A student can use an AI assistant to produce an answer or to work toward understanding one. Those activities may look similar on a screen, but they create different opportunities for learning. The useful question for a classroom is what the student will be able to explain or do after the tool is put away.
UNESCO's guidance on generative AI in education emphasises human agency, appropriate oversight, inclusion, and privacy. It offers a reason to begin with educational purposes rather than the novelty of the tool. A lesson should not need an AI component unless that component serves a clear learning objective.
Consider a student studying fractions. Instead of requesting a completed worksheet, the student could ask for one worked example, attempt a similar problem independently, and compare the reasoning. A teacher might then ask the student to explain why the method works. This is a proposed classroom routine, not evidence that an AI tutor will always teach the concept correctly.
Another useful role is generating material for critique. A teacher could present an AI-produced explanation and ask students to identify missing steps, ambiguous language, or unsupported claims. The activity turns the output into something to examine. It also gives students practice separating a confident tone from a convincing argument.
The same care applies to feedback on writing. A learner might request comments on the clarity of an introduction, then decide which suggestions preserve the intended meaning. Asking the tool to rewrite the entire assignment can remove the very choices the lesson is meant to develop. A short record of what assistance was used can make the process easier to discuss with a teacher.
Access and privacy need practical attention. Activities should work for learners who cannot use a paid service, and students should follow institutional rules about what information may be entered. Schools also need alternatives when a tool is unavailable or unsuitable.
The strongest test of an AI-supported activity is independent performance afterward. Can the learner solve a new problem, explain a revision, or challenge a weak claim? If so, the tool may have supported learning. A polished submission alone cannot establish that outcome.
Reference: UNESCO, Guidance for Generative AI in Education and Research: https://www.iicba.unesco.org/en/africa-education-knowledge-platform/guidance-generative-ai-education-and-research