Generative AI fabricates species in citizen science platforms, threatens research integrity
AI-generated fake observations of animals are infiltrating iNaturalist and eBird, potentially contaminating biodiversity research and scientific papers.
Conversation activity · last 2 days peak 1/hr
Summary, timeline and people extracted by Claude from 4 items across 3 sources · 4h ago. Quotes are verbatim.
Researchers including Kris Anderson and Alexander Lees have documented AI-generated fake species observations appearing on citizen science platforms like iNaturalist and eBird, including fake mantises and a fabricated willow tit. As generative AI tools improve, scientists warn that unverified AI-generated observations could contaminate biodiversity research, scientific papers, and environmental assessments if verification processes aren't strengthened.
- Generative AI is now fabricating convincing but entirely nonexistent species observations on citizen science platforms like iNaturalist and eBird, with at least two documented cases in 2025-2026.
- Researchers warn AI-generated observations could infiltrate peer-reviewed scientific papers, environmental assessments, and surveys without adequate verification, contaminating the scientific record.
- AI hallucinations in biological research are particularly dangerous when synthetic data replaces experimental measurements, as subtle corruptions in complex datasets may be nearly impossible to detect.
- The vulnerability exists because verification processes on citizen science platforms and in research workflows are not equipped to detect increasingly sophisticated AI-generated content.
How it unfolded
-
Report Anderson documents second fake mantis on iNaturalist
Kris Anderson reports finding an AI-generated Deroplatys (dead leaf mantis) on iNaturalist with inconsistent anatomical features. The image has since been removed from the platform.
“Because so much modern biodiversity research increasingly incorporates citizen science records, I felt it was important to document this vulnerability early, before AI-generated observations became commonplace.”
Kris Anderson · Hacker News ↗ -
Analysis Warning of cascading AI contamination in science
Anderson warns the problem could spiral and that fake observations might infiltrate scientific papers, surveys, and environmental assessments without adequate verification. Thomas Burger notes that AI hallucinations could be particularly hard to detect when processing genuine data in complex computational workflows.
“AI might disregard a drug candidate that would have worked, direct researchers toward an ineffective treatment, conceal a genuine biological effect, or make a nonexistent disease mechanism look like a discovery.”
Thomas Burger · Mastodon ↗ -
Report ScienceAlert covers AI-invented biological discoveries
ScienceAlert publishes article on how generative AI could create plausible-looking but nonexistent biological discoveries.
-
Report Burger publishes AI hallucination risks in biology
Computational biologist Thomas Burger published an Opinion article in Patterns exploring 10 potential uses of generative AI in biological research and the risks of AI hallucinations creating false biological claims.
- 40 weeks quiet
-
Event Fake willow tit detected in Scotland
Alexander Lees identified an AI-generated image of a willow tit posted to a Scottish birding site, noticing deformed feet and beak that revealed it as fabricated.
What people are saying verbatim
“Because so much modern biodiversity research increasingly incorporates citizen science records, I felt it was important to document this vulnerability early, before AI-generated observations became commonplace.”
Kris Anderson, Mantis specialist, independent researcher · Phys.org ↗
“AI might disregard a drug candidate that would have worked, direct researchers toward an ineffective treatment, conceal a genuine biological effect, or make a nonexistent disease mechanism look like a discovery.”
Thomas Burger, Computational biologist · ScienceAlert / Mastodon ↗
“Most of the time, the problem is not about comparing a hallucination and a genuine biological discovery side by side. It is more about real data having been corrupted by hallucination along the course of the complex computational (genAI-aided) workflow that makes it possible to turn raw signals acquired with complex biotechnologies into biologically valid descriptions of molecular mechanisms.”
Thomas Burger, Computational biologist · ScienceAlert / Mastodon ↗
“I have never thought about that so far, but I guess it is possible to have hallucinations that lead to genuine discoveries.”
Thomas Burger, Computational biologist · ScienceAlert / Mastodon ↗