The AI experiences in education that stand out in Europe

What the best AI education experiences in Europe have in common

A French girl practices reading with exercises that automatically change according to her mistakes. In a Czech school, secondary students learn civic education by talking with an artificial intelligence inspired by Václav Havel while checking whether its responses contain false information. In the Netherlands, sensors integrated into Montessori materials allow teachers to observe in greater detail how primary school students solve math problems.

All of these experiences are part of Good Practice in AI for Education: Spotlight on EU Case Studies and Insights, a study commissioned by the Directorate-General for Education, Youth, Sport and Culture of the European Commission (DG EAC), which identified 209 practices involving the ethical and responsible use of artificial intelligence in education through document reviews, surveys, and interviews conducted between January and October 2025. From that group, the report selected fourteen experiences for deeper analysis in order to explore what results they are achieving, what problems they are trying to solve, and which conditions seem to support their success.

Although the study includes university and vocational education initiatives, many of the most interesting experiences take place in basic education. The result gives us concrete evidence about real uses in classrooms and training centers and, based on this evidence, allows us to infer patterns of implementation.

Best practices: what is already working in European classrooms

As we have already explained, the European Commission report identified more than 200 experiences and analyzed 14 of them in depth. The initiatives are very different from one another: some focus on adaptive reading in primary school; others on virtual assistants, simulations, or teacher support tools. The educational levels, technologies used, and implementation contexts also vary. But they all share one feature: they try to solve very specific and measurable pedagogical problems. In this section we will examine five of them.

A platform that detects reading difficulties before they become school failure

Lalilo was created to address a very specific problem: the enormous differences in reading skills with which many students arrive in the first years of primary school. The platform, used in French schools, automatically adapts phonics, word recognition, and comprehension exercises according to each student’s performance. If it detects specific difficulties, it modifies both the type and sequence of activities.

One of the most interesting aspects of the project is that personalization is not presented as a substitute for the teacher, but as a monitoring tool. Teachers receive dashboards with detailed information about each student’s progress and can intervene before problems accumulate.

The report also cites an experimental study conducted in 2023 that found significant improvements in reading fluency, especially among students who needed the most support. Teachers highlight another less quantifiable but important element: many children use the platform as an engaging activity rather than as remedial instruction.

The chatbot that teaches civic education… and also how to question AI

How can critical thinking about artificial intelligence be taught using artificial intelligence? DigiHavel attempts to answer that question through civic education. The tool, developed in the Czech Republic, uses a chatbot inspired by the historic Czechoslovak leader Václav Havel to work on topics related to democracy, human rights, and citizenship.

But the goal is not simply to talk with AI. Students also learn to verify information and detect errors or “hallucinations” generated by the system. In fact, one of the project’s central ideas is that students understand that artificial intelligence can produce responses that sound convincing while still being incorrect.

The design of the tool reinforces that critical approach. Instead of building a hyper-realistic representation of Havel, the developers deliberately chose a caricature that constantly reminds users that this is a simulation.

The project also incorporates accessibility functions, such as text-to-speech reading and English-language support for Ukrainian refugee students.

Sensors and virtual reality to better understand how children learn

In the Netherlands, the NOLAI laboratory works on educational AI projects in a rather unusual way: the ideas do not come from technology companies, but from problems identified by teachers and schools. From there, researchers, teachers, and developers jointly design the tools.

One of the projects uses brief virtual reality sessions to help primary school students expand vocabulary through immersive experiences. Children can virtually “visit” environments that are difficult to reproduce in the classroom and develop language skills through concrete contexts.

Another pilot incorporates sensors into Montessori math materials to observe how students interact with tasks and provide more detailed information about their learning processes.

The project is especially interesting because of its teacher co-design approach and its commitment to open-source tools. The report also highlights that all projects must include plans for scalability and ethical data management from the beginning, something that is still relatively uncommon in many educational AI initiatives.

The Austrian pilot that made teacher training the center of its AI strategy

While many educational AI projects focus directly on tools for students, Austria chose to begin somewhere else: teacher training. The KI-Pilotschulen program, promoted by the Ministry of Education, selected 114 schools to experiment with different AI platforms and analyze how pedagogical practices change when teachers receive specific support.

