A level of peer review can happen for manuscripts submitted to preprint servers.Credit: Wasan Tita/iStock via Getty
The COVID‑19 pandemic had a major effect on scholarly communication, accelerating data sharing, peer review and the publication of preprints. Although preprints have been around for decades, the pandemic underscored the part they play in the rapid, free and open dissemination of scientific knowledge.
The proliferation of artificial-intelligence-generated manuscripts and junk science in the past few years has forced preprints to adapt once again. Some repositories seem to be tightening their quality-control and moderation procedures in an effort to protect against these threats.
Preprint moderators have to strike “a really, really difficult balance” between rapidly sharing new research and protecting the community from flawed or harmful material, says Natascha Chtena, who studies scholarly communication and open science at Simon Fraser University in Vancouver, Canada.
Chtena and Alice Fleerackers, a social scientist at the University of Amsterdam, have been looking into moderation practices at preprint servers. They’ve found that, under specific circumstances, moderators make judgements about the quality of submissions, often using criteria similar to those applied by academic journals.
Although these efforts are well intentioned, legitimate submissions can get caught in the crossfire, says Fleerackers.
Fleerackers, who has been researching the preprint ecosystem for six years, has experienced this herself. In 2025, she submitted a qualitative study on science journalism to a major repository, only to be told that it contained no new data or analysis. It was her second rejection by that repository that year – a stark change from previous years, when she’d had several submissions accepted by the repository in similar research areas.
Making it more difficult for genuine research to be published ins preprints “fundamentally goes against the spirit of open science and the commitment to advancing inclusive, rapid knowledge sharing that is at the heart of preprints’ value — for both science and society”, Fleerackers says.
This raises the question: how much moderation by a repository is too much?
No one-size-fits-all approach
Over the past three decades, dozens of preprint servers have been launched by academic communities, research institutes and publishers, with the goal of making new research publicly available before it has undergone peer review. Speed and accessibility are key qualities of these platforms, many of which emphasize in their policies that they avoid making judgements on the quality and significance of submissions.
It’s something that John Inglis, co-founder of preprint servers bioRxiv and medRxiv, takes seriously. “If there are signals that the provenance of a manuscript is not reliable, we investigate and may decline” it, he says. “But we do not judge scientific quality or value: that is for the community to assess once the manuscript is available.”
A hands-off approach to moderation is becoming increasingly difficult, however. During the pandemic, bioRxiv decided to automatically reject manuscripts that made predictions about treatments solely on the basis of computational work, for example, owing to concerns over misleading health information being shared.
In a preprint posted on the server MetaArXiv in 20241, Chtena, Fleerackers and their co-authors highlight the variety of measures that these repositories are now using to screen submissions. The study, which was based on interviews with 14 managers and moderators of these repositories, including Inglis, found that some sites use advisory boards to make decisions on difficult manuscripts, whereas others rely on the moderators’ “gut feeling” as to whether to accept or reject a submission.
Many interviewees said they used reputation markers or “trust cues”, such as formatting that resembles that of a scientific paper or details of an author’s affiliations and publication record, to judge whether a submission was legitimate.
A repository’s size, scope and capacity can often influence its moderation practices. The manager of RINarxiv, an Indonesian preprint server that receives fewer than 20 submissions each month, speaks directly to the submitting authors to learn about the intention and story behind their manuscript before accepting or rejecting it, Chtena and Fleerackers’ study found.
Nature Index 2025 Research Leaders
At LatArXiv, a preprint server for Latin American research, basic integrity checks, such as verifying the authors’ ORCID profiles and publication history, are performed by a small group of editors, founder and publisher Patricio Pantaleo told Nature Index.
Such methods would be impossible to implement at arXiv, which receives more than 25,000 submissions a month. Instead, the repository uses an ‘endorsement system’ to verify that a submission comes from members of the scientific community. The system requires new submitters to have both an institutional e-mail address and previous authorship of a paper in the same scientific domain. If they do not, they must get a personal endorsement from an established arXiv author.
These requirements were tightened to stem “the flood of low-quality, non-scientific submissions”, arXiv announced in a blogpost in January.
