One in 277 biomedical papers published this year cited research that doesn't exist.
That's from a Lancet audit of PubMed papers in 2026, and it captures something the publishing industry has been quietly panicking about. Fabricated citations in medical literature have jumped 12-fold in two years, and the source isn't a mystery. AI-generated papers are slipping through peer review faster than anyone can catch them.
"Your doctor could be making decisions around treatment based on studies that never existed," Maxim Topaz, the Columbia researcher who led the audit, said of the findings.
The scale isn't subtle:
- Paper mills, the operations that mass-produce fake studies for hire, are doubling their output every 1.5 years.
- Retractions are only doubling every 3.5 years.
- Only 15 to 25% of paper-mill articles ever get pulled from the literature.
The math gets ugly fast. Springer Nature said it ran nearly 60 AI tools across 1.5 million submissions last year and flagged 25,000 for fake references, fabricated text, or manipulated images. Hindawi, another major publisher, pulled more than 8,000 paper-mill articles in a single year. Those are the ones that got caught.
"The rise of AI has made it easier for unethical individuals to generate fake content," said Chris Graf, Springer Nature's director of research integrity. The tools to spot the pattern are now essential, he said, particularly when fabricated content "appears legitimate at first glance."
Images are a separate problem with the same root cause. Wiley, which runs integrity platforms for its journals, has flagged AI-generated figures being passed off as real scientific evidence as a serious risk to journal credibility. The old detection methods were built to spot copied or doctored photos, not synthetic ones generated from a prompt.
The part that should worry anyone outside the journals is that detection mostly doesn't work yet. The Authors Guild tested five of the most popular AI detectors and found them largely unreliable. One flagged 100% of human-written articles in the test as AI-generated. Another scored a Joan Didion obituary at 66% AI. If editors can't trust the tools, neither can a researcher running a literature review, and neither can a search engine that surfaces a fake paper as a credible source.
Which is where this stops being a publishing problem.
"That's not a problem for me, that's not a problem for the journals," Matt Spick, a health and biomedical data lecturer at the University of Surrey, said. "That's a problem for society."
The publishers can fight this with better detection, and they are. But every fabrication that slips through gets indexed, cited, and eventually pulled into the training data for the next round of models, which means tomorrow's AI is going to be a little more confident about studies that were never run. The cleanup is real, the contamination is faster. The question worth sitting with isn't whether journals can catch up. It's what a citation is even worth when a meaningful share of them point to nothing.
