Climate-driven impacts on marine trophic pathways worldwide are compounded by sea-ice loss at northern latitudes. For the Arctic, current information describing food-web linkages is fragmented, and there is a need for tools that can describe overarching trophic structure despite limited species-specific data. Here, we tested the ability of a mass-balanced ecosystem model (Ecopath with Ecosim, EwE) to reconstruct the trophic hierarchy of 31 groups, from primary producers to polar bears, in the Canadian Beaufort Sea continental shelf. Trophic level (TL) estimates from EwE were compared with those derived from two nitrogen stable isotope (SI) modelling approaches (SI linear and scaled) to assess EwE accuracy, using a data set of 642 δ
15
N observations across 282 taxa. TLs from EwE were strongly, positively related to those from both SI models (R
2
 > 0.80). EwE performed well (within 0.2 TL) for groups with relatively well-known diets or for taxa characterized by fewer trophic connections (e.g., primary consumers). Performance was worse (>0.5 TL) for species groups aggregated at coarse taxonomic levels, those with poorly documented diets, and for anadromous fishes. Comparisons with SI models suggested that the scaled approach can overestimate the TL of top predators if ecosystem-specific information is not considered.