Background <p>Accurately representing the spatial distribution of organisms is essential for understanding ecological systems and managing natural environments. Conventional surveys for estimating macroinvertebrate abundance in tidal flats need labor-intensive burrow counting. Although model-based and remote sensing approaches have attempted to improve abundance accuracy, they do not fully reflect actual abundance.</p> Results <p>We introduce an advanced approach for estimating the abundance of multiple macroinvertebrate species using drone-based high-resolution orthophotos combined with artificial intelligence (AI)-driven object detection to identify their burrow openings. We constructed training and validation datasets for three macroinvertebrate species (<i>Laomedia</i> sp., <i>Uca arcuata</i>, and <i>Macrophthalmus japonicus</i>) using high-resolution drone orthophotos of biogenic features, including burrow openings and feeding or excretion traces of each species. The datasets were then used to train the YOLO network, and detection accuracy was evaluated through the multi-class identification of the burrow openings of each species in the validation datasets. The multi-class detection accuracy for <i>Laomedia</i> sp., <i>U. arcuata</i>, and <i>M. japonicus</i> averaged 75%, with variation attributed to species-specific burrow characteristics. Notably, the multi-class detection accuracy for the three species was comparable in two regions, with and without pre-established training datasets, suggesting its potential broad applicability and scalability beyond the study area.</p> Conclusions <p>This study represents the first attempt to estimate the abundance of multiple macroinvertebrate species using drones and AI, indicating the potential to develop a method that derives spatial information based on precise measurements reflecting actual abundance.</p>