Aim: The small-island effect (SIE) describes a different relationship between island area and species richness on smaller compared to larger islands. The pattern has recently gained widespread support. However, few studies have attempted to identify the actual mechanisms driving the SIE. Here, we use a phylogenetic community framework to study the SIE, based on the assumption that if the dominant assembly processes differ between small and large islands, patterns of phylogenetic community structure should shift across the area and habitat diversity gradient. Location: The Aegean Archipelago, Greece. Methods: We used a large dataset of 3262 vascular plant species distributed across 173 islands, in combination with a species-level phylogeny. The phylogenetic community structure of each island was calculated using a null modelling framework and was quantified using effect sizes (ES); negative values indicating phylogenetic clustering and positive values overdispersion. Habitat diversity, species richness, phylogenetic diversity (PD) and ES values were regressed against log10 -transformed area and we tested for a SIE using piecewise regression models. We also assessed differences in taxonomic and phylogenetic composition between small and large islands using a beta-diversity framework. Results: We found evidence of a SIE using species richness, PD, and phylogenetic community structure (ES values). Small islands displayed low variation in habitat diversity and tended to be more phylogenetically clustered, while large islands shifted from phylogenetic clustering towards phylogenetic overdispersion with increasing area and habitat diversity. In addition, we showed that phylogenetic composition tended to be more similar between small islands than expected. Main conclusion: Overall, our results provide an example of a SIE in the analysis of island phylogenetic community structure, and point to a role of habitat diversity in driving the SIE more generally.