In the near future, automated vehicles (AVs) should take on the responsibility to communicate with other road users, such as pedestrians. However, in complex urban scenarios, AV’s external Human Machine Interfaces (eHMIs) may lead to confusion and danger while communicating with multiple road users. The present virtual reality (VR) study investigated the interaction between an AV and two pedestrians. 31 participants were instructed to cross an unsignalized two-lane street with the existence of an additional pedestrian. The approaching AV exhibited 3 behaviors and communicated with an eHMI concept. The two pedestrians started crossing on different roadsides with different lateral distances from the approaching AV. Each participant experienced 24 trials. Participants generally found the eHMI concept useful. However, misunderstandings and feelings of uncertainty were revealed, especially when participants were unsure whether the communication was intended for them or the other pedestrian. The other pedestrian’s crossing behavior was regarded as a signal for a safe crossing; thus, a second brake for the farther pedestrian was wrongly expected. Therefore, considering the reciprocal affection among pedestrians on their crossing behavior, when there is more than one pedestrian in the traffic environment, AV should be able to communicate with multiple pedestrians and dynamically adjust the driving behavior and communication content.