Ecological monitoring is essential for understanding biodiversity amid rapid global change. Within wildlife monitoring, Passive Acoustic Monitoring (PAM) has become a powerful and commonplace tool. Advances in computational methods—fuelled by affordable data collection and machine learning—have accelerated its growth, yet challenges persist, particularly in data integration, species coverage, and classification model evaluation. In this paper, we systematically review the state of PAM, analysing approximately 80 studies (2005–2024) spanning diverse taxa and ecosystems. We evaluate its strengths, limitations, and applications. Our findings highlight significant increases in dataset sizes over time, with spectrograms dominating preprocessing methods. However, there remains a lack of consensus on evaluation methods for machine learning models, and code sharing has outpaced data availability. We propose a roadmap for future research for closing identified gaps, which includes advocating for open-sourced data and code, improving data preprocessing, and providing accessible model hosting. Broad uptake of these recommendations will enhance impact of computational bioacoustics on ecological science.