Objective Cognitive risk figures prominently in models predicting health behaviors, but affective risk is also important. We examined the interplay between cognitive risk (personal likelihood of COVID-19 infection or death) and affective risk (worry about COVID-19) in predicting COVID-19 precautionary behaviors. We also examined how outbreak severity bias (overestimation of the severity of COVID-19 in one's community) predicted these outcomes. Design In a representative sample of U.S. adults (N = 738; M-age = 46.8; 52% women; 78% white), participants who had not had COVID-19 took two online surveys two weeks apart in April 2020. Main outcome measures We assessed cognitive risk, affective risk, and outbreak severity bias at baseline and at follow-up two precaution variables: prevention behaviors (e.g. social distancing) and behavioral willingness (e.g. vaccinations). Results Overall, affective risk better predicted precautions than cognitive risk. Moreover, overestimating the severity of the outbreak predicted more affective risk (but not cognitive risk) and in turn more precautions. Additional analyses showed that when affective risk was lower (as opposed to higher) greater cognitive risk and outbreak severity bias both predicted more precautions. Conclusion These findings illustrate the importance of affective risk and outbreak severity bias in understanding COVID-19 precautionary behavior.