge and critical thinking.
The training gap is the part most likely to shape policy. Beyond the 29.7 per cent who had received AI-related training, 53 per cent of teachers said they had neither taken part nor felt a strong need to do so. That is a warning sign as much as a statistic, because it suggests that a large share of the profession does not yet see AI in education as part of its core job, even as pupils bring it into homework and classrooms. The report argues that teachers could use AI to prepare lesson plans, materials and assignments, but that doing so responsibly creates new demands that initial and continuing teacher education have not yet caught up with.
The second publication, a Eurydice report called Digital education at school in Europe 2026, maps how 38 education systems are handling the wider digital shift, drawing also on 24 systems that took part in the ICILS 2023 survey. It finds that almost two-thirds of systems now address AI through national strategies, guidance or policy frameworks, and that around one-third, mostly in Western and Southern Europe, include AI-related skills in their national curricula. Digital competence is now taught in lower-secondary education in most systems, and roughly three-quarters offer it as a dedicated subject.
Policy on paper, however, is not the same as practice in the classroom. The Eurydice report finds that schools do not have equal infrastructure, technical support or access to advanced tools, and that evaluation lags behind planning, with only a minority of systems building digital education into quality assurance. Pupils from disadvantaged backgrounds may have fewer chances to develop digital skills and can struggle to learn at home without reliable devices or internet access. In other words, generative AI risks widening gaps that the pandemic already exposed.
Commissioner Roxana Mînzatu framed the findings as groundwork for action, saying that digital tools must provide tangible support for learning. The evidence will feed into the Union of Skills agenda and the forthcoming Education Package, which the Commission says will address basic skills, teachers and trainers, and digital education infrastructure.
The test for that package will be whether it treats teacher training as the priority it now appears to be. Education remains a national competence, so Brussels can fund, coordinate and set benchmarks but cannot write curricula. What it can do is put money behind continuing professional development through Erasmus+ and push for the long-term, representative research the Friend or foe paper says is missing. Without that, AI in education in Europe will be shaped less by schools and ministries than by whichever tools pupils happen to download.
