There is increasing evidence that regime shifts occur at several scales in ecosystems (from the spatiotemporal alternation of two species to large-scale, ecosystem-level rearrangements). Yet, the theoretical background for understanding these changes is far from clear. Since fishing down in marine ecosystems is well-documented trend, and its top-down cascading effects in food webs have been richly documented, it is a current question whether overfishing, in general, can also influence regime shifts at lower levels. We model simple marine ecosystems by dynamical food webs and investigate the probability of regime shifts emerging among primary consumers. We considered cases where only one of the primary consumers is persistent in the stationary state. By perturbing the death rates in the food web, we studied the circumstances when the previously persistent primary producer is indirectly changed by the previously non-persistent one. Whether and how regime shifts (e.g., change in primary consumers) can occur depends on (1) food web topology (presence of top-predator and alternative producer), (2) the relative strength of perturbation of primary consumers’ death rates, and (3) the dynamical parameters of the recovering consumer. We found that overfishing, food web topology and dynamical parameters together determine the probability of regime shifts. Thus, integrative and complex models are needed in multispecies fisheries.