Clara Bennett: Before we get into anything else — did you see the Guardian piece from August 2026?
Max Rivera: The paper mills thing? Yeah, I did, and I — wait, walk me through it, because I want to make sure I'm not overstating what actually happened.
Clara Bennett: One journalist, tracking these networks, documented over twenty unsolicited solicitations in a single 24-hour stretch — Telegram, WhatsApp, Facebook groups — each one offering to sell authorship on a fabricated research paper. Not hacking. Not back-channel academic politics. Openly advertised, like spam.
Max Rivera: So people are — I mean, the messaging app is just the storefront. They're selling fake papers like someone sells refurbished laptops in a Facebook group.
Clara Bennett: Exactly that. Paper mills are organized commercial operations. The product is fraudulent scientific credibility. And the demand engine is publish-or-perish — careers and grant funding contingent on publication volume.
Max Rivera: So the incentive structure basically created a black market and nobody closed the kitchen.
Clara Bennett: In practice, peer review was never built for this. It's a health inspector who only visits when scheduled. These paper mills are cooking at three in the morning, no inspection, no oversight.
Max Rivera: And a journalist had to be the one to walk into that kitchen.
Clara Bennett: And what the journalist found wasn't just the paper mills — it was the scale underneath them. Retraction Watch has been tracking this since 1975, and the fraud-related retractions are growing at 13.3% annually. That's not drift. That doubles the problem roughly every five years.
Max Rivera: Wait — doubles every five years? So whatever we're counting today, in a decade it's four times that?
Clara Bennett: In principle, yes. And 2023 already crossed 10,000 retractions in a single year. New record.
Max Rivera: Ten thousand. In one year. I mean — okay, I want to steelman the other read here, because the PMC analysis actually flags this: maybe rising retractions mean the system is catching more? Like, heightened scrutiny, not more fraud necessarily?
Clara Bennett: That framing is real. Detection improving is not the same as fraud shrinking. But the retraction lag complicates both versions of that story.
Max Rivera: The lag — yeah, this is actually the part that — wait, this is what keeps me up, honestly. A paper gets published. It circulates. Gets cited by three other papers. Maybe it's shaping a clinical protocol somewhere. And then, what, two years later someone pulls it?
Clara Bennett: Retraction Watch data shows the median lag spans a 40-fold range depending on the publisher. Forty-fold. Some papers get caught in months. Others circulate for years, cited the whole time.
Max Rivera: A 40-fold range — so the retraction is kind of a fiction of cleanup, right? The bad science is already embedded in the literature. Other studies built on it. The correction runs in a journal nobody outside the field reads.
Clara Bennett: That's the structural problem. Peer review was never adversarial by design — it assumed honest actors. So now the question is whether post-publication review, including journalism, can actually close that gap before the damage compounds.
Max Rivera: But that gap — post-publication, journalism filling it — that's not a designed handoff. That's an accident. And the truth is: there are historical cases where journalism didn't just help. It was the whole mechanism. Cold fusion, 1989. The vaccine-autism fraud, 1998. Those weren't caught internally. Scientists didn't flag them. Reporters did.
Clara Bennett: And that's exactly what Michele Catanzaro and Luiz Peres-Neto at the Autonomous University of Barcelona formalized. Their study in Media and Communication defines this as watchdog science journalism — not just covering findings, but actively scrutinizing funding sources, ethical considerations, reproducibility. It's a distinct practice with a name now.
Max Rivera: Wait — so there's a formal definition? Like, someone drew the line between science reporter and watchdog?
Clara Bennett: Catanzaro and Peres-Neto did, yes. And Alice Fleerackers built on it empirically — 21 semi-structured interviews with working science journalists, published in Journalism Practice, documenting that these reporters are actively probing misconduct. Not accidentally. Deliberately.
Max Rivera: Twenty-one interviews — I mean, that's qualitative, not a census. But what it tells you is there are journalists who see this as their actual job, not a side effect of covering a story.
Clara Bennett: Right — and the Fleerackers study links directly to post-publication peer review as the layer journalism helps constitute. That's the framing: journalists aren't replacing peer review, they're forming a second accountability layer after publication. Citizens, by the way, explicitly endorse this — particularly in health and climate reporting.
Max Rivera: So here's what's weird to me — picture a researcher somewhere, say a cardiologist in Lyon, whose dataset is quietly fabricated. Peer reviewers read the methods section and it looks clean. Then a journalist, not a cardiologist, reads the funding disclosure and finds a pharmaceutical contract that wasn't disclosed. That's — the journalist caught what the expert missed.
Clara Bennett: Because peer review wasn't designed adversarially. Watchdog journalism is. That's the structural difference Catanzaro and Peres-Neto are pointing at. And honestly — the part that comes later makes this whole picture more complicated, because the same newsrooms doing this work are making story decisions based on feasibility, not public importance.
Max Rivera: Yeah — and that's, I mean, that's the version of this story that should make everyone a little uncomfortable.
Clara Bennett: And the Fleerackers study is where that gets uncomfortable in a very specific way. It's not just that newsrooms are stretched — it's that when resources are thin, story feasibility starts outweighing public importance in coverage decisions. That's documented. Not assumed