This study investigates the factors influencing continuance use intention toward generative AI chatbots by focusing on the human-like characteristics of conversational quality. Specifically, it examines how conversational quality influences trust formation and post-adoption evaluation within the framework of the Expectation Confirmation Model (ECM). Conversational quality is conceptualized as a multidimensional construct consisting of understanding humanness, perceived contingency, and response humanness.
Data were collected from 231 ChatGPT users and analyzed using PLS-SEM. The results show that all three conversational quality dimensions significantly influence user trust. However, only perceived contingency has a direct effect on perceived usefulness, while understanding humanness and response humanness do not significantly affect usefulness perceptions. Trust positively influences perceived usefulness, indicating that trust functions as a cognitive mechanism that shapes users’ evaluation of AI-generated outputs. Consistent with the ECM, confirmation significantly affects both perceived usefulness and satisfaction, while satisfaction has the strongest effect on continuance use intention.
These findings suggest that conversational interaction quality plays a critical role in trust formation in generative AI environments, and that contextual responsiveness is more important than linguistic humanness in shaping perceived usefulness. By integrating conversational quality and trust into the ECM, this study extends post-adoption research in generative AI contexts and provides practical implications for the design of AI-driven conversational systems in electronic commerce.