This research establishes attitude ambivalence as an explanation for previous inconsistent results concerning two-sided messages’ effectiveness. The results of a regression study show that the more ambivalence a two-sided message evokes, the less effective it is in terms of attitudes and intentions (study 1a). In a second study (study 1b), we validate this effect in a between-subjects factorial design, and confirm that when a two-sided message evokes high (vs. low) ambivalence, message effectiveness deteriorates. Study 2 shows that these dire effects of high ambivalence in two-sided messages can be resolved by explicitly refuting the negative attributes in the message. Refutational messages overcome the negative effects of high ambivalence on attitudes and intentions. This research advances and deepens literature on message sidedness and refutational messages, while also offering practical recommendations on how to remedy cases of high ambivalence in two-sided messages.