AI-Generated Images Threatening Credibility of Citizen Science Platforms
Key Takeaways
- ▸Hundreds of AI-generated or AI-enhanced images have been discovered on citizen science platforms, with the true scale of the problem likely much larger due to undetected fakes
- ▸AI image editing can inadvertently introduce biological inaccuracies by mixing features from different bird species, creating false sightings of species outside their normal ranges
- ▸The credibility of popular platforms like iNaturalist and Macaulay Library—which scientists rely on for ecological research and monitoring species habitat range—is being compromised by AI-generated content
Summary
Scientists are warning about a growing threat to citizen science platforms like iNaturalist and Macaulay Library: AI-generated and AI-altered images that create false species sightings and undermine the credibility of research data.
The rise of generative AI tools such as ChatGPT and Google Gemini has led to hundreds of fake or enhanced bird photos appearing on wildlife forums, with researchers noting that the true scale of the problem remains unknown as many fake images may go undetected.
Dr. Alexander Lees from Manchester Metropolitan University, who authored a recent commentary in Nature, highlighted how AI editing can inadvertently introduce parts from different bird species to create false records—such as a red-winged blackbird sighting in Brazil that turned out to be an epaulet oriole after AI enhancement.
While outright hoaxes are relatively rare and easy to spot, the more insidious threat comes from photographers using AI to 'improve' photos, which then contaminates the large-scale datasets that scientists rely on to monitor species ranges and track responses to climate change.
- Citizen science organizations are still working to understand the magnitude of the issue, with only 1,400 of iNaturalist's 610 million images flagged for AI use so far
Editorial Opinion
The collision between generative AI's capability for rapid image creation and the scientific community's reliance on crowdsourced data reveals a fundamental challenge for the AI era: powerful tools designed for creativity and efficiency can easily be misused—intentionally or accidentally—to undermine the very systems that depend on human trust and accuracy. While the most egregious fakes are easy to spot, the real danger lies in subtle alterations that slip through, corrupting datasets that inform critical environmental research. As these systems become more powerful and accessible, guardrails around their use in data-sensitive domains must evolve in tandem.
