Large language models in school education: emerging evidence

Large language models (LLMs) have entered European classrooms more quickly than any educational technology before them,and outpacing the capacity of schools to guide their use. In 2025, 63.8 % of 16–24-year-olds in the EU reported using generative AI (GenAI), with 41 % of this age group using GenAI for formal education (Eurostat, 2026). A survey of 12–17-year-olds across seven European countries found that nearly half of them had used ChatGPT, half of them at their teacher’s instruction (Vodafone Foundation, 2025). The policy question is therefore not if LLMs should enter secondary education; they already have. The question this report addresses is under what conditions LLMs support learning, and under what conditions their use undermines it.

 

Authors:

Stéphanie Crêteur, PPMI (Part of the Verian Group)

Please cite as:

European Commission: Directorate-General for Education, Youth, Sport and Culture and Crêteur, S., Large language models in school education – Emerging evidence – ENESET ad hoc report, Publications Office of the European Union, 2026, https://data.europa.eu/doi/10.2766/7329817