In this research, we develop a research model explaining the adoption of conversational agents for disease diagnosis. Healthcare is challenged by a parallel increasing demand for healthcare services and a decreasing supply of healthcare professionals. Mobile Health is proposed to overcome geographical, temporal, and organizational barriers of healthcare services. Conversational agents (CA), i.e. software programs that interact with users through natural language, are developed that are even able to diagnose a disease based on an individuals' input using a chat interface. However, these systems face an adoption challenge. To understand that, we use UTAUT2 as theoretical lens and 35 semi-structured interviews with potential users of a CA for disease diagnosis. Based on that we propose a research model that contains (1) well-known UTAUT2 factors (performance and effort expectancy, facilitating conditions), (2) re-defined other factors to better fit the context (social influence, price value, habit), and (3) newly identified ones (privacy risk expectancy, trust in provider and system, compatibility, experience in e-diagnosis, access to health system). We also reveal that hedonic motivation is not relevant for CA adoption. The newly proposed model addresses research gaps in CA research in general, but also in mHealth and especially the use of CA in healthcare research in particular. We also discuss rather general implications for technology acceptance research and provide some suggestions for providers of CA in healthcare to increase the diffusion rates of these systems.