The first systematic map of evidence syntheses on the use of artificial intelligence in ecology

Sergio Poo Hernandez, Eduardo S. A. Santos , Santiao Ortega, Ayumi Mizuno , Sarah Young, Malgorzata Lagisz , Shinichi Nakagawa 

Artificial intelligence (AI) is increasingly used in ecology to automate data-intensive tasks, from species identification and environmental monitoring to ecological prediction. As primary studies have proliferated, evidence syntheses reviewing AI applications have also increased, but their thematic coverage, methodological emphasis, and reporting transparency remain unclear. We conducted a systematic map, critical appraisal, and bibliometric analysis of 72 evidence syntheses published between 2017 and 2025 on AI applications across the broadly defined field of ecology. Synthesis coverage was strongly concentrated on supervised machine learning and deep learning, particularly image-based classification and prediction workflows. In contrast, reviews of AI applications using acoustic, video, sensor time-series, and multimodal data were comparatively scarce. Explicit comparisons between AI methods and conventional statistical or ecological approaches were rare, as was the synthesis of performance moderators such as data availability, class imbalance, transferability, interpretability, and computational cost. Reporting transparency was generally low to moderate, with recurrent shortcomings in protocol availability, screening and extraction reporting, search validation, language coverage, and sharing of data or code. Bibliometric analyses further indicated uneven geographic representation among authors and collaboration networks. Overall, the review literature on AI in ecology is expanding rapidly, but remains better at cataloguing applications than at evaluating when, why, and under what conditions AI methods improve ecological inference or practice. More transparent, reproducible, geographically inclusive, and benchmark-oriented reviews are needed to support robust and decision-relevant ecological informatics.