Adequate nutrition is a fundamental requirement for the maintenance and growth of populations, but complex interactions between nutrients and plant toxins make it difficult to link variation in plant quality to the ecology of wild herbivores. We asked whether a 'foodscape' model of habitat that uses near-infrared spectroscopy to describe the palatability of individual trees in the landscape, predicted the foraging decisions of a mammalian browser, the koala (Phascolarctos cinereus). Specifically, we considered four behavioural decision points at which nutritional quality may influence an animal's decision. These were: which tree to enter, whether to feed from that tree, when to stop eating, and how long to remain in that tree. There were trends for koalas to feed in eucalypt trees that were more palatable than unvisited neighbouring conspecific trees, and than trees that they visited but did not eat. Koalas ate longer meals in more palatable trees, and stayed longer and spent more time feeding per visit to these trees. Using more traditional chemical analyses, we identified that an interaction between the concentrations of formylated phloroglucinol compounds (a group of plant secondary metabolites) and available N (an integrated measure of tannins, digestibility and N) influenced feeding. The study shows that foodscape models that combine spatial information with integrated measures of food quality are a powerful tool to predict the feeding behaviour of herbivores in a landscape.