Does gesturing primarily serve speaker internal purposes, or does it mostly facilitate communication, for example by conveying semantic content, or easing social interaction? To address this question, we asked native speakers of Dutch to retell an animated cartoon to a presumed audiovisual summarizer, a presumed addressee in another room (through web cam), or an addressee in the same room, who could either see them and be seen by them or not. We found that participants produced the least number of gestures when talking to the presumed summarizer. In addition, they produced a smaller proportion of large gestures and almost no pointing gestures. Two perception experiments revealed that observers are sensitive to this difference in gesturing. We conclude that gesture production is not a fully automated speech facilitation process, and that it can convey information about the communicative setting a speaker is in.