Abstract
People increasingly use conversational AI for support and companionship. Yet, current discourse on AI companions reveals a stark divide: some scholars argue that feeling connected to AI is impossible due to AI's inability to experience emotions, while others contend that AI's ability to provide the illusion of emotions is enough. Across two experiments (one pre-registered; Total N = 1274), we investigated whether accounting for individual differences in anthropomorphism could help bridge these two perspectives. Participants completed a measure of anthropomorphism and were then randomly assigned to discuss their past month through a conversation with a chatbot (chatbot condition) or by journaling (control condition) and then completed a measure of social connection. Our results suggest that for some individuals, AI's artificial nature may pose an insurmountable barrier to meaningful connection; for others, however, this artificiality may be a minor obstacle, easily overcome by a tendency to anthropomorphize.