Hallucinations Don’t Submit Papers — People Do (and Journals Still Publish Them) Nature's latest "News Feature" manages to turn something embarrassingly simple into a quasi-mystical phenomenon: “hallucinated citations” drifting into the literature as if carried by some invisible atmospheric current. https://lnkd.in/ey7Risw2 Tens of thousands of papers affected, we are told—cue the wide-eyed concern. But nowhere is the obvious stated plainly: these references don’t hallucinate themselves. People put them there. Journals publish them. Instead, the article tiptoes around responsibility with remarkable elegance. Maybe it’s “human error,” maybe it’s “machine error,” maybe it’s somewhere in between. Maybe authors were in a rush. Maybe policies differ. Maybe we just don’t yet understand the “scale of the problem.” It’s all very soothing—if the goal is to ensure that no one is actually held accountable. And then comes the truly astonishing part: we are apparently still “debating” whether citing non-existent papers counts as misconduct. Seriously? A reference list containing fabricated sources—presented as evidence in a scientific argument—and we’re not sure what to call that? Data fabrication seems like a reasonable starting point, but even that is mentioned only cautiously, almost apologetically. Meanwhile, the one solution that would eliminate most of this overnight is never seriously discussed: journals checking whether references exist before publication. Not with AI, not with probabilistic scoring tools, not with elaborate pipelines—just basic verification. The kind of thing one might assume is already part of editorial quality control. Apparently not. The most revealing moment in the article is almost accidental: an editor manually checking references rejects 25% of submissions because of fake citations. Twenty-five percent. Which suggests that this is not some subtle, hard-to-detect edge case—it’s widespread and trivially identifiable if anyone bothers to look. But instead of confronting this, Nature frames the issue as complex, emerging, and technologically challenging—conveniently aligning with the interests of publishers who would prefer new tools, new workflows, and new layers of abstraction over simply doing the job they are already paid to do. This isn’t a mysterious new threat to science. It’s a basic failure of editorial responsibility. And Nature’s refusal to say that outright makes the article less an analysis of the problem and more an exercise in carefully avoiding its most obvious conclusion.
Andrea Rocci
20h
Nature citing journals mega-factory and quasi-predator "Frontiers" as a publisher on the forefront of integrity checking is the cherry on the 🍰. It really shows the direction they want to go. Shame on you "Nature" for this piece!
I died 🤣🤣🤣 But so spot on. The corporatese empty language is everywhere so much so that stating the obvious will soon be nothing short of revolutionary.
cc Vrije Universiteit Amsterdam, the employer of Ronald Meester https://vu.nl/nl/onderzoek/wetenschappers/ronald-meester See below for some backgrounds [in Dutch]. https://www.linkedin.com/posts/klaas-van-dijk-a40672297_in-het-stikstofrapport-van-vrije-universiteit-activity-7438957871565991936-P4gO
“What can be done?” We need to focus on deeper causes if we want to find solutions: https://doi.org/10.2139/ssrn.6223538 [How we drive ΑΙ against Science] The concerns have already begun: https://www.linkedin.com/posts/panagiotis-gioannis-3b8605186_for-open-research-to-work-research-institutions-share-7444688122069573632-iT6Y [For open research to work, research institutions and publishers need to collaborate] https://www.linkedin.com/posts/panagiotis-gioannis-3b8605186_soon-publishers-wont-stand-a-chance-literary-share-7446954093719293952-z7bb [‘Soon publishers won’t stand a chance’: literary world in struggle to detect AI-written books]