A machine learning-generated and expert-validated bird occurrence dataset, based on 10,691 hours of passive audio recordings systematically collected across different types of dense and compact urban areas in Gothenburg, Sweden. The dataset includes 239,597 occurrence records of 61 species from April 21 to June 16, 2024, across 30 sites in central Gothenburg.In addition to the Darwin Core Archive (DwC-A), the data package includes: species_accuracy_and_misclassification.txt: A supplementary table that includes a list of 206 species validated by the ornithologist. For each species, the table includes its classification accuracy (based on a randomly validated sample of detections) and the probabilities of it being misclassified as other species. species.txt: A list of the scientific names of the 61 species included in the occurrence dataset, along with their taxon IDs (unique Life Science Identifiers) according to the Swedish taxonomic database (Dyntaxa; dyntaxa.se).