Why do current chatbots sound unnatural? It is old-news that the more human traits a technology has, the more social expectations the users create around it. When people talk to chatbots, they unconsciously establish coherence between agent’s utterances and plausible utterances that a person might form in that context. In this talk, we argue that chatbots that use coherent, appropriate language are more likely to meet users’ expectations. We will discuss the effect of chatbot’s linguistic choices on user’s experience and explain why considering sociolinguistic aspects for chatbots’ design would potentially empower them as social individuals.