Ecological restoration aims at recovering biodiversity in degraded ecosystems, and it is commonly assessed via species richness. However, it is unclear whether increasing species richness in a site also recovers its functional diversity (FD), which has been shown to be a better representation of ecosystem functioning. We conducted a quantitative synthesis of 30 restoration projects and tested whether restoration improves FD. We compared actively and passively restored sites with degraded and reference sites with respect to four key measures of FD (functional richness, evenness, dispersion, and turnover) and two measures of species diversity (richness and evenness). We separately analyzed longitudinal studies (which monitor degraded, reference, and restored sites through time) and space-for-time substitutions (which compare at one point in time degraded and reference sites with restored sites of different ages). Space-for-time studies suggested that species diversity and FD improved over time. However, replicated longitudinal data showed no sustained benefits of active or passive restoration for FD measures, relative to degraded sites. This could suggest that the positive results in space-for-time designs may have been unreliable, but the relatively short duration of longitudinal studies suggests a need for longer-term longitudinal research to robustly demonstrate the absence of any effect. These differences across study designs may explain the variable results found in recent studies directly measuring the response of FD to restoration. We recommend that future assessments of ecological community dynamics include control sites in monitoring, to ensure that the consequences of treatments, including but not limited to restoration, are correctly partitioned from unassisted temporal changes.