<p>While both low social connectedness and an AI agent’s emotional expression can increase anthropomorphism, their interaction remains underexplored in conversational AI (CAI) contexts. Accordingly, this study examined whether social exclusion predicts anthropomorphism toward CAI and whether this association is moderated by the AI agent’s emotional expression. Participants interacted with a fictitious AI agent conversation. Participants (<i>N</i> = 120) were randomly assigned to either an emotional AI condition where emotional words are expressed by AI agent (<i>n</i> = 53) or to a control condition (<i>n</i> = 67), where no emotion words are expressed. Social exclusion was also measured using a social exclusion scale. Results showed that the main effect of emotional AI condition on anthropomorphism was not significant. However, social exclusion was positively associated with anthropomorphism. Furthermore, the moderation effect of social exclusion and emotional AI condition on anthropomorphism was significant. Specifically, social exclusion was positively associated with anthropomorphism only in the control condition. These findings tentatively suggest that an emotionally expressive CAI may function as a form of “social snacking,” and may be associated with a weaker relationship between social exclusion and anthropomorphism.</p>