During feeding trips, central-place foragers make decisions on whether to feed at a single site, move to other sites and/or exploit different habitats. However, for many marine species, the lack of fine-resolution data on foraging behaviour and success has hampered our ability to test whether individuals follow predictions of the optimal foraging hypothesis. Here, we tested how benthic foraging habitat usage, time spent at feeding sites and probability of change of feeding sites affected feeding rates in European shags (Gulosus aristotelis) using time–depth–acceleration data loggers in 24 chick-rearing males. Foraging habitat (rocky or sandy) was identified from characteristic differences in dive patterns and body angle. Increase in body mass was estimated from changes in wing stroke frequency during flights. Bout feeding rate (increase in body mass per unit time of dive bout) did not differ between rocky and sandy habitats, or in relation to the order of dive bouts during trips. Bout feeding rates did not affect the duration of flight to the next feeding site or whether the bird switched habitat. However, the likelihood of a change in habitat increased with the number of dive bouts within a trip. Our findings that shags did not actively move further or switch habitats after they fed at sites of lower quality are in contrast to the predictions of optimal foraging theory. Instead, it would appear that birds feed probabilistically in habitats where prey capture rates vary as a result of differences in prey density and conspecific competition or facilitation.