More than 10,000 research papers were retracted in 2023, setting a new record for scientific publishing as journals grapple with historical fraud amplified by artificial intelligence.
Nature reported that old fabricated studies, such as the extensive Himalayan geology papers by Vishwa Jit Gupta, are now being resurfaced by AI search summaries that direct readers to uncorrected falsehoods.
Gupta published more than 450 publications featuring nonexistent collection sites, plagiarized images, and stolen specimens. Retired geologist Ghulam Bhat completed his PhD on Himalayan geology in 1987, the same year anomalies in Gupta’s work came to light. “It destroyed the image of Indian science,” Bhat told Nature.
The AI Threat to Historical Records
Gupta died in 2022 without any of his discredited papers ever receiving a formal retraction. Many of his articles remain unindexed in major bibliometric databases like Scopus, Web of Science, and Dimensions, yet Google’s AI overview tools pull them directly when users search for geological landmarks such as the Tandi Limestone.
Bhat now plans to rectify the digital record and says that all other journals that have published Gupta’s work should do the same. “If this problem is left as it is, unattended, it will pollute all of the geological literature in future,” Bhat told Nature.
Research-integrity specialist Andrew Grey said that when investigations of old papers are hampered by limited information, an editorial expression of concern is generally the preferred mechanism for alerting readers. Publishers often hesitate to judge past science by modern standards, especially when original authors are dead or records are missing.
Eighty Years From Prank to Retraction
Standardized retraction processes are relatively recent, with the Committee on Publication Ethics developing its first guidelines in 2004 when journals published only around 30 retraction notices annually. Digitization has since enabled journals to reach much farther back into archives.
Data shows that retractions of papers older than 20 years remain rare but are increasing. Analysis indicates that such actions accounted for roughly 0.7% of total retractions since 2015, up from 0.2% before 2010. Meanwhile, about 3% of expressions of concern target works older than 20 years.
When Old Icons Face Re-evaluation
Mass retractions often follow investigations that cast doubt over an entire body of work. A 2019 inquiry by King’s College London into psychologist Hans Eysenck and collaborator Ronald Grossarth-Maticek led to the flagging of papers dating back to the 1940s.
“What is problematic is if false conclusions lead other scientists and the community as a whole astray,” Pääbo told Nature. “I don’t think anyone is currently misled by the mummy paper,” he said, arguing that correcting very old papers with errors would create more confusion than clarity.
