It remains unknown whether the virtual agents’ implicit behaviors (e.g., the agent’s eye-gaze shifting) would also cause such an observable effect on the users’ experiences even when the users are not consciously aware of the implicit behaviors of the agents. More importantly, it is unclear whether the effect of the implicit behaviors of agents can be amplified as the agents’ group size increases. To answer these questions, we asked two participant groups ($\mathrm{N}=12$ each) to deliver a public speech in front of the virtual agents showing either an explicit (head movement) or implicit (eye movement) behavior as we manipulated the agents’ group sizes (small, medium, and large). We measured the users’ perceptions, emotions, and behavior (head and eye movements) in the study. A posterior psychophysical experiment confirmed the implicit nature of the eye-movement behavior of the virtual agents (i.e., participants’ detection rate of the eye movements was less than the chance level). The results showed that the explicit behavior of the virtual agents caused the expected participant behavior changes along with negative emotions. In contrast, the implicit behavior of the agents caused largely positive emotions, yet, without significant behavior changes of the participants. Most importantly, as the group size of the virtual agents increased, the degree of their effect on users increased in both the explicit head-movement condition and the implicit eye-movement condition.