With advancements in Large Language Models, AI chatbots are now widely used, offering near-human-like interactions. This study examines the emotional and psychological impacts of AI in communication, particularly when paired with human-like avatars and how AI-based communication compares to human interaction when users are either aware or unaware of their interaction partner's identity. A total of 42 female participants were randomly assigned to an informed group (aware of whether they were speaking to an AI or a human) or an uninformed group (unaware of speaker's identity). Participants interacted with a virtual avatar that conveyed messages generated by either AI or a human while discussing their last holiday. In the uninformed group, participants could not correctly identify whether they had interacted with an AI or a human. In the informed group, interacting with a human led to a stronger increase in positive affect, a greater sense of social presence, and stronger attribution of emotions and intentions to the avatar compared to interacting with AI. In the uninformed group, no significant differences in emotional or psychological responses were observed between AI and human interactions. Regardless of awareness, messages generated during human interactions were more positive than those from AI. Additionally, messages sent by the AI or human experimenter were more positive than those sent by participants. These findings provide insights into user perceptions of interactions with virtual avatars that are either AI- or human-controlled, and can contribute to the design of AI systems in order to increase user satisfaction and engagement. • People struggle to distinguish interactions with an AI from those with a human. • Awareness of the identity of the communication partner affects the user experience. • Communication with a human generates more positive sentiment than with an AI. • The results underline the importance of user awareness when communicating with AI.
Kober et al. (2026) studied this question.