We develop an individual‐based, stochastic population model to evaluate the potential effectiveness of captive‐breeding and release programmes, illustrated by the critically endangered Ardeotis nigriceps Vigors great Indian bustard. Successful implementation of captive breeding, involving permanent retention of 20 breeding females and release of surplus juveniles, required collection of many wild eggs and consistent ‘best possible’ performance across all aspects of the programme. Under ‘full‐range’ and ‘above‐average’ scenarios, captive population extirpation probabilities were 73–88% and 23–51% respectively, depending on egg collection rates. Although most (73–92%) ‘best possible’ programmes supported releases, re‐establishment of free‐living adults also required effective in situ conservation. Incremental implementation of effective conservation measures over the initial 10 years resulted in more free‐living adults within 35 years if eggs were left in the wild without attempting captive breeding. For the great Indian bustard, rapid implementation of in situ conservation offers a better chance to avoid extinction than captive breeding. Demographic modelling of threatened species should be used to examine whether captive breeding will bring net benefits to conservation programmes.