Islands make up a large proportion of Earth’s biodiversity, yet are also some of the most sensitive systems to environmental perturbation. Biogeographic theory predicts that geologic age, area, and isolation typically drive islands’ diversity patterns, and thus potentially impact non-native spread and community homogenization across island systems. One limitation in testing such predictions has been the difficulty of performing comprehensive inventories of island biotas and distinguishing native from introduced taxa. Here, we use DNA metabarcoding and statistical modeling as a high throughput method to survey community-wide arthropod richness, the proportion of native and non-native species, and the incursion of non-natives into primary habitats on three archipelagos in the Pacific - the Ryukyus, the Marianas and Hawaii - which vary in age, isolation and area. Diversity patterns largely match expectations based on island biogeography theory, with the oldest and most geographically connected archipelago, the Ryukyus, showing the highest taxonomic richness and the lowest proportion of introduced species. Moreover, we find evidence that forest habitats are more resilient to incursions of non-natives in the Ryukyus than in the less taxonomically rich archipelagos. Surprisingly, we do not find evidence for biotic homogenization across these three archipelagos: the assemblage of non-native species on each island is highly distinct. Our study demonstrates the potential of DNA metabarcoding to facilitate rapid estimation of biogeographic patterns, the spread of non-native species, and the resilience of ecosystems.