Trust me like a human: Unpacking the functional and non-functional dynamics of human-like trust in generative AI adoption

Users’ trust in technologies is influenced by technology’s perceived humanness. To advance acceptance of generative artificial intelligence (GenAI) in knowledge work, it is crucial to understand how trust influences workers’ attitudes and use. Acknowledging anthropomorphic features and human-like characteristics of GenAI, this study examines how human-like trust, consisting of benevolence, competence, and integrity, affects knowledge workers’ perceptions and adoption of GenAI. Drawing on the Artificial Intelligence Device Use Acceptance (AIDUA) model, we investigate how these trust dimensions change functional (perceived performance, perceived effort) and non-functional perceptions (emotions), which subsequently influence willingness to use and use intensity. Survey data from 1,503 knowledge workers in Germany, South Africa, and the United Kingdom were analyzed using four structural equation models. Results show that competence strongly affects functional and non-functional perceptions, integrity influences only functional perceptions, and benevolence primarily influences emotional responses. These findings position human-like trust as a multi-layered socio-technological appraisal mechanism and reveal trust calibration, where users rely on different trust cues when evaluating GenAI’s utility versus relational qualities.