1. Assemblages of insect herbivores are structured by plant traits such as nutrient content, secondary metabolites, physical traits, and phenology. Many of these traits are phylogenetically conserved, implying a decrease in trait similarity with increasing phylogenetic distance of the host plant taxa. Thus, a metric of phylogenetic distances and relationships can be considered a proxy for phylogenetically conserved plant traits and used to predict variation in herbivorous insect assemblages among co-occurring plant species.

2. Using a Holarctic dataset of exposed-feeding and shelter-building caterpillars, we aimed at showing how phylogenetic relationships among host plants explain compositional changes and characteristics of herbivore assemblages.

3. Our plant–caterpillar network data derived from plot-based samplings at three different continents included >28,000 individual caterpillar-plant interactions. We tested if increasing phylogenetic distance of the host plants leads to a decrease in caterpillar assemblage overlap. We further investigated to what degree phylogenetic isolation of a host tree species within the local community explains abundance, density, richness and mean specialisation of its associated caterpillar assemblage.

4. The overlap of caterpillar assemblages decreased with increasing phylogenetic distance among the host tree species. Phylogenetic isolation of a host plant within the local plant community was correlated with lower richness and mean specialisation of the associated caterpillar assemblages. Phylogenetic isolation had no effect on caterpillar abundance or density. The effects of plant phylogeny were consistent across exposed feeding and shelter-building caterpillars.

5. Our study reveals that distance metrics obtained from host plant phylogeny are useful predictors to explain compositional turnover among hosts as well as host-specific variations in richness and mean specialisation of associated insect herbivore assemblages in temperate broadleaf forests. As phylogenetic information of plant communities is becoming increasingly available, further large-scale studies are needed to investigate to what degree plant phylogeny structures herbivore assemblages in other biomes and ecosystems.

The sampling was conducted in Lanžhot (Czech Republic; 48°42’ N, 16°57’ E; 152 m a.sl.), Toms Brook (Virginia, USA; 38°55’ N, 78°25’ W; 220 m a.sl.), and Tomakomai (Hokkaido, Japan; 42°43’ N, 141°36’ E; 90 m a.sl.). We employed plot-based sampling to investigate insect-plant interactions and the assemblage structure of insect herbivores. At each site, we set up two 0.1 ha plots and sampled caterpillars from all deciduous trees with a DBH ≥ 5 cm. The sampling was carried out during the vegetation seasons between 2013 and 2017, but usually finished within two years at a given site (Czech Republic: May, 2013 – April, 2015; Japan: May, 2014 – July, 2015; USA: April, 2016 – August, 2017).

Folivorous caterpillars were collected by tree felling (USA), by using a canopy crane (Japan), or by using a mobile elevating work platform (Czech Republic). To capture seasonal changes in species composition, the sampling of con-specific tree individuals was spread across the vegetation period. Caterpillars were sampled manually from accessible foliage, assigned a morphotype and unique number, reared, and subsequently identified. Species identifications were based on external morphological characteristics of caterpillars and reared adults, and verified by comprehensive DNA barcoding (i.e., sequencing of a 658 bp fragment of the COI gene). Adult identification was further aided in many cases by genitalia dissections.

Caterpillar species sampled in less than three individuals were excluded from the matrices in order to eliminate erroneous interaction records.