Background
Case-Based Learning (CBL) is a student-centered educational approach, based in constructivism, where students rely on prior knowledge, participate as active knowledge seekers and co-create new experiences in case scenarios. In our M1 fall semester, CBL includes a facilitator, online resources as needed and cases with guiding questions to teach basic sciences. However, as students use chat-based AI platforms in their CBL practice, the value of constructing new knowledge through collective discussion may be impacted.

Objective
This work hopes to evaluate how effectively AI chats perform CBL tasks compared to medical students. This data will help educators understand the role of AI in CBL activities in medical education without diminishing the educational value in fostering collaborative, student-centered learning.

Methods
Eleven learner group responses and three AI-generated responses were scored for 10 CBL questions by four blinded evaluators. Quality of responses was assessed by scoring the depth, scope, clarity of communication and quality of references and likelihood of being AI generated on a 5-point Likert scale. AI platforms used in this study included ChatGPT 4.0, Microsoft CoPilot, and AMBOSS Medical Knowledge.

Results
A mixed model analysis of summative scores revealed that AI models performed better than student groups (p < 0.05). Among the AI models tested, only AMBOSS GPT significantly outperformed student responses (p < 0.001) compared to copilot and ChatGPT. Comparison of individual AI models reveals that one model (AMBOSS GPT) performs higher (P< 0.001). Overall, evaluators could distinguish between AI and student responses.

Conclusions
Our findings indicate that AI can support CBL activities without diminishing work quality, but it is essential to emphasize to students that meaningful learning stems from the constructivist process of actively engaging with and building knowledge—not merely from producing correct answers.