Forest biodiversity assessment relies mostly on indirect indicators based on elements of forest structure used as a surrogate for species. However, most of those indirect indicators used in national and international evaluation processes lack scientific evidence documenting their quantitative link with the biodiversity they are supposed to assess. In addition, those for which the link is established generally concern only one taxon and are rarely considered at the community level. In this context, tree microhabitats – defined as tree-borne singularities such as cavities, conks of fungi or bark characteristics – have recently gained attention as a potential forest biodiversity indicator for species that depend on them for at least a part of their life cycle and that are difficult to detect. However, like most other potential indicators, precise quantitative information on the link between biodiversity indices and microhabitat variables remains scarce. We explored the link between the richness of three taxonomic groups (bats, birds, and saproxylic beetles) and microhabitat indices using a nation-wide multi-taxonomic sampling design comparing 213 plots located inside and outside forest reserves. We hypothesized that the positive effect of setting aside forest reserves on biodiversity conservation was indirectly due to the increase in the proportion of large structural elements (living trees and standing deadwood) which in turn favoured the quantity and diversity of microhabitats. We also hypothesized that microhabitat indices would be better indicators than large and standing dead tree densities, two classical and often used indirect forest biodiversity indicators. We first analysed the response of total species richness and species richness of different groups and guilds (e.g. red-listed species, forest specialists, cavity dwellers...) to several microhabitat indices (density, diversity) to identify the best microhabitat metric. We then used confirmatory structural equation models to assess the indirect effects of management, large structural elements and microhabitats on the diversity of the target species. We first showed that microhabitat indices based on diversity better described biodiversity than indices based on abundance. Structural equation models confirmed that setting aside forest reserves increased the quantities of large trees and standing deadwood, which in turn drove higher quantities and diversity of microhabitats. This significantly increased the richness of several taxonomic groups including notably total richness and cavity dweller richness of both birds and bats. However the magnitude of the link between microhabitat indices and species richness was relatively small. Conversely, the biodiversity of saproxylic beetles was not driven by microhabitat indices and, for this taxon only, models involving direct effects of large structural elements or management performed better. Tree microhabitats appear to be an indicator for part of the taxonomic groups we analysed, but do not constitute a universal biodiversity indicator. They rather have a complementary role compared to large structural elements. Currently, their moderate correlation with biodiversity, as well as thei sensivity to other factors not taken into account in this study (e.g. observer effects), question their transferability to management and policy. Finally, this study calls for replicable and standardized multicriteria methods to validate biodiversity indicators.