With recent developments in state-of-the-art conversational chatbots which levarage powerful language models such as GTP-3, evidently there has not been many studies conducted for its presentation to the user as an entity. In this study, we utilized an open source state-of-the-art chatbot, Blenderbot by Facebook, and accompanied with it 10 avatar configurations varying in realism, humanity and gender. Participants were tasked to observe interactions with the 10 different configurations and evaluate their impressions. Results show that avatar quality and humanity may be detrimental to the experience as they raise user expectations - furthermore, participants reacted better to avatars of the opposite gender.