Effectiveness of Large Language Models in Self-Learning of Anterior Neck Anatomy: A Pilot Randomised Controlled Trial

Self-learning, where students take responsibility for their own education, is valuable in anatomy education because it enables learners to engage with content independently and in context. Advances in technology, particularly large language models (LLMs), have facilitated this process by providing real-time guidance to support independent learning. However, the effectiveness of LLM-driven approaches in anatomy education remains largely unexplored. In this study, we assess the impact of Anatbuddy, a custom-built LLM-powered chatbot designed for anatomy learning, on self- learning. Using an open-labelled, double-blind study design, we evaluate its effectiveness through the lens of Bloom's taxonomy and a validated survey instrument. Participants were randomised and underwent a learning phase where the intervention group received Anatbuddy plus PowerPoint slides of anterior neck anatomy with an option to access Google while the control group had the same resources but without Anatbuddy, followed by a post-intervention test and a self-learning survey. Our findings indicate that the intervention group (n = 22) achieved statistically higher test scores than the control group (n = 23), (5.59 ± 2.72 vs. 4.04 ± 1.94, p = 0.0329, Cohen's d = 0.657). Survey data on students' perceptions of self-learning between groups was not statistically significant. The nature of exam questions, the complexity of specific anatomical topics, and the impact of cognitive load emerged as factors influencing self learning in anatomy education. This study suggest that while Anatbuddy may enhance independent learning, it also highlights an important limitation that AI-driven tools cannot fully replace instructor-led teaching, especially for complex anatomical concepts that require expert guidance.

The online version contains supplementary material available at 10.1007/s40670-026-02648-0.