Trust formation across AI embodiments: how robotic, virtual, and embedded systems shape user behavior in healthcare

Purpose Trust in artificial intelligence (AI) systems shapes user behavior and technology acceptance, yet limited research compares how different AI embodiments influence trust formation. This study aims to examine how robotic, virtual, and embedded AI affect cognitive trust in healthcare contexts. Design/methodology/approach An experimental study was conducted to assess the effects of tangibility, transparency, reliability, task characteristics, and immediacy behaviors on cognitive trust across three AI embodiments. Participants evaluated healthcare scenarios involving robotic, virtual or embedded AI systems, and trust-related perceptions were analyzed across conditions. Findings Significant perceptual differences emerged across AI embodiments. Virtual AI was perceived as less tangible than robotic and embedded AI, while embedded AI was rated highest in technical intelligence but lowest in social intelligence. Despite these differences, overall cognitive trust did not significantly vary across embodiments. However, the mechanisms of trust formation were embodiment-specific. Reliability consistently predicted trust across all conditions. Transparency and immediacy behaviors enhanced trust in robotic AI, tangibility and social intelligence drove trust in virtual AI, and immediacy behaviors were especially critical for embedded AI. Originality/value This study advances AI trust research by positioning embodiment as a boundary condition that shapes mechanisms of trust formation rather than overall trust levels. It extends prior work by demonstrating that trust antecedents operate differently across AI forms and by embedding this analysis within healthcare, where risk and task sensitivity fundamentally influence trust dynamics.