What happens when students can openly discuss AI boundaries, then reflect on their own use through a simple survey?