Rebuilding depleted fish stocks is an international policy goal and a 2020 Aichi target under the Convention on Biological Diversity. However, stock productivity may shift with future climate change, with unknown consequences for sustainable harvesting, biomass targets and recovery timelines. Here we develop a stochastic modelling framework to characterize variability in the intrinsic productivity parameter (r) and carrying capacity (K) for 276 global fish stocks worldwide. We use models of dynamic stock productivity fitted via Bayesian inference to forecast rebuilding timelines for depleted stocks. In scenarios without fishing, recovery probabilities are reduced by 19%, on average, relative to models assuming static productivity. Fishing at 90% of the maximum sustainable rate depresses recovery probabilities by 42%, on average, relative to static models. This work reveals how a changing environmental context can delay the rebuilding of depleted fish stocks, and provides a framework to account for the potential impacts of environmental change on the productivity of wildlife populations more broadly.