Bats are ecologically and economically important mammals, yet their global decline underscores the need for effective monitoring strategies. Acoustic identification through echolocation calls has become a cornerstone for bat research and conservation, but the reliability of automated methods remains uncertain. We conducted a systematic literature review of 531 articles published between 1990 and 2022, of which 207 met the inclusion criteria. Across these studies, we identified 34 echolocation analysis methods, including commercial software and supervised learning approaches. Accuracy was reported in 28 studies for 18 methods. Most tools achieved over 70% correct classifications, but performance varied widely depending on the species, environment, and software version. Comparative evaluations revealed that machine learning techniques, particularly convolutional neural networks and random forests, consistently outperformed commercial software packages such as Kaleidoscope and SonoChiro. Nevertheless, even the best-performing automated tools were not sufficiently reliable to be used without expert verification. Factors such as call type, recording conditions, and reference library quality strongly influenced classification outcomes. The evidence highlights the importance of complementing acoustic monitoring with other methods, such as capture-based methods, and building comprehensive reference call libraries that capture inter- and intra-specific variation. Overall, automated approaches substantially improve data processing efficiency but must be applied with caution to avoid misinterpretations. We recommend integrating acoustic methods with traditional techniques and advancing collaborative efforts to expand global call libraries, ensuring more robust monitoring of bat populations in the face of accelerating biodiversity loss.
Pino-Aedo et al. (2026) studied this question.