Student Perceptions and Preferences Regarding AI-Generated Instructional Videos in Computing Education

As generative AI tools become more accessible, instructors are increasingly using them to produce educational content at scale. This study investigated how 170 computing students at two U.S. universities perceived AI-generated instructional videos. Participants watched three short, AI-generated videos on the Markdown markup language and subsequently completed a survey evaluating the quality of the videos and their broader attitudes toward AI in the classroom.

The results show a clear distinction between how students evaluate specific AI-generated content versus the general adoption of AI in education. When watching the Markdown tutorials, students rated the videos as high-quality, accurate, and easy to follow. Remarkably, nearly half of the participants could not distinguish whether the videos were AI-generated or human-produced. This suggests that, in terms of production value and clarity, current AI video tools are capable of meeting student expectations for supplemental learning resources.

Despite the positive reception of the specific videos, students expressed significant hesitation regarding the widespread use of AI in their courses. A majority of respondents stated they would not prefer AI-generated videos over those created by their instructors, nor would they trust the information in an AI video as much as that presented by a human. Students identified specific "safe" use cases for AI, such as summarizing foundational concepts, providing visual demonstrations, or replacing dense text-based documentation. However, they voiced concerns that over-reliance on AI could lead to inaccurate content, reduced opportunities for interaction with course staff, and a general decline in the perceived value of their education.

This research provides empirical evidence that while AI-generated videos are a viable tool for enhancing learning, they are currently viewed as a supplement rather than a replacement for human instruction. For educators, these findings suggest that AI videos are best deployed for routine, procedural, or visual tasks, while core conceptual instruction remains a domain where students prioritize the trust and personal connection associated with human faculty.