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

Synthesis: Studies student perceptions of AI-generated instructional videos in computing education. Finds students value personalization and rapid production but express concerns about accuracy and the loss of instructor presence. Identifies clear preferences for hybrid approaches where AI generates draft content that instructors review and refine.

Key Findings

• In a descriptive post-test survey, 170 computing students at two U.S. institutions watched three 3-minute AI-generated videos on the Markdown markup language, created with the Knowlify tool, and then reported their perceptions of the videos and of AI-generated video use in education more broadly.
• Students rated the specific AI-generated videos as high-quality, accurate, and usable, and nearly half were unable to determine that the videos were AI-generated.
• Despite positive ratings of the videos themselves, students expressed limited comfort with the widespread adoption of AI-generated videos in the classroom.
• Students preferred AI videos for simple, supplemental, and visual use cases, while expressing concerns about lower-quality or inaccurate content, reduced instructor interaction, and diminished educational value.
• The study positions AI video generation as a way for instructors to create personalized educational videos efficiently and cost-effectively, extending computing-education research beyond text-based AI tools.

Study Design & Method

The study addresses a gap in computing education research, which has focused largely on text-based AI tools for developing learning resources even as advances in AI video models make high-quality personalized videos feasible. Using a descriptive post-test survey design, the researchers recruited 170 computing students across two U.S. institutions. Participants watched three three-minute AI-generated videos on Markdown and completed a survey covering both perceptions of those videos and broader views on AI-generated videos in education. Outcomes were analyzed descriptively, with attention to whether students could detect AI authorship and how stated preferences varied by use case.

What this means for practice

• Instructors. Restrict AI-generated video to simple, supplementary, visually oriented content and keep instructor-led interaction in place: students rated the three Markdown videos as high-quality, accurate, and usable, yet 26 students (16% of the coded responses) argued against replacing instructor-led lectures outright.
• Instructors. Add a human validation step before publishing any AI-produced video; nearly half of the 170 students could not determine that the videos were AI-generated, so undetected inaccuracies would not be challenged by viewers.
• Instructors. Disclose AI authorship and pair it with demonstrated quality — the study's own contrast between positive video ratings and limited comfort with classroom adoption points to transparency as the way to preserve student trust in Generative AI resources.
• Designers. Adopt the hybrid production workflow students preferred: AI generates draft media that the instructor reviews and refines, treating the tool as an augmentation of the instructor's role in CS Education rather than a replacement.
• Researchers. Test AI videos with less experienced learners before generalizing; participants were enrolled in higher-level CS courses and were already fluent in technical notation.

Limitations

• Descriptive post-test survey with no comparison condition: 170 consenting students (143 at institution 1, 27 at institution 2) watched three 3-minute videos and reported perceptions afterward, so the design cannot show the videos were more effective than other instructional materials.
• The knowledge measure was brief and incidental: the 86% average on the five-question post-survey Markdown quiz reflects immediate acquisition only, not retention or transfer.
• Easy content, experienced learners: the stimulus covered Markdown in three-minute clips and most participants reported little or no prior Markdown familiarity but came from higher-level CS courses, limiting generalization to novices and to more complex content.
• Qualitative coding is open to multiple interpretations: the authors report their coding process and representative quotes rather than an inter-rater reliability statistic.

Citation

Esse Ciego, Shubbhi Taneja, Wilson Wong, Amanpreet Kapoor (2026). Student Perceptions and Preferences Regarding AI-Generated Instructional Videos in Computing Education.