Abstract The conservation of natural habitats requires an accurate assessment of both their degradation and their potential for restoration. Among the tools available, dark diversity is promising, since it estimates all the species that could be present in a community due to favourable environmental conditions but are currently lacking. However, current experimental approaches measure dark diversity mainly from species co‐occurrences, which rely partly on neutral processes, and also suffer from several problems concerning the inclusion of rare species. Here, we propose a novel method to estimate dark diversity, which is based on more deterministic co‐occurrence and also strongly limits the problems of rarity: the co‐occurrence of species' functional features. We adapted the Beals' co‐occurrence index using functional groups, and we estimated functional dark diversity based on co‐occurrence of functional groups. To illustrate the interests of our approach, we then addressed as an example a major issue of restoration ecology: how does passive rewilding (i.e. spontaneous restoration of woodlands) impact the ecological integrity of recovered communities? We compared post‐agricultural woodlands with ancient forests, in terms of taxonomic and functional dark diversity of vascular plants and spiders. For both taxa, the patterns of dark diversity in recent woodlands and ancient forests changed depending on whether we used the taxonomic or functional approach. This was due to the detection of functional features even if they were hosted by rare species, and to the consideration of ecological redundancy between species in functional groups. Notably, recent woodlands harboured plant communities with a similar amount of taxonomic dark diversity compared to ancient forests, but with less functional dark diversity. Also, the composition of functional dark diversity clearly differed between forest types. Synthesis and applications. Functional dark diversity brings novel perspectives for ecological diagnostic and restoration. In particular, it provides a more mechanistic understanding of the constraints which limit the re‐assembly of natural communities during restoration practices (e.g. dispersal limitation). The functional approach usefully complements the taxonomical one to identify habitat conservation and restoration interests, especially for highly degraded sites where the detection of missing groups via species is biased because of their rarity.