Across all species, a dyadic interaction entails that an emitter sends signals to a receiver through one or more channels. The receiver decodes the signals and reacts accordingly, potentially turning into an emitter himself. Like dogs and horses, cats can integrate visual and auditory signals sent by humans and modulate their behaviour according to the valence of the emotion perceived. However, the specific patterns and channels governing cat-to-human communication are poorly understood. This study addresses whether, in an extraspecific interaction, cats adapt their communication channel to those used by their human interlocutor. We examined three types of interactions: vocal, visual and bimodal (visual plus vocal), through coding video clips of 12 cats living in cat cafés. There was a significant effect of the modality of communication on the latency for cats to approach the human experimenter. Cats interacted significantly faster in the visual and bimodal compared to the “no communication” pattern as well as to the vocal condition. There was a significant effect of communication modality on the tail wagging behaviour. Cats significantly displayed more tail wagging when the experimenter engaged no communication compared to the visual and bimodal conditions. Cats also displayed more tail wagging in the vocal compared to the bimodal condition. Taken together, these results suggest that cats display a marked preference for both visual and bimodal cues addressed by non-familiar humans, over vocal cues only. Our data bring further evidence for the emergence of human-compatible socio-cognitive skills in cats, that favour their adaptation to a human driven niche.