Evaluating expert opinions on using a virtual patient app based on artificial intelligence for teaching brief smoking cessation advice to medical students: a cross-sectional study

BACKGROUND: Smoking is one of the main preventable causes of premature disease and death. Anti-smoking counseling continues to be a challenge to cover the educational gap in this topic. Artificial intelligence and virtual-patient technologies are a promising next step in this evolving context, allowing for realistic, adaptive, and standardized simulations of clinical counseling scenarios. The objective of this study was to evaluate the perceptions and acceptance of a Generative Artificial Intelligence-based Virtual Patient Application for training medical students in brief smoking-cessation counseling among respiratory medicine professionals in Romania. METHODS: A cross-sectional survey, primarily descriptive, with additional analytical components to explore associations between key variables was conducted between May and October 2025 among respiratory medicine professionals in Romania. A structured, validated electronic questionnaire assessed perceived usefulness, feasibility, preferred counseling models, interface design, and expected educational and public health benefits of the GenAI-ViP application. Descriptive statistics summarized participant responses, while inferential analyses included Mann-Whitney U, Kruskal-Wallis, Chi-square tests, Spearman correlations, and binary logistic regression to identify predictors of perceived usefulness. Statistical significance was defined as p < 0.05. RESULTS: A total of 147 clinicians were included, most of whom were pulmonologists (89.1%), and digital engagement was high (95.3% reporting regular use of medical applications). Most respondents rated the application as useful (81.6% Likert scores 4-5) and reported high receptiveness (91.1% Likert scores 4-5). The preferred interface included multimodal interaction with voice and video (65.3%). The 5 A counseling model was identified as the most relevant framework (80.3%), followed by motivational interviewing (37.4%). Significant predictors of perceived usefulness included higher perceived student receptiveness (OR 2.535, 95% CI 1.416-4.539, p = 0.002), while clinicians without prior smoking cessation program involvement reported greater perceived usefulness (OR 0.266, 95% CI 0.087-0.815, p = 0.020). CONCLUSIONS: The Generative Artificial Intelligence-based Virtual Patient Application for smoking cessation training was highly accepted and perceived to be useful by respiratory medicine professionals and could address existing gaps in tobacco counseling education. Integration of such tools into undergraduate and postgraduate medical curricula could enhance smoking cessation competencies and support broader tobacco control strategies.