Technological advances in artificial intelligence allow chatbots to mimic human interactions and take on an anthropomorphic appearance in the form of an avatar. Because of the human-like characteristics of chatbot avatars, users tend to identify with them. However, it has not yet been fully explained which characteristics of a chatbot avatar contribute to user identification with it and how this affects the user-provider relationship. Against this background, the avatar identification model is applied in the context of chatbot avatars to fill the identified research gap. The empirical results show that perceived similarity, embodied presence, and wishful identification affect user-chatbot identification, which in turn has an impact on parasocial preference for the chatbot. It was also shown that a parasocial preference leads to brand preference and loyalty, and that a user's emotional empathy traits have a moderating effect.