Expansion of the wildland–urban interface (WUI) and the increasing size and number of wildfires has policy-makers and wildfire managers seeking ways to reduce wildfire risk in communities located near fire-prone forests. It is widely acknowledged that homeowners can reduce their exposure to wildfire risk by using nonflammable building materials and reducing tree density near the home, among other actions. Although these actions can reduce the vulnerability of homes to wildfire, many homeowners do not take them. We examined the influence of risk factors on homeowners’ perceived wildfire risk components using a survey of WUI homeowners in central Oregon (USA) and biophysical data that described wildfire risk as predicted by wildfire simulation models, past wildfire, and vegetation characteristics. Our analysis included homeowners’ perceptions of the likelihood of wildfire and resulting damage, and examined how these factors contribute to homeowners’ likelihood to conduct mitigation actions. We developed an empirical model of homeowners’ risk perceptions and mitigation behavior, which served as input into an agent-based model to examine potential landscape and behavior changes over 50 years. We found homeowners’ wildfire risk perceptions to be positively correlated with hazardous conditions predicted by fuel models and weakly predictive of mitigation behavior. Homeowners’ perceived chance of wildfire was positively correlated with actual probability of wildfire, while their perceived chance of damage to the home was positively correlated with potential wildfire intensity. Wildfire risk perceptions also were found to be correlated with past wildfire experience. Our results suggest that homeowners may be savvy observers of landscape conditions, which act as “feedbacks” that enhance homeowners’ concerns about wildfire hazard and motivate them to take mitigation action. Alternatively, homeowners living in hazardous locations are somehow receiving the message that they need to take protective measures. Mitigation compliance output from the agent-based model suggests that completion of mitigation actions is likely to increase over 50 years under various scenarios.