QGen: Dynamic AI Question Generation During Live Classroom Lectures

University students occasionally struggle to ask questions in class. While AI can support learning, little is known about how it can support in-class question asking. We developed an AI system that generates questions in real time from lecture content and a mobile interface that allows students to view, vote on, and submit their own questions alongside AI-generated ones. We deployed this system in four computer-science lectures, and found that AI-generated questions helped students consider lecture content from different perspectives. For some, this prompted deeper reflection on what to ask. However, others reported reduced motivation to form their own questions, suggesting risks of overreliance. Although students found the AI helpful, they more frequently voted for questions asked by their peers, revealing a gap between AI-generated questions and students' in-the-moment learning needs. We provide design recommendations for systems that promote interactive collaboration with AI and scaffold question asking without undermining student agency.