AI Images Trigger 3 Science Paper Retractions as Journals Struggle to Verify Visual Evidence
Updated
Updated · The Conversation · Jun 22
AI Images Trigger 3 Science Paper Retractions as Journals Struggle to Verify Visual Evidence
3 articles · Updated · The Conversation · Jun 22
Summary
Three high-profile retractions — two in 2024 and one by the New England Journal of Medicine in April 2026 — have spotlighted how AI-generated or AI-manipulated images are entering scientific publishing.
AI tools now help researchers create illustrations, synthetic data and edited lab images, but they also blur the line between legitimate enhancement and outright fabrication, especially in fields that rely on visual evidence.
Publishers are adopting AI-detection systems, yet those tools tend to trail fast-improving image generators and can miss subtle distortions that still look scientifically plausible in peer review.
That erosion of reliable visual cues matters beyond journals: when polished images and institutional branding no longer signal authenticity, audiences may judge science visuals more by prior beliefs than by evidence.
Researchers argue the answer is disclosure rather than bans — documenting whether AI was used, what an image represents and how it was verified — so scientific images retain traceable links to reality.