The world of science is built on curiosity, collaboration, and collective progress—at least in principle. In reality, however, it has always been marked by competition, inequality, and the potential for error. The scientific community has long feared that these pressures could divert some researchers from the fundamental mission of science: creating credible knowledge. For a long time, fraud appeared to be mainly the work of lone perpetrators. In recent years, however, a troubling trend has emerged: growing evidence suggests that fraud is no longer a series of isolated missteps but an organized, industrial-scale activity, according to a recent study.
The explosive spread of the internet and open-access publishing has facilitated the flow of scientific information—but has also created space for networks specializing in fraud. These “paper mills” mass-produce low-quality or outright fake research, while brokers connect fraudulent researchers with publishers.
The functioning of the scientific world can be compared to a collective “game of trust,” in which scientists, universities, funders, and society work together to expand knowledge. The system is built on good faith: researchers rely on each other’s work, publishers ensure quality control, and institutions and funders depend on professional evaluation.
However, this chain of trust can easily be broken if someone fails to uphold their part. Moreover, scientific performance is now increasingly measured by quantitative indicators—such as the h-index, journal impact factors, or university rankings. These metrics have rapidly become the main criteria for allocating resources and rewards, intensifying competition and inequality, and making the system more vulnerable to fraud.
“Scientific desertion” occurs when someone benefits from the system without conducting genuine scientific work. In a 2002 survey, 0.2% of researchers funded by the US NIH admitted to falsifying data. A more recent study, analyzing over 20,000 articles, found that 3.8% contained improperly duplicated images—half of which were intentionally manipulated. According to some publishers, up to 14% of submitted manuscripts likely originate from “paper mills.”
Recent research shows that the networks carrying out scientific fraud are large, resilient, and growing rapidly. These entities not only produce fake articles but also act as intermediaries in an extensive network where editors and authors collaborate to bypass the traditional peer-review process.
A case study analyzing data from the journal PLOS ONE revealed that certain editors were significantly more likely to accept articles that were later retracted or criticized on PubPeer, a post-publication peer-review site. The study identified 22 editors who accepted a disproportionately high number of subsequently retracted articles and 33 who did the same for articles commented on at PubPeer. The research also identified authors who conspicuously often submitted their work to these “flagged” editors. This points to collusion between editors and authors. A particularly tightly connected group of editors was found to have reviewed each other’s articles between 2020 and 2023, with more than half of the articles they accepted later being retracted. Similar patterns were found in journals from the publisher Hindawi and in IEEE conference proceedings.
The industrial scale of the fraud is also evident from falsified images. A network of 2,213 articles was mapped, linked by shared, duplicated images. The network broke down into large, interconnected clusters, suggesting the articles were produced in large batches from a common image bank. These articles typically appeared in the same journals during the same period, indicating coordinated activity between paper mills and cooperating journals. Despite the clear evidence of fraud from image duplication, only 34.1% of these articles have been retracted.
Organizations specializing in fraud are extremely adaptable. When a journal they use is de-indexed by major scientific databases (such as Scopus or Web of Science)—meaning it is removed due to quality concerns—the fraudsters simply target new outlets. This phenomenon is known as “journal hopping.”
The case of an organization called ARDA illustrates this well. ARDA advertises on its website that it can guarantee publication in certain journals. Their portfolio changes dynamically: in 2018, they offered only 14 journals; by March 2024, this number had grown to 86. When Scopus de-indexed a group of journals they used in 2020 or 2021, ARDA removed them from its portfolio in May 2021 and replaced them with new ones. Of the Scopus-indexed journals listed by ARDA, 33.3% were later de-indexed—compared to just 0.5% for all Scopus journals. This indicates that such organizations deliberately target journals with weak quality control that are still indexed.
Scientific fraud is not evenly distributed across disciplines. Certain rapidly growing and popular fields are particularly attractive targets. One study compared six closely related subfields of RNA biology. While the retraction rate for articles on CRISPR-Cas9 technology was only 0.1%, the rate was dramatically higher for circular RNAs (2.5%), for studies linking microRNAs to cancer (4%), and for research on long non-coding RNAs (lncRNA) (4%). These rates far exceed what could be explained by honest mistakes and clearly point to concentrated fraudulent activity.
Perhaps most alarming is that scientific fraud is growing much faster than legitimate scientific output. While the total number of scientific publications doubles approximately every 15 years, the number of retracted articles doubles every 3.3 years, and the number of suspicious products from paper mills doubles in just 1.5 years. This means fraud is spreading exponentially.
Current punitive measures, such as de-indexing journals, are vastly insufficient given the scale of the problem. While Scopus and Web of Science together de-index about a hundred journals annually, the number of journals publishing content from paper mills is ten times higher. Only a fraction (about 28.7%) of suspicious articles are retracted. Based on current trends, only about 25% of such articles will ever be retracted, and just 10% will appear in a de-indexed journal.
The severity of the situation is compounded by the fact that scientific fraud undermines trust in science and endangers future research. Artificial intelligence and large language models, intended to summarize and analyze scientific literature, are currently unable to distinguish genuine science from fakes. If the literature becomes saturated with fraudulent work, these technologies could unintentionally amplify and spread false information, with unforeseeable consequences. The situation demands urgent, coordinated action from all stakeholders in the scientific community.
