Based on a series of experiments with AI in education, the VU Education Lab has gained considerable insight into what works well and what doesn’t. AI tutors can effectively enhance learning, but only when supported by sound educational design and integration. These tips explain how to achieve that.

Tip 1: integrate AI tutors into your course design - not as an add-on
It’s best not to treat an AI tutor as a bonus feature, but as an integral part of your course. The AI tutor should use the same terms and notations as in your lectures, so students don’t have to translate between systems. The exercises the AI offers should align directly with your exam content and follow the same order and pace as your course. When the AI tutor’s feedback is consistent with your explanations, the two reinforce each other rather than cause confusion.

Include in your syllabus how the AI tutor fits into your course and how students should use it. Demonstrate it during the first session so students immediately see how to work with it effectively. Research from Harvard (Kestin & Miller, 2024) shows that students using an AI tutor tailored to the course structure not only achieved higher scores but also felt more motivated than in traditional active learning sessions. 

Tip 2: tailor to your discipline
AI tutors perform best in subjects with clear problem-solving structures, such as mathematics, physics or programming. They are less effective for essays, creative work or philosophical discussions. In STEM courses, AI can guide students step by step through problems; in the humanities and social sciences, it can support brainstorming or critical questioning. In practice-oriented courses, AI can offer realistic scenarios for students to practise their skills. 

Tip 3: keep didactic control
An AI tutor is a tool, not a replacement for you as a teacher. You remain responsible for determining learning goals and how students may use AI - as a support tool and not as a robot that gives all the answers. You decide when human guidance is needed and how AI use is reflected in assessment. This didactic control is key to successful implementation. A systematic review from the University of Queensland (Létourneau, A et al., 2025) shows that AI tutors are most effective when they provide immediate feedback, personalise learning, and are embedded in a blended approach with active teacher involvement. 

Tip 4: decide whether you want insight into the tutor’s data
As a teacher, you may want to see what your students have mastered or where they struggle. However, if you cannot monitor the AI tutor’s interactions, you may lose part of that insight. At the same time, monitoring raises privacy and autonomy concerns - students might feel inhibited if you look over their shoulders. Be clear about whether you will access the data and what you will use it for. Also consider whether you might prefer an alternative, such as a quiz or paper-based exercise during class, which avoids some ethical dilemmas of AI use.

Tip 5: train students in effective AI use
Students often use AI tutors as ‘answer generators’ rather than learning tools. This can lead to what critics call ‘de-skilling.’ Explicitly address AI literacy: teach students to ask good questions, evaluate AI responses critically, and use the AI as a sparring partner rather than a cheat sheet. Help them recognize when human support is more appropriate than AI assistance. You can use, for example, the AI Literacy Companion - student edition by VU Amsterdam.

Interested in using AI in your teaching? Contact the VU Education Lab for tailored advice.