Beyond Traditional Simulation: An Exploratory Study on the Effectiveness and Acceptability of ChatGPT‑4o Advanced Voice Mode for Communication Skills Practice Among Medical Students

Background Effective communication skills are essential for quality medical practice and patient care, yet providing sufficient practice opportunities remains challenging in medical education. Traditional approaches using standardized patients (SPs) face limitations, including high costs and logistical constraints. This study investigates the potential of conversational artificial intelligence (AI) to supplement communication skills practice among medical students. Methodology We conducted a mixed-methods study employing a convergent parallel design with 27 medical students from three UK universities between November 2024 and March 2025. Participants engaged in communication skills roleplay sessions using freely available ChatGPT as an SP. Pre- and post-session assessments measured perceived usefulness and self-reported confidence in three communication domains: dealing with challenging patients, breaking bad news, and counselling anxious patients. The instrument was adapted from the Immersive Technology Evaluation Measure (ITEM). Quantitative data were analyzed using non-parametric tests and ordinal logistic regression, while qualitative responses underwent reflexive thematic analysis. Results Self-reported confidence increased significantly across all communication domains following the ChatGPT simulation. Confidence in dealing with challenging patients increased from a median of 3 to 4 (<i>P </i>< 0.001), breaking bad news from 2 to 4 (<i>P </i>< 0.001), and counselling anxious patients from 3 to 4 (<i>P </i>< 0.001). University affiliation influenced willingness to repeat the experience (odds ratio (OR) = 0.19; 95% confidence interval (CI): 0.05-0.83) and perceived effectiveness in creating realistic scenarios (OR = 0.18; 95% CI: 0.04-0.79). Qualitative analysis revealed four themes of varying experience and expectations, valuable features (particularly detailed feedback and safe practice environment), limitations (technical challenges and restricted emotional range), and AI's complementary role in education. Conclusions This exploratory study provides preliminary evidence that conversational AI may enhance medical students' confidence in challenging communication skills while offering structured feedback in an accessible format. While we observed significant pre-post improvements in self-reported confidence measures, these findings should be interpreted cautiously given the small sample size, absence of a control group, and reliance on subjective self-assessments. Limitations in emotional authenticity and transcription accuracy were identified, alongside notable institutional variations in perceived realism. Despite these constraints, participants valued the psychological safety and convenience afforded by AI simulation. Our findings suggest that AI simulation may have potential as a complementary tool to expand communication skills practice opportunities, though larger controlled studies are needed to establish efficacy relative to traditional approaches. Future research should incorporate objective outcome measures and investigate the transferability of skills to real clinical encounters.