I study a model of Bayesian persuasion involving a benevolent sender with commitment power and a mass of receivers whose strategic behaviour is underpinned by a model of culture as shared cognition to investigate how communication varies as a function of culture. Receivers resolve strategic uncertainty in a pure coordination game by reasoning about others in introspective equilibrium (Kets and Sandroni, 2021). In the baseline model where all receivers belong to a single cultural group, I characterise the sender’s optimal information structure and find that social welfare with information is piecewise over three intervals and non-monotone in cultural strength when the receivers are suciently pessimistic. This is due to conflicting coordination and persuasion e↵ects: stronger cultures coordinate better, but force the sender to persuade them with bad news more frequently. I then investigate the implications of two cultural groups for Bayesian persuasion. Under this framework, I first investigate public persuasion and show that the sender’s optimal information structure depends on the cultural distance between the two groups. I then consider targeted persuasion, where one or both cultural groups receive information without knowing the other group’s. In this scenario, it is sometimes better for the sender to target a single group: the sender increases the strategic uncertainty faced by receivers who do not have information, which improves social welfare by making them more likely to choose the high action.