Chit-chat is an innate human activity that facilitates familiarization, particularly during first interactions. This study represents a preliminary investigation of ChitT5, a conversational model developed for casual chit-chat between social robots and humans in their first encounters. Two transformer-based architectures, T5 and Flan-T5, were fine-tuned using datasets such as DailyDialog, Conversation Question Answer, and a custom small-talk dialogue dataset. Preliminary evaluations by five participants favored the Flan-T5 model for its interaction capabilities, leading to its selection and renaming as “ChitT5” for an in-person lab study. In this study with 22 English-speaking university students interacting with the EMAH system on the Ameca robot, findings showed that rapport and overall robot rating strongly correlated with the robot’s performance as a conversationalist. Despite being perceived as a “Stranger”, ChitT5 facilitated effective communication and safety perceptions. This research highlights the potential of custom chit-chat models like ChitT5 in enhancing casual interactions with social robots, providing insights for further conversational model development.