What separates an AI tutor from a chatbot?

AI tutoring tools are becoming more common, but I think there’s still a big gap between answering questions and actually teaching. A chatbot can explain a concept, but a tutor identifies misconceptions, asks guiding questions, adapts explanations, and helps learners build confidence over time. I’m curious how other builders and educators think about this. If you were designing an AI tutor today, what capabilities would you consider essential before calling it a “tutor” instead of a chatbot?