Automated plant-identification applications serve as tools in the hands of both enthusiasts and professional botanists, potentially revolutionizing the way we interact with the botanical world. These applications use machine-learning algorithms and image-recognition technology to identify plant species from a simple snapshot. Their utility could extend across various domains, aiding gardeners in identifying and caring for plants, assisting researchers in field surveys for biodiversity assessments, and empowering nature lovers to deepen their understanding of flora. For AI applications to be useful, however, their reliability must be confirmed across diverse species and regions, and in taxa whose identification is challenging. I tested 4 popular and widely available plant-identification applications using images of 518 professionally identified species, representing 174 genera in 36 orders, growing in New York State. Approximately 51% of species were correctly identified by the primary suggestion across all 4 applications. Species indigeneity status (i.e., native vs non-native), local genus representation, and order designation were all found to have significant impacts on identification accuracy for some or all applications. All applications performed reasonably well at identifying plant specimens to genus with the 2 best-performing applications correctly suggesting generic-level identifications for at least 95% of the species. I conclude that automated plant identification applications are useful tools for identifying species at the familial or generic level but require further examination for species-level identification, especially if the species belongs to a species-rich genus or challenging order (e.g., Aquifoliales, Fagales, Gentiales, Sapindales).