Rationale: Increasing demand for fish and seafood means that traceability of marine products is becoming ever more important for consumers, producers and regulators. Highly complex and globalised supply networks create challenges for verifying the claimed catch region. Atlantic cod is one of the most commercially important species in the northeast Atlantic. Several regional fisheries supply cod into the trade network, of which some are at more risk of overexploitation than others. Tools allowing retrospective testing of spatial origin for traded cod products would significantly assist sustainable harvesting of wild fish, reducing incentives for illegal fishing and fraud. Methods: Here we investigate whether stable isotope ratios of carbon, nitrogen and sulfur in muscle tissue can be used to identify the catch region of Atlantic cod ( Gadus morhua). We measured the isotopic composition of muscle tissue from 377 cod from ten known catch regions across the Northeast Atlantic and Northeast Arctic, and then applied three different assignment methods to classify cod to their region of most likely origin. The assignment method developed was subsequently tested using independent known-origin samples. Results: Individual cod could be traced back to their true origin with an average assignment accuracy of 70-79% and over 90% accuracy for certain regions. Assignment success rates comparable to those using genetic techniques were achieved when the same origin regions were selected. However, assignment accuracy estimated from independent samples averaged c25% overall. Conclusion: Stable isotope techniques can provide effective tools to test for origin in Atlantic cod. However not all catch regions are isotopically distinct. Stable isotopes could be used in conjunction with genetic techniques to result in higher assignment accuracy than could be achieved using either method independently. Assignment potential can be estimated from reference datasets, but estimates of realistic assignment accuracy require independently collected data.