We study the effects on intimate partner violence (IPV) of new information received by women only, men only, or both, relevant to a high-stakes joint household decision. We model communication between spouses as Bayesian persuasion where disagreements elevate the risk of IPV. Our framework predicts that IPV will be lower when only one spouse is informed, compared to when both are, as the opportunity for persuasion by one spouse leads to more agreement. To test the model’s predictions we leverage an existing randomized controlled trial of an edutainment intervention addressing child marriage decisions for girls in rural Pakistan, targeted at men, women, or both. Our empirical findings confirm the prediction that the likelihood of IPV is highest when men and women are jointly targeted. Due to systematic gender differences in preferences, our persuasion model further predicts that marriage delays are largest when targeting men alone or jointly with women and smallest when targeting women alone, predictions that are also confirmed in the data.