AI can write the answer—but what does the student learn?

Two Commission reports presented on September 22 indicate that the educational benefit of generative AI depends on how it is used, while questions remain about lasting skills. Teacher preparation and institutional rules matter.

Editorial analysis: A better assignment does not necessarily demonstrate better understanding. If a system does the reasoning for the learner, the finished work reveals little about independent knowledge. A student checking the generated answer, finding mistakes and explaining corrections performs a different task.

For Hungarian education and workplace training, designing the exercise is therefore as important as buying the tool. Performance with assistance should be assessed separately from knowledge that participants can subsequently apply unaided.

The reports do not prove that a selected product improves outcomes in every classroom. Comparable exercises, clear assessment rules and follow-up can provide evidence for a local deployment. The next question is not simply how often the tool is used, but what knowledge remains.

If improvement appears only with the tool enabled, the institution must assess whether that meets its learning objective. The quality of generated work is not a complete measure of the participant's ability.