Objective: This study aims to develop and evaluate an AI chatbot virtual patient system dedicated to nursing, addressing the shortcomings of traditional nursing clinical thinking training, such as high costs and limited case diversity. Methods: Based on the Transformer architecture, the chatbot was fine-tuned using a nursing-specific dataset (NANDA-I nursing diagnoses, clinical guidelines) to simulate three typical case scenarios (chronic diseases, acute and critical illnesses, elderly care). Sixty nursing students were randomly assigned to the experimental group (n = 30) or the control group (n = 30). The Lasater Clinical Judgment Rating Scale (LCJR) was used to assess clinical judgment ability, and the SEGUE framework was utilized to evaluate communication skills. Mixed methods analysis included statistical testing using SPSS and thematic analysis of interviews. Results: The total LCJR score for clinical judgment in the experimental group (28.72 ± 3.21) was significantly higher than that in the control group (23.52 ± 2.82) (t = 6.32, p < 0.01), and the SEGUE communication skills score in the experimental group (19.86 ± 3.05) was significantly better than that in the control group (11.16 ± 2.14) (t = 14.787, p < 0.05). Qualitative interviews revealed that 82% of the experimental group students believed that the system effectively improved their consultation skills. Student satisfaction survey results indicated that 93.17% of students felt that the “case design was close to clinical practice,” 87.26% expressed a “willingness to continue using it,” and 76.35% believed that it “can assist traditional teaching.” Conclusion: The AI virtual patient system can effectively enhance the clinical thinking and communication skills of nursing students, promoting their clinical work abilities. Additionally, artificial intelligence technology provides scalable new teaching ideas and technologies for nursing education reform with limited resources.