Social service robots are increasingly being deployed in public spaces such as libraries, hospitals, and offices to assist and engage people. A critical yet underexplored factor in their success is nonverbal communication (the robot’s use of gaze, gestures, interpersonal distance and other cues) which profoundly shapes human–robot interaction. Equally important is user self-efficacy, or the confidence users have in their ability to interact with these robots effectively. This paper investigates how integrating nonverbal behaviors in a social service robot (TEMI) can enhance human interaction in real-world public spaces. We present findings from 15 qualitative interviews with end users, collaborators (on-site staff), and robot sellers, structured to explore users’ experiences, self-efficacy levels in controlled vs. real settings, and perceptions of the robot’s nonverbal cues. The results identify key categories of nonverbal behavior that influence user trust and comfort: Gaze & Eye Contact, Proxemics, Gestures & Movement, Posture/Height, Movement Speed, Expressiveness, and Cultural Sensitivity, as well as a notable drop in user self-efficacy when moving from training to real-life use. We discuss design and deployment recommendations to address these findings, emphasizing that successful social robot integration requires not only technical proficiency but also careful attention to nonverbal “social skills” and user support.