Monitoring programmes are needed to assess the conservation status of species and to quantify the effectiveness of conservation effort. Rare species are usually poorly studied due to difficulties in monitoring. We evaluated the use of autonomous recording units (ARUs) coupled with automated song recognition for monitoring the presence of Dupont's Lark Chersophilus duponti, a rare and patchily distributed species. We surveyed 49 potential habitat patches for the species in central Spain never censused before, from May to June 2017. In each habitat patch we deployed an ARU that recorded one continuous hour during two consecutive nights. In addition, we also performed line transect censuses in 22 out of the 49 potential habitat patches, to evaluate the efficacy of ARUs. Audio analyses revealed the existence of eight previously unrecorded Dupont's Lark populations. Line transect censuses always confirmed the presence or absence of the species in agreement with audio analysis results. ARUs required 44 fewer working days than human-based surveys for monitoring the presence of Dupont's Larks in potential habitat patches. Our results suggest that the use of ARUs coupled with automated song recognition can provide an effective alternative to human-based surveys for monitoring the presence of bird species in large-scale surveys. We conclude that a combined methodology using ARUs and field censuses, in order to estimate densities or spatial patterns, may be considered the most effective method for monitoring large numbers of potential occupancy sites.