AbstractAimWe examined potential environmental drivers of broad‐scale spatial patterns in the trophic structure of marine ecosystems as represented by nitrogen stable isotopes in globally distributed marine predators. Additionally, we assessed the effects of spatial scale on the predictive capabilities of environmental variables.LocationGlobal oceans.Time period2000 to 2015.Major taxa studiedTunas: Thunnus albacares, Thunnus obesus, Thunnus alalunga.MethodsWe undertook a global compilation and meta‐analysis of the bulk nitrogen stable isotope ratios (δ15N values) of three tuna species (n = 4,281). After adjusting for regional variations in baseline δ15N values using a global ocean biogeochemistry model, generalized additive mixed models were employed to infer global‐scale oceanographic controls of trophic structure, using cosmopolitan tuna species as a model.ResultsFor the three tuna species, variation in trophic position estimated using bulk δ15N values was largely explained by geographical location and the corresponding oxygen minimum layer depth. Tuna trophic positions declined in areas with reduced oxygen at depth. Food‐chain length, as captured by maximum trophic position, was longer in areas of the western Pacific Ocean and shorter in the northern Atlantic and eastern Pacific Oceans. Trophic adaptability of the tuna predators, as indicated by intraspecific variability, was highest in the western and central Pacific Ocean and lowest in the northern Atlantic Ocean. Our analysis demonstrated that while tunas share similar functional trophic roles, deeper‐foraging tuna species had higher trophic positions globally. The predictive capacity of environmental variables decreased at finer (regional) spatial scales.Main conclusionsOur work suggests that habitat compression resulting from the predicted global expansion of oxygen minimum zones with ocean warming will impact the trophic structure of marine food webs and the corresponding foraging habits of marine predators. Spatial scale analyses highlighted the importance of representing differences in regional ecological dynamics in global‐scale trophic and ecosystem models.