The increasing adoption of digital beauty technology through Virtual Try-On (VTO) platforms powered by artificial intelligence (AI), augmented reality (AR), and virtual reality (VR) necessitates a precise understanding of user preferences. This study aims to model user responses to VTO features using Item Response Theory (IRT), specifically the Graded Response Model (GRM) for ordinal-scale items and the Nominal Response Model (NRM) for non-ordinal categorical items. A total of 15 items were analyzed using GRM to assess users’ perceptions of AI accuracy, feature convenience, and visual satisfaction. Additionally, five non-ordinal items were modeled using NRM to capture explicit preferences for specific VTO functionalities, such as digital interaction and visual simulation. The results show that the average user preference score was 3.02 out of 5 (SD = 1.03), while the total preference score averaged 60 out of 100 (SD = 10.7), with response distributions approximately normal (skewness = 0.12; kurtosis = -0.24). The item information curves from the GRM indicated that several items contributed high information (above 2.0), effectively differentiating users within the −1 to +1 ability range. In the NRM analysis, the “real-time color visualization feature” exhibited the highest utility parameter (c = 1.98), indicating a strong user preference for this feature. The integration of GRM and NRM enables a more nuanced and flexible mapping of user preferences. This study presents a novel psychometric approach by combining GRM and NRM within a single VTO context to simultaneously capture both ordinal and nominal dimensions of user experience, which has not been extensively explored in prior research. These findings provide meaningful insights for the development of adaptive, personalized, and intelligent AI-based VTO systems to enhance the quality of user experience in digital beauty applications.