The COVID-19 pandemic threatens millions of lives, and an effective response will require individuals to take costly and difficult measures to prevent infection. How should public health messaging frame these measures, which can reasonably be conceptualized as either self-interested actions or cooperative efforts? We measured the influence of three messaging treatments on coronavirus prevention intentions among Americans from Amazon Mechanical Turk. All treatments presented identical COVID-19 information, but emphasized either personal, public, or both personal and public benefits of prevention behaviors. In studies (n = 2176) conducted early in the pandemic (March 14-16, when there were under 2,000 confirmed U.S. cases), we found support for prosocial framing: the Public treatment was more effective than the Personal treatment, and no less effective than the Personal+Public treatment. In studies (n = 3985) conducted later (April 17-30, when there were over 500,000 confirmed U.S. cases), all three treatments were similarly effective. Additionally, across both sets of studies, the perceived public threat of coronavirus predicted prevention intentions more strongly than the perceived personal threat. Together, our results highlight the potential value of prosocial framing.