In April 2026, a photograph of Earth glowing in deep space, taken by astronauts aboard NASA's Artemis II mission, drew comparisons to the iconic Apollo 8 "Earthrise" image. But that same month, academic journals retracted papers over AI-manipulated figures, and researchers warned that generative tools are making it impossible for nonexperts to tell real scientific images from fabricated ones. The problem isn't just misinformation - it's a growing crisis of trust in the visual evidence science depends on.
Nan Li, an associate professor of science communication at the University of Wisconsin-Madison, studies how people judge the credibility of scientific visuals. In a recent analysis, she argues that AI tools are eroding the three mental shortcuts audiences have long used to evaluate science images: technical sophistication, institutional source, and alignment with prior beliefs. "When visual quality and institutional attribution become unreliable cues for judging the credibility of science images, people tend to fall back on something else: their own prior beliefs," Li wrote.
Fake images are already reaching journals
The problem is not hypothetical. In 2024, two papers were retracted after publishing AI-generated figures with biologically impossible structures. In April 2026, the New England Journal of Medicine retracted a paper after discovering a clinical image had been manipulated with AI. Li notes these cases are likely "just the tip of the iceberg," with researchers warning that fields heavily dependent on visual evidence, such as materials science, face the greatest risk.
Academic publishers are beginning to adopt AI-detection tools, but Li cautions that detection systems will always lag behind generation systems. Detectors can only identify patterns they were trained to recognize, and as new AI models emerge, developers must constantly retrain their systems. The biggest concern, she says, is realistic-looking visuals that subtly distort scientific details while remaining believable enough to pass initial review.
Transparency over restriction
Li doesn't argue for banning AI in scientific communication. Instead, she proposes that researchers treat image provenance with the same seriousness they apply to data provenance. Scientists already disclose funding sources, methodologies, and conflicts of interest - similar standards should now apply to images. Was AI used to generate or modify this image? Is it a direct observation, a simulation, or an illustration? Can it be replicated by other researchers?
Her research suggests disclosure can help. In one study, Li and her colleagues found that people familiar with AI tools were more likely to view AI disclosure as a sign of transparency, and some rated clearly labeled AI-generated content as more credible than unlabeled content. "Transparency gives audiences the necessary context to evaluate what they are seeing, but it may not resolve every dispute about how images are made," she wrote.
For professionals working in AI for Science & Research, the implications are direct: the same Generative AI and LLM tools that speed up scientific communication can also undermine it. Responsible use requires honesty, adherence to professional norms, and the collective development of evidence-based standards across fields.
Why authentic images still matter
Li distinguishes between an image's visual quality and its authenticity. "Authenticity is a documented relationship between an image and the world," she wrote. The Apollo 8 photograph carries emotional weight not just because it's beautiful, but because viewers know there were astronauts, physical cameras, documented missions, and verifiable observations behind it. The Artemis II images carry the same traceable connection.
In the age of generative AI, scientific institutions can no longer assume audiences will automatically trust their visuals. Without guidelines and standards, Li warns, "science risks entering a world where every image can be questioned and no image carries inherent credibility."
Why this matters for science and research professionals
If you publish or review scientific work, the practical takeaway is to document image provenance now, before it becomes a retraction issue. That means recording whether AI was used to generate or modify any visual, what the image represents versus what it illustrates, and how it was verified. It also means treating AI disclosure as a professional norm rather than an admission of weakness - Li's research shows audiences increasingly read it as a sign of integrity. For researchers who rely on visual evidence to communicate findings, adopting these standards is the difference between an image that persuades and one that invites suspicion.
Your membership also unlocks:
