Comparative evaluation of the effectiveness of NotebookLM-generated and delivered lectures for dental interns: an interventional study

Abstract Background and objective The introduction of AI into the area of dental education provides a new opportunity through which to improve student learning outcomes. This study compared the effectiveness of traditional ‘live’ lectures with a video lecture produced and delivered by NotebookLM (Gemini Notebook) to teach endodontics to dental interns. Methods Two groups of dental interns were randomly assigned to receive either live lectures or NotebookLM videos on 3 progressively difficult endodontic subjects (diagnosis and treatment planning – basic; vital pulp therapy – intermediate; and root resorption – advanced). Both instructional delivery methods utilized identical source materials (current journal articles, books, and clinical practice guidelines). All students completed 15 multiple-choice questions designed to measure higher-level thinking skills, as well as self-assessment questionnaires measuring perceived learning outcomes and satisfaction after attending their respective training sessions. Results There were no significant differences in the diagnosis ( p = 0.327) or vital pulp therapy ( p = 0.411) post-test scenarios. However, the AI lecture group performed significantly better in the root resorption scenario (87.8% vs. 83.0%, p = 0.029). Students in the AI group reported significantly higher engagement on 2 items: the logical flow of the presentation ( p = 0.041) and the relevance of the topic to their training ( p = 0.024). A significant difference was also found for self-reported distraction ( p = 0.003), higher in the AI group (74.5% vs. 38.3%). The AI group was also favored in delivery style, content for note preparation, overall satisfaction, confidence on the topic after lecture, clinical application capability, clarification of complex concepts, and meeting of learning objectives (all p ≤ 0.012). For all other items in the questionnaire, there was no statistically significant difference between the groups. Conclusions NotebookLM-generated video lectures matched human-delivered lectures on immediate post-test performance for diagnosis and vital pulp therapy and yielded higher scores for root resorption, a difference that may be topic-specific. Interns also reported a favorable experience on the self-assessment questionnaire. Future research could examine the efficacy of student-AI educator interaction during interactive sessions, as well as long-term knowledge retention.