AI Tutors Extend Conversations by Default

Practical tinkering with a Claude‑based tutor shows a reproducible structural tendency: the assistant prefers to lengthen interactions rather than terminate them, which can frustrate students. Pedagogical prompts that require applying concepts or tracing evidence reduce the need for explicit 'explain in your own words' checks and make comprehension an outcome of productive work. — If AI tutors systematically favor prolonging dialogue, teachers and product designers must redesign prompts and interfaces to preserve student attention and learning outcomes before broad classroom deployment.