The European Commission published a study on how generative AI is actually being used in secondary schools across five EU countries.
121 people participated: teachers, students, school leaders, teacher educators, and policymakers from Finland, Germany, Ireland, Luxembourg, and Spain. All early adopters. All with direct experience using GenAI in education.
Here’s what jumped out to me.
Students are ahead of their teachers. Across all five countries, students reported using GenAI more frequently and more naturally than their educators. They’re using it to practice foreign language conversations, generate summaries, simplify difficult concepts, simulate exams, brainstorm creative projects, and build flashcards. Some already treat it as a personal assistant for both schoolwork and daily life.
And they’re not naive about it. Students in the study raised their own concerns about becoming too dependent on AI. One student in Ireland put it perfectly:
“we do learn by doing. You can read 1000 English essays, but still, we have to be able to write one. If you don’t practise writing it, you’re not going to learn anything. And if the AI’s doing it for you, you’re not learning. And on the day of the Leaving Cert [exam] you don’t have the ChatGPT in your pocket to pump out, whatever, that Platt, or that Yeats poetry. So, you have to write.”(p. 19)
That’s a level of self-awareness that many adults in education haven’t reached yet.
Now here’s the part that should concern us. Teachers across all five countries said the same thing: they feel underprepared. Professional development on GenAI was either scarce or nonexistent. Initial teacher education programs had barely started to acknowledge GenAI exists. In Germany, teachers predicted it would take years before GenAI finds its way into curricula and training programs. Years.
I’ve said this before and I’ll say it again. The biggest risk to students right now is not AI. It’s teachers who don’t know how to guide them through it.
The study also found something I’ve been arguing for a long time: GenAI is forcing a rethink of assessment. Traditional essay-based assignments are becoming unreliable as measures of learning. A school leader in Germany described a shift from product-oriented to process-oriented assessment, where you evaluate how students think and work through problems, not just the final output. That’s competence-based education finally getting its moment.
One Irish teacher shared a story that perfectly captures what thoughtful AI integration looks like. She built a “paragraph writing buddy” to help struggling students get started with their writing. When one student remained disengaged, she created a “smart Alec version” that gave sassy responses. That got him writing. Same learning outcome, different support. That’s personalization.
Another Irish teacher described how she uses GenAI for feedback. She reads through a student essay, records rough spoken notes on her phone with all her “butts and stutters,” then asks the AI to clean those notes into structured, constructive feedback. The expertise is hers. The efficiency comes from AI.
These are not futuristic scenarios. These are teachers doing this right now, in real classrooms, with real students.
The study also raised concerns that don’t get enough attention. Students in Spain noted that GenAI is eroding collaboration. Before AI, they’d turn to classmates for opinions and ideas. Now everyone just asks ChatGPT, and “no one really has their own original idea anymore.” That’s a social skills problem hiding inside a technology problem.
And a STEM teacher in Spain made a point that stopped me:
if we are not training or educating our young people first [in the good use of GenAI], it’s like giving them a gun and then telling them don’t hurt them. If [as teenagers] they are not able to control their emotions, they are not able to control their day-to-day life: how are they going to be able to manage their emotions, their reality with tools as powerful as artificial intelligence? It’s a bit of a strong topic, but that’s how it is. (p. 31)
Strong language. But fair.
The report’s policy recommendations are worth reading in full, but the core message is clear: AI literacy needs to be in the curriculum now, teacher training needs a complete overhaul, and schools need actual infrastructure and guidelines to support this work.
What struck me most about this report is the gap between what students are already doing and what schools are prepared to handle. Students didn’t wait for permission. They didn’t wait for guidelines. They started using AI the moment it became available. And many of their teachers are still trying to figure out what ChatGPT is.
That gap is where the real damage happens. Not from AI itself, but from the absence of adults who know how to help young people use it well.
Reference
Villar Onrubia, D., Cachia, R., Rietz, C., Feltrero, R., Niemi, H., Hallissy, M., & Reuter, R. (2025). Generative artificial intelligence in secondary education: Uses and perceptions from the perspective of early adopters across five EU Member States. Publications Office of the European Union. https://publications.jrc.ec.europa.eu/repository/handle/JRC144345