In natural conversation, the minimal gaps and overlaps of the turns at talk indicate an accurate regulation of the timings of the turn-taking system. Here we studied how the turn-taking affects the gaze of a non-involved viewer of a two-person conversation. The subjects were presented with a video of a conversation while their eye gaze was tracked with an infrared camera. As a control, the video was presented without sound and the sound with still image of the speakers. Turns at talk directed the gaze behaviour of the viewers; the gaze followed, rather than predicted, the speakership change around the turn transition. Both visual and auditory cues presented alone also induced gaze shifts towards the speaking person, although significantly less and later than when the cues of both modalities were available. These results show that the organization of turn-taking has a strong influence on the gaze patterns of even non-involved viewers of the conversation, and that visual and auditory cues are in part redundant in guiding the viewers’ gaze.