The current study explores the association between place features (e.g., schools, public transportation nodes, bars and restaurants) and perceived unsafe locations using data from two open-ended items in a cross sectional random sample community survey in Malmö, Sweden. Perceived unsafe locations in respondents’ own neighborhood and in other parts of the city are geocoded and merged with a 200 by 200-m grid-cell network. The data are analyzed using logistic regressions inspired by the Risk Terrain Modeling approach, where geographical representations of place features are used as predictors and geographical representations of unsafe locations used as outcomes. The results show that several place features are associated with unsafe locations, but the importance of different features seem to vary between outcomes. While grid-cells being near an elementary school, convenience store or park demonstrates significant associations with unsafety when respondent’s report locations in their own neighborhood, these patterns are weaker when considering unsafe locations in other parts of the city, where the most important variable is whether a grid-cell is located within a neighborhood defined as vulnerable by the national police agency. This indicates a qualitative difference in what types of locations are being reported as unsafe when respondents are asked about locations in their own neighborhood versus other parts of the city. These results contribute to the current body of research on place-based fear of crime, and the findings may offer guidance in better understanding why some places are perceived as unsafe and how public perceptions of safety may be improved.