Much concern has been raised about the unique power of political microtargeting to sway voters’ opinions, influence elections, and undermine democracy. Yet little empirical research has directly estimated the persuasive advantage of microtargeting over more conventional messaging strategies. Here, we do so using two large-scale studies focused on U.S. policy issue advertising (total n = 32,695 U.S. adults; 74 distinct persuasive messages). To simulate a microtargeting approach, we combine machine learning with message pre-testing to determine which messages to show to which individuals to maximize persuasive impact. Using randomized experiments, we then compare the performance of microtargeting against two alternative messaging strategies: (i) the naïve strategy, in which people are exposed to a random message from a set of relevant messages, not guided by pre-testing, and (ii) the single-bestmessage strategy, in which people are exposed to the message that was the most persuasive overall in pre-testing. We estimate that microtargeting outperforms these alternative strategies by an average of 70% or more in a typical political advertising context where all messages are attempting to influence the same policy attitude (Study 1). In a less typical context, where microtargeting is used to identify both which policy attitude to target and which message to show (Study 2), microtargeting outperforms the naïve strategy but is no better than the single-best-message strategy. Together, these results suggest that political microtargeting can confer a sizable persuasive advantage over more traditional messaging strategies and furnish new insights about the potential societal impacts of this increasingly common campaign practice.