Although monitoring is an essential tool for biodiversity conservation, monitoring programmes are often poorly designed and thus unlikely to produce results that are meaningful for management. Monitoring is especially challenging when dealing with rare and elusive species in areas where conservation resources are particularly limited. In such cases, monitoring techniques aimed at estimating occupancy represent an attractive alternative to traditional methods concerned with estimating population size, as the collection of detection/non-detection data is in general less costly and easier to implement. In this study, we evaluated the use of occupancy as a state variable for the monitoring of the Alaotran gentle lemur Hapalemur alaotrensis, a Critically Endangered primate exclusively inhabiting the dense marshes around Lake Alaotra in Madagascar. We used a likelihood-based modelling approach that explicitly accounts for detectability. This showed that the probability of detection of H. alaotrensis was extremely low and depended on site characteristics that can vary in space and time, confirming the need to account for imperfect detection when monitoring this species. We used our models to explore factors affecting the probability of occupancy and detection to identify management implications, and also developed recommendations for the ongoing monitoring of this species. The method applied in this study provides an efficient tool for the monitoring of an elusive species and has the potential to provide a flexible sampling framework for local community-based monitoring initiatives.