Civic engagement is considered an important element of a healthy polity. Yet, many attempts to induce it fail, and experimental evidence on ‘what works’ to induce it is limited. Further, most experimental studies in this area of research focus on self-reported or low-cost proxy behavior outcomes. This paper describes a laboratory experiment with 809 participants to measure the impact of short behaviorally-informed messages on civic engagement in Uganda. We randomly assign participants to three treatment audio messages, applying different self-efficacy enhancing techniques, and one control audio, and measure their effect on two primary components of civic engagement: a lab measure of attention to political and pre-political information, and a real-world activity measure: attendance at a community meeting. We find that short audio messages can prompt people to pay more attention to pre-political information, but do not encourage overall civic engagement behavior, and some treatments produce a significant negative effect on our measures of engagement. We also find that a rights-focused message, closely modelling current civil society practice, does not generate increased civic engagement. We conclude that one-off behaviorally informed audio messages are insufficient to generate civic engagement in this context.