Marine ecosystem models that incorporate fisheries and climate change are essential for forecasting and guiding sustainable ecosystem management decisions. A key challenge in developing and applying ecosystem models that are able to provide robust predictions for management is to accurately represent the structure and dynamics of food webs. Ecosystem models vary in complexity and formulation and there is no set method routinely used to evaluate the skill of a model to correctly represent food web characteristics. One approach for evaluation is the comparison of modelled food web attributes with measures of stable isotope composition of taxa. While this approach has been used in some studies, its full potential has not been realised. Critically, directly modelling the assumed underlying processes that give rise to stable isotope signatures in ecosystem models has only just begun to be explored. Here, we examine the process of building ecosystem models and assess the potential for incorporating stable isotope results into this process, including the evaluation of model skill. We consider both size- and species- based ecosystem modelling approaches for their potential in this regard. We discuss that whilst conceptually achievable, in practice this is highly challenging through highlighting the advances and challenges in using stable isotope data, including the implications of precision associated with isotope-based measurements. We conclude with a proposed framework for explicitly integrating stable isotope data into both size- and species- based ecosystem models as an example of how signatures may be more powerfully used in the modelling process, and highlight key needs for future work.