European Commission examines classroom AI: better task performance is not always better learning

Editorial recommendation

What to do

Review school AI guidance, task design, evidence of learning, teacher development and equitable access. These are editorial recommendations, not new European legal requirements.

Verified facts

What was published and what it does

On 22 September 2026 the European Commission presented “Friend or foe? Evidence from the use of generative artificial intelligence in learning and teaching” and the Eurydice report “Digital education at school in Europe 2026: Bridging gaps in access, teaching and learning”. They provide research and policy evidence; neither changes Spanish law or imposes a new school or university protocol.

The Eurydice report draws on policy information from 38 education systems and ICILS 2023 data from 24 systems. The Commission says the findings will inform its forthcoming Education Package, which has not yet been adopted.

European Commission overview and findingsGenerative AI paper · EU Publications OfficeEurydice school digital education report

Verified facts

Findings for education leaders

• Eurostat data cited in the paper show that four in ten EU residents aged 16–24 used generative AI for formal education in 2025. This is an EU-wide figure, not a Spanish or school-only rate.
• AI may improve task performance without improving learning. Students benefit more when they examine and work with AI output; delegating cognitive steps may harm knowledge development.
• Only 29.7% of EU teachers surveyed in TALIS 2024 had received professional development on AI for teaching and learning.
• Eurydice finds widespread digital planning but uneven infrastructure, technical support, home access and evaluation. Nearly two-thirds of the systems examined address AI through strategies, guidance or policy frameworks, while school-level use remains uneven.

Official summary of findings

Editorial analysis

Implications for schools and universities in Spain

For school leaders, the useful question is not simply whether pupils use AI, but where, with what supervision, and what evidence demonstrates learning. The research does not establish that any single tool improves outcomes for every learner.

The European comparison can inform purchasing and training decisions: devices and connectivity can be necessary, but they do not replace teaching support, technical assistance or a review of access gaps. Universities may draw on the AI evidence when reviewing assessment and academic integrity, although the Eurydice report focuses on schools.

Editorial recommendation

A proposed review for education leaders

These are prudent editorial suggestions. The reports themselves impose no new legal duties on institutions.

• Set out permitted AI uses by age, subject and task, and explain how assistance should be disclosed and checked.
• Use process evidence, oral explanation or draft review where a generated answer alone cannot show learning.
• Train staff to select tools, spot errors and bias, and protect pupil data; review vendor terms and assess impacts where appropriate.
• Measure access, support and outcomes before wider rollout; follow the Commission’s forthcoming Education Package and any resulting Spanish or regional measures.