Voice Assistant (VA) is an advanced smart technology that responds to voice commands with the power of Artificial Intelligence (AI). This technology is embedded into various devices such as smartphones and smart speakers, which facilitate a wide range of daily tasks. In the context of healthcare, VAs offer significant potential for enhancing accessibility to health-related information and services. Therefore, this study explores the factors governing the intention to utilize VAs in the context of health by proposing an integrated framework that assimilates relevant constructs from the Technology Acceptance Model (TAM), including perceived ease of use (PEU), perceived usefulness (PU), and AI-specific factors which include perceived anthropomorphism (PA), perceived intelligence (PI), with perceived usefulness serving as a mediating variable. Structural equation modeling (SEM) is employed to analyze the relationships between the constructs. The results indicate that perceived usefulness, perceived intelligence, and perceived anthropomorphism significantly influence usage intention, while perceived ease of use does not exert a direct effect. Furthermore, PU acts as a mediating factor between PI, PEU, and usage intention, which indicates the importance of PU in VA adoption. These findings provided several implications for both theory and practice.