Abstract Background In the light of urbanization and aging, a crucially relevant policy question is how to shape neighborhoods to foster healthy aging. An important debate is whether older adults should group in neighborhoods, or whether a more mixed neighborhood age composition is more beneficial to health and well-being. We therefore assessed the association between neighborhood age structure and mental health and the mediating role of individual perceptions of neighborhood social factors. Methods We conducted multivariable linear regression models and causal mediation analyses in 1255 older adults of the Dutch Globe study. The neighborhood age structure was measured in 2011 as the homogeneity of the age composition (using the Herfindahl-Hirschman index, range from 0 to 100, a higher score indicating more homogeneity) and the percentage of specific age groups in a neighborhood. Mental health was measured in 2014 by the Mental Health Inventory-5 score (range 0 to 100, a higher score indicating better mental health). Potential mediators were assessed in 2011 and included perceptions of neighborhood social cohesion, feeling at home in a neighborhood, and social participation. Results A more homogeneous age composition (not specified for age) and a higher percentage of children living in a neighborhood were associated with better mental health, the other age categories were not. Social cohesion, feeling at home and social participation did not mediate the associations. Conclusions The neighborhood age composition may be an interesting but currently insufficiently understood entry point for policies to improve older adult’s mental health status.