The link between poor reproductive success and diet was investigated in yellow-eyed penguins Megadyptes antipodes, by assessing diet at two localities separated by about 30 km: the north coast of Stewart Island where breeding success is low (0.38–0.67 chicks per pair in recent years), and Codfish Island where breeding success is higher (0.96–1.51 chicks per pair), and relating this to published data from South Island localities, where average breeding success was 1.1 chicks per pair. Diet composition, meal sizes and energetic content of meals and prey were determined from stomach contents, and stable isotope analyses of chick down, fledgling feathers and adult blood provided information on diet throughout the fledging period. The high proportion of stomachs that were empty or lacked diagnostic remains reduced sample size considerably, and variability between samples reduced the power to detect significant differences in meal size, proportions of empty stomachs and prey diversity of meals. Energetic content of Stewart Island meals was less than Codfish Island meals, and there was a non-significant trend for smaller meal sizes and reduced prey diversity among Stewart Island samples. Both localities had lower prey diversity and smaller meals than South Island penguins. Blue cod Parapercis colias accounted for 99% of prey biomass in Stewart Island and 70% in Codfish Island stomach samples, where 27% of prey biomass was opalfish Hemerocoetes monopterygius. Isotopic mixing models carried out on larger sample sizes indicated that opalfish comprised a large proportion of the diet at both locations, with adults selectively provisioning chicks with opalfish while feeding mainly on blue cod themselves. We suggest the large blue cod consumed by Codfish Island and Stewart Island penguins, larger than those consumed by South Island penguins, is difficult to transfer to chicks by regurgitation. Oyster dredging around Stewart Island may have reduced the availability and abundance of alternative prey to Stewart Island penguins.