Purpose: This study aimed to develop a Generative Pre-trained Transformer-based model specialized in nuclear medicine education and patient guidance and to evaluate its effectiveness as a supplementary learning tool Materials and Methods: The specialized system was designed to provide tailored interactions for students and the general public through role-specific prompts and structured explanations based on nuclear medicine textbooks. Thirty-two participants experienced the system and assessed its performance across four domains: user experience, suitability, information quality, and reliability, using a 5-point Likert scale. Results: All evaluation domains achieved mean scores of approximately 4.0, indicating overall positive responses. Participants highly rated the system for its structured explanations (mean 4.15±0.71), interactive follow-up questions (mean 4.34±0.64), and efficient information retrieval (mean 4.03±0.72). However, visual formatting (mean 3.90±0.80) and the risk of receiving incorrect information (positive response rate 62.5%) were identified as areas requiring further improvement. Conclusion: Domain-specific GPT models can effectively supplement traditional textbooks by enhancing understanding and supporting self-directed learning in nuclear medicine.