Dominance is a key factor in interpersonal relationships, shaping others’ decision-making processes. This dynamic extends to human-agent interaction (HAI), where an agent’s expression of dominance can influence human perception. However, the specific effects of agent dominance cues on human decision-making within HAI―particularly the relative impact of verbal versus nonverbal cues―remain insufficiently explored. To address this gap, we conducted an experiment employing a 2 (verbal cues: dominant vs. submissive) × 2 (nonverbal cues: dominant vs. submissive) full factorial design. In a virtual reality (VR) city environment, participants interacted with agents exhibiting different combinations of these cues, each providing navigational suggestions. The results revealed that while verbal cues significantly influenced perceived agent dominance, only nonverbal cues affected participants’ decision-making (i.e., which agent’s suggestion they followed) and their liking of the agent. Furthermore, both verbal and nonverbal cues shaped perceptions of the agent’s intelligence, competence, trustworthiness, and warmth. Moreover, participants’ own trait dominance moderated their perception of the agent’s dominance. These findings advance our understanding of how dominance cues operate in HAI and offer practical insights for the strategic design of embodied conversational agents (ECAs) to enhance user perception and interaction outcomes.