Image: nature.com
The Story
AI-assisted writing is widespread; proven AI-generated fraud is real but much harder to measure.
AI-assisted language is already common, but language signals do not prove that data, methods, or conclusions were fabricated. More concrete evidence comes from retractions, fabricated citations, paper mills, and copycat papers. Current studies show a serious integrity threat, but their estimates cover different populations and kinds of misconduct.
- ScalePrevalence estimates depend on whether researchers count AI-edited prose, fabricated references, copied papers, or wholly fake studies.
- DetectionText-pattern methods can identify likely AI involvement, but they cannot by themselves establish research misconduct.
- RiskFabricated references and mass-produced papers can enter databases and slow the discovery of reliable evidence.
- Fourteen percent of biomedical abstracts indexed in PubMed in 2024 contained words commonly associated with large language models, according to an analysis of 1.5 million abstracts.
- A 2026 audit examined fabricated citations across 2.5 million biomedical papers and identifies nonexistent references as a direct threat to evidence evaluation.
- An earlier systematic study estimated 75 SCIgen papers per million in information and computing sciences, while only 19% of detected papers received formal retraction or silent removal.
Deeper Insight
Scientific Reports, a Springer Nature research journal, is on hold. With 1,400 papers retracted in one day, output is overtaking knowledge.
One day brought 1,400 retractions. Now Scientific Reports, a major research journal from Springer Nature, is on hold in PubMed while the National Library of Medicine reevaluates it. Add fabricated citations, paper mills, and AI-assisted drafting, and the problem is bigger than bad papers. A crowded record can bury the good ones.
- What it could changeReliable research could become harder to find than polished research built to imitate it.
- How it touches todayDoctors and researchers already sift through papers that may contain fake citations or fabricated work.
- Our futureScience could become a numbers game where more published papers mean less trustworthy knowledge.
- The Lancet commission estimates that around 3% of recent biomedical articles are fake publications and says the rate is increasing because of paper mills and inappropriate generative-AI use.
- A review of the fraud industry estimates that several hundred thousand fake papers may circulate, while noting that the exact scale is difficult to determine.
- Research on GPT-fabricated papers found that questionable AI-produced papers can appear in Google Scholar beside reputable research, especially on health, environmental, and computing topics.
Sources · 8
Featured · Primary source · thelancet.com
The Lancet–World Conferences on Research Integrity Foundation Commission on Research Integrity - The Lancet
The Lancet–World Conferences on Research Integrity Foundation Commission on Research Integrity - The Lancet # The Lancet–World Conferences on Research Integrity Foundation Commission on Research Integrity Correspondence Send email to [email protected] Department of Epidemiology and Data Science, Amste
Read sourceAnalysis · misinforeview.hks.harvard.edu
GPT-fabricated scientific papers on Google Scholar: Key features, spread, and implications for preempting evidence manipulation | HKS Misinformation Review
GPT-fabricated scientific papers on Google Scholar: Key features, spread, and implications for preempting evidence manipulation | HKS Misinformation Review September 3, 2024 # GPT-fabricated scientific papers on Google Scholar: Key features, spread, and implications for preempting evidence manipulat
Read sourceAnalysis · theconversation.com
Fake papers are contaminating the world’s scientific literature, fueling a corrupt industry and slowing legitimate lifesaving medical research
Fake papers are contaminating the world’s scientific literature, fueling a corrupt industry and slowing legitimate lifesaving medical research Assistant professor Frank Cackowski, left, and researcher Steven Zielske at Wayne State University in Detroit became suspicious of a paper on cancer research
Read source