Study Finds Students Can't Tell AI-Generated Slides from Human-Made Ones
A new study published on arXiv (2605.13532) investigates how generative AI can help create effective educational slides from teachers' notes. The research looked at NotebookLM along with two general-purpose language models, Claude and M365 Copilot, and two coding assistants, Cursor and Claude Code. According to feedback from educators, the coding assistants produced the most accurate and helpful slides for teaching. Interestingly, in real classroom settings, students found it hard to tell the difference between slides generated by AI and those made by their teachers. This research highlights how generative AI could be a valuable tool for teachers in developing their presentations.
Key facts
- Study published on arXiv with ID 2605.13532
- Tested NotebookLM, Claude, M365 Copilot, Cursor, and Claude Code
- Coding assistants (Cursor, Claude Code) produced best slides
- Students rated AI slides similar to human-made slides
- Students could not reliably identify AI-generated slides
Entities
Institutions
- arXiv