AI-Altered Images on Citizen Science Platforms Are Putting Scientists on Alert
Scientists have warned that AI-manipulated photographs are producing false records on birdwatching platforms, threatening the reliability of scientific research. In a commentary published in Nature, researchers reported that hundreds of fake or AI-altered images had been identified on databases such as iNaturalist and the Macaulay Library. The paper’s lead author, Dr. Alexander Lees, an ecologist at Manchester Metropolitan University, argues that the core problem is not deliberate fraud but users asking AI tools to touch up their photos. During such edits, however, the AI can graft features from another species onto the original image.
As a concrete example, Lees points to a reported sighting of a red-winged blackbird in the Brazilian interior, a species with no prior record in the region and normally found in North America. The bird in the photo was in fact an epaulet oriole, a species in the Icteridae family. The photographer had used AI to enhance the image, and the software added blackbird-like features, producing a false observation record. Researchers say such AI interventions are undermining trust in the platforms scientists rely on.
Citizen science platforms have yet to fully gauge the scale of the problem. Of iNaturalist’s more than 610 million images, only around 1,400 have been flagged for AI use. Officials acknowledge that many AI-touched images likely pass through moderation undetected. iNaturalist’s Outreach and Community Coordinator, Tony Iwane, notes that most cases aren’t malicious, which is exactly what makes the problem harder to catch.