Conservation policies are starting to consider functional diversity (“FD” hereafter) as a key biodiversity component to safeguard for future generations. However, in practice, it is challenging to measure FD; we have imperfect knowledge about which, and how many, traits and functions are important in a given context, how these traits and functions vary among species and across space, and how they may change in the future. Many researchers have therefore advocated for using a “phylogenetic gambit”; if species traits reflect their shared evolutionary history, then the pattern of that evolutionary history – their phylogeny – should serve as a useful stand-in for unmeasured and unmeasurable traits. The phylogenetic gambit implies that maximizing phylogenetic diversity (PD), i.e. the breadth of evolutionary history, will ensure that a wide variety of forms and functions (e.g. high FD) are captured. Perhaps surprisingly, the validity of this gambit has never been evaluated theoretically nor tested empirically. To be able to fairly compare the conservation outcomes based on the maximization of FD and PD, we first need to use comparable diversity metrics. To do so, we have synthetized research on the “jungle” of PD and FD indices by producing a unifying conceptual framework (Mazel et al, Ecography, 2015; Tucker, […] and Mazel [senior author], 2017, Biological Reviews). Then, we have demonstrated that maximizing PD often fails to maximize FD, by comparing global sets of phylogenetic and functional hotspots (Mazel et al., Glob. Ecol and Biogeo; see also Thuiller, Moraino, Mazel et al, Proc B). To complement my empirical studies on the efficacy of PD as a surrogate for FD, we dug deep into the theory underpinning these tests and discovered some surprising results. Using simulations, we have shown that even when traits harbour phylogenetic signal, selecting species to maximize PD can actually capture less FD than a random set of species (Mazel et al, Sys Bio, 2017). Finally, we have recently conducted a meta-analysis to show that maximizing PD capture, on average across real datasets, 18% of FD compared to maximizing FD directly (Mazel et al, submitted to PNAS). However, while the strategy of maximizing PD does a good job compared to a random strategy in terms of capturing FD across datasets, the effect is rather small and never significant for a given dataset. This means that, for a given conservation action, maximizing PD can be a relatively inefficient strategy to capture FD. I hope these results will be extended to a broader set of functions and clades so that the usefulness of PD in conservation can be finally rigorously assessed.