AI-Generated Scientific Images Are Fueling a Crisis of Trust in Research

In April 2026, a breathtaking new space photograph captured widespread public attention: it showed Earth glowing vividly against the inky darkness of deep space, with the Moon’s crater-pocked horizon stretching across the entire foreground. Shot by astronauts during NASA’s Artemis II mission, the image echoed the famous 1968 Apollo 8 “Earthrise” photo – and like that iconic shot, it immediately struck viewers as deeply authentic and profoundly inspiring.

But today, artificial intelligence lets nearly anyone generate a visually near-identical version of this same image from a simple text prompt in seconds. With that ability now widely accessible, how can ordinary people even distinguish which images are rooted in real observation, and which are entirely fabricated?

The growing spread of AI-generated scientific imagery in public spaces is far more than just a standard misinformation problem. As a researcher who studies visual science communication and public trust in research, I argue this trend is fueling a full-blown trust crisis in science in the AI era. The long-standing tools scientists have relied on for decades to establish the credibility of their visual work are rapidly losing their power.

AI tools are already reshaping how scientific visuals are created, shared and promoted to the public. Researchers use these systems for everything from generating conceptual illustrations and building synthetic datasets to editing raw lab images and creating content for education and public outreach. While AI can help scientists explain complex, nuanced ideas more creatively and efficiently than ever before, these same tools blur the once-clear lines between conceptual illustration, minor image enhancement, and full data fabrication.

High-profile cases have already made these risks clear: in 2024, two peer-reviewed research papers were retracted after reviewers discovered they included AI-generated figures that depicted biologically impossible structures. In April 2026, the New England Journal of Medicine retracted a published paper after confirming one of its core clinical images had been manipulated with AI. These are only the incidents that have gained widespread public attention, and they are almost certainly just the tip of the iceberg. Researchers have already warned that AI-generated visuals pose a fast-growing threat to integrity in fields that depend heavily on visual evidence, such as materials science.