Endangered yellow-eyed penguins (Megadyptes antipodes) are central-place, benthic-diving foragers that search for prey in the productive marine areas off the coast of the South Island, New Zealand. Like other seabirds, they target specific, reliable areas of high prey abundance, which are often associated with oceanographic characteristics such as bathymetry, seafloor sediment type, and sea surface temperature. Employing GPS tracking data collected between 2003 and 2021, we created species distribution models using maximum entropy modelling (Maxent) to determine foraging space use and habitat suitability for yellow-eyed penguins across their entire South Island range and within five distinct subpopulations: Banks Peninsula, North Otago, Otago Peninsula, the Catlins, and Stewart Island. We quantified the importance of environmental variables for predicting foraging site selection during and outside the breeding season. Significant regional variation existed in predicted probability of penguin presence, and proximity to the nearest breeding area was a key predictor of suitable foraging habitat. When distance was not included in the models, dissolved oxygen concentration was the most important predictor in the overall South Island model and the North Otago, Otago Peninsula, and the Catlins subpopulation models, whereas water current speed and mean monthly turbidity were most important in Banks Peninsula and Stewart Island subpopulation models, respectively. Dynamic variables related to prey availability were often the most important variables in model predictions of the habitat selection of yellow-eyed penguins. Visualisations and findings from this study, particularly of the observed interactions between penguins and their marine habitat, can be used to direct conservation and resources during marine spatial planning and species management.