Sentences without meaning, copy and paste, doctored images and graphics. According to a study carried out by the specialist firm Clear Skies and revealed by the scientific journal Naturearound 70,000 articles that could amount to false content intended for disinformation or to boost careers were published in 2022 in scientific journals, or around 2% of total scientific production that year.
Shaken by the digital transition, which led to an explosion in the number of published papers, shortened proofreading times and simplified cheating methods, scientific publishing houses initially took time to react to this type of content. But pressed by repeated scandals, more and more publishers are calling on software, staff or subcontractors trained specifically to detect these fakes, such as Clear Skies.
Born from this desire to catch up in the fight against fraud, Clear Skies has developed an intelligent algorithm for publishers. The tool is capable of identifying signs of fraud. It is based on a semantic and visual analysis of articles published on OpenAlex, a scientific archive which brings together 48 million articles published since the 2000s. In total, 400,000 of these contents could be fraudulent, suggests the company .
This software, which has been booming in recent years, mainly detects massive fraud and more particularly the use of AI to write false papers in large numbers on the basis of real papers reformulated automatically. The proportion identified by Clear Skies corroborates previous work, such as that of the NGO Committee on Publication Ethics, which also mentions a share of “false papers” of around 2%.
Often the prerogative of scientists ready to do anything to launch their careers, of companies wanting to give a scientific side to their products, or of structures wishing to influence scientific literature and public opinion, these fakes are found in varying proportions of a research field to another and according to the review criteria of the different journals – some give little importance to rigor checks. The medical field appears to be the most affected today.
These are not the only frauds that scientific publishers must respond to: numerous manipulations by researchers themselves have also been revealed in recent years. Even forcing the president of Stanford, one of the best universities in the world, to resign last July. Faking had been detected in papers published under the responsibility of this neuroscientist.