The project included tools for content generation, lesson planning, and teaching support, all of them previously reviewed to ensure compliance with European data protection regulations.

The initiative was part of a broader national strategy that also incorporated large-scale AI training courses for teachers across the country. According to the report, more than 350 schools applied to participate in the program.

One of the most relevant findings was that teachers’ confidence and familiarity with technology directly influenced the quality of implementation.

The Italian pilot using AI to personalize feedback and monitoring

In several Italian secondary and vocational schools, students began working with virtual assistants designed to provide personalized feedback and continuous support in STEM subjects and Italian language learning. The project, promoted by the Ministry of Education and Merit, combined different AI tools integrated into platforms already commonly used in classrooms.

The goal was not simply to automate exercises, but to create more individualized learning pathways within large groups. Some tools helped generate personalized exercises; others monitored progress or provided support during independent practice.

The report especially highlights the effort devoted to teacher training. Participating schools organized courses, shared sessions, and spaces for exchange among teachers before incorporating the tools into everyday practice.

The first comparative results were striking: classes using AI systematically showed better academic results than groups working with innovative methodologies without artificial intelligence support.

What the most successful experiences have in common

Beyond the differences between tools, countries, and educational levels, the experiences analyzed by the European Commission are beginning to reveal some shared patterns. The strongest initiatives do not seem to be defined so much by technological sophistication as by the clarity of the educational problem they are trying to solve and by the way they integrate AI into broader pedagogical strategies.

Taken together, the practices make it possible to extract several interesting lessons about where artificial intelligence appears to add the most value in education and under what conditions it does so.

AI works best when it responds to very specific educational needs

The most interesting experiences in the report do not start from the technology itself, but from specific problems: reading difficulties in primary school, lack of opportunities for practice, language barriers, or the need for individualized monitoring. Artificial intelligence appears subordinated to clear pedagogical objectives rather than added superficially into the classroom.

Practice and continuous feedback are among the most promising uses

Many tools show especially interesting results when they expand opportunities for rehearsal, simulation, or feedback. Adaptive platforms adjust exercises according to student progress; virtual assistants provide continuous support; conversational simulations allow students to practice complex situations in safe environments. In several cases, the educational value seems to arise less from content delivery than from the possibility of practicing more and receiving more immediate guidance.

The role of the teacher remains decisive

The most stable experiences incorporate specific training, support, and active teacher participation. In the Netherlands, several projects emerge directly from needs identified by teachers. Austria turned AI training into one of the pillars of its national strategy. Rather than reducing the importance of teachers, many initiatives seem to increase the need for pedagogical mediation and contextual adaptation.

AI literacy is beginning to become part of learning itself

Some of the most interesting practices do not use artificial intelligence only as a support tool, but also as an object of critical analysis. Several experiences incorporate exercises in verification, error detection, or reflection on the limits of generative systems. The objective is not only to learn with AI, but also to understand how it works, when it can fail, and why its answers need to be questioned.

Utility matters more than technological sophistication

The report also points to a less obvious conclusion: the educational value of AI does not necessarily seem to depend on particularly complex tools. Some of the most promising practices use relatively simple technologies — conversational chatbots, adaptive platforms, or support assistants — integrated very concretely into well-defined pedagogical dynamics. In many cases, what matters is not technological spectacle, but the tool’s capacity to respond usefully and consistently to real classroom needs.

Knowing how to integrate AI where it works

After years of debates dominated by hypothetical scenarios, the experiences collected by the European Commission make it possible to observe how artificial intelligence is beginning to be used in real educational contexts. The report shows a landscape characterized by fairly pragmatic attempts to solve persistent educational problems: how to provide more feedback, how to better personalize learning, how to expand opportunities for practice, or how to support students who are often left outside more standardized models of teaching.

The initiatives analyzed remain heterogeneous, experimental in many cases, and still limited in scale. But taken together they offer an interesting conclusion. AI appears to add the most value when it stops occupying the center of the conversation and instead becomes integrated as one more tool within broad pedagogical strategies, carefully designed and guided by clear educational objectives.

Perhaps that is one of the most relevant lessons emerging today from European classrooms. The future of artificial intelligence in education will probably depend less on the technological sophistication of the tools than on the ability of schools and education systems to use them where they genuinely succeed in improving something important about learning.