Inaccurate perceptions of COVID-19 (coronavirus disease 2019) risk may decrease compliance with public health mitigation practices, in turn increasing disease burden. The extent to which public perceptions of COVID-19 risk are inaccurate is not well studied. This study investigates the relationship between preferred information sources and inaccurate COVID-19 risk perception. A cross-sectional online survey of adults in the United States using online snowball techniques was administered between April 9, 2020 and July 12, 2020. Raking techniques were used to generate a representative U.S. sample from 10,650 respondents. Respondents who did not provide an answer to key questions were excluded. The remaining sample included 1,785 health care workers (HCW) and 4,843 non-HCW. Subjective risk was measured as the product of perceived likelihood of COVID-19 infection and perceived harm from infection. Objective risk was measured as a function of the presence of known COVID-19 risk factors. Discrepancies between subjective and objective risk were compared between respondents with different preferred information sources. Chi Square contingency tables and pair-wise correlation were used to evaluate differences to 95% confidence. For HCW and non-HCW, the greatest overestimation of personal COVID-19 risk assessment ( < .05 for all differences) were found in those whose preferred source of information was social media (HCW: 62.1%; non-HCW: 64.5%), followed by internet news sources (HCW: 59.6%, non-HCW%: 59.1%), government websites (HCW: 54%, non-H = 51.8%), other sources (HCW: 50.7%, non-H = 51.4%), and television news (HCW: 46.1%, non-HCW: 47.6%). Preferred information sources correlate with inaccuracies in personal COVID-19 risk assessment. Public health information campaigns should consider targeting groups whose preferred information sources correlate to higher inaccuracies in COVID-19 risk perceptions. [].