Risk perception and consequently engagement in behaviors to avoid illness often do not match actual risk of infection, morbidity, and mortality. Unrealistic optimism occurs when individuals falsely believe that their personal outcomes will be more favorable than others' in the same risk category. Natural selection could favor overconfidence if its benefits, such as psychological resilience, outweigh its costs. However, just because optimism biases may have offered fitness advantages in our evolutionary past does not mean that they are always optimal. The current project examined relationships among personal risk for severe COVID-19, risk perceptions, and preventative behaviors. We predicted that those with higher risk of severe COVID-19 would exhibit unrealistic optimism and behave in ways inconsistent with their elevated risk of morbidity and mortality. Clinical risk scores for severe COVID-19 were calculated and compared with COVID-19 threat appraisal, compliance with shelter-in-place orders (March 13-May 22, 2020) and travel restrictions, compliance with public health recommendations, and potential covariates like self-rated knowledge about COVID-19 in a robust dataset including 492 participants from McLennan County, TX, USA. While those with high clinical risk acknowledged their greater likelihood of experiencing severe illness if infected, they actually reported lower perceived likelihood of becoming infected in the first place. While it is possible that those with higher clinical risk scores truly are less likely to become infected, the pattern and significance of these results held after controlling for possible occupational exposure, household size, and other factors related to infection probability. Higher clinical risk also predicted more recent travel within Texas and lower distress during the pandemic (i.e., feeling less stressed, depressed, and helpless). Additional behavioral data suggested that those with higher clinical risk scores did not generally behave differently than those with lower scores during the shelter-in-place order. While unrealistic optimism may provide some short-term psychological benefits, it could be dangerous due to improper assessment of hazardous situations; inferring that optimism bias has evolutionary origins does not mean that unrealistic optimism is "optimal" in every situation. This may be especially true when individuals face novel sources (or scales) of risk, such as a global pandemic.