Juho Leinonen, Lisa Zhang, Arto Hellas (2026) — WCCCE 2026.
Key Findings
This study evaluated five generative AI tools for creating instructional slides from instructor-authored course notes: NotebookLM, Claude, M365 Copilot, Cursor, and Claude Code. Educators assessed slides for accuracy, completeness, and pedagogical soundness.
- Coding assistants (Cursor, Claude Code) produced the best slides — most accurate, complete, and pedagogically sound.
- Students rated AI-generated slides as similar in quality to instructor-created slides.
- Students could not reliably identify which slides were AI-generated.
- A negative correlation emerged: high quality ratings were associated with lower "AI-generated" guesses, suggesting students associate poor quality with AI origin even when AI slides are good.
Pedagogical Implications
The finding that coding assistants outperform dedicated education tools and general-purpose LLMs suggests that the scaffolding and structured output formats inherent to coding tools may translate to better instructional design outputs. The student bias finding — associating poor quality with AI — connects to ai-literacy research: if students penalize content they suspect is AI-generated, transparency about AI use may backfire unless accompanied by demonstrated quality.
Related Pages
- slidesqaqa-pedagogical-question-generation — question generation from slides vs. slide content generation
- cognitive-shift-ai-education — 471 students surveyed 2020–2026 show shift from AI preference to human intellige
- ai-tpack-teacher-multi-agent-workflow — Teacher archetypes shape AI content design and student perception
- faculty-development-genai — GenAI for instructional design workflows
- student-experience — Student perception and trust in AI-generated content
- higher-ed — University course context
- ai-literacy — Student ability to recognize and evaluate AI outputs
- scaffolding — Structured tool outputs as pedagogical scaffolds
- pedagogy-ai-mistakes — Student reactions to AI imperfections
- llm-tts-dialogue-lesson-generation - prior AI-generated instructional content study
Citation
APA: Leinonen, J., Zhang, L., & Hellas, A. (2026). AI-generated slides: Are they good? Can students tell? Proceedings of the Western Canada Conference on Computing Education (WCCCE 2026). arXiv:2605.13532.