Abstract
Citizen science platforms like iNaturalist play a crucial role in biodiversity monitoring. However, the identification of plants from the genus Castilleja (Indian Paintbrush), which comprises about 200 species with often similar features and frequent introgression, presents considerable challenges. Our study examines the agreement between initial identifications (often made by computer vision algorithms), subsequent identifications, and the final Research-Grade identifications for Castilleja species on iNaturalist. We focus on prevalent identification problems within this genus, particularly noting that Castilleja densiflora and Castilleja exserta are most frequently confused. This study highlights the need for improved algorithms to enhance initial species identification accuracy, especially for complex genera like Castilleja. Our findings have implications for the efficiency of the identification process on citizen science platforms and underscore the importance of expert verification in challenging taxonomic groups.