Declaring that the business of expert publishing and knowledge curation is under attack is hardly breaking news. What is alarming is that not only is the use of artificial intelligence part of the problem, it is also limited in its usefulness as a tool for publishers to help combat nefarious activity or disseminate editorially sound content.
Take papermills – a report by Frontiers journal publishers found that owing to different approaches to detecting red flags for suspicious submissions in three AI-powered detection tools, only 4.5% of such submissions were detected by all three tools (https://www.frontiersin.org/news/2026/05/06/mind-the-detection-gap-why-publishing-needs-a-multilayered-defense-against-industrial-scale-papermills). The heavy lifting in identifying and eliminating papermill activity still, therefore, mainly rests with human editorial oversight and peer review expertise.
Researching and processing publications to cite supporting references is gradually being supplanted by a reliance on AI-generated overviews in Google (https://scholarlykitchen.sspnet.org/2026/05/06/zero-click-readership-are-ai-overviews-changing-the-way-we-discover-research/), which summarise and interpret selected publications instantly – known as ‘zero-click readership’. Overviews are certainly fast and efficient, but the emphasis here is on ‘selected’, raising questions about bias and oversimplification eroding researchers’ abilities to critically evaluate publications. And that’s before we even talk about meaningless metrics and weakened visibility for journals.
As nobody said ever, but ought to, the price of algorithms is eternal scepticism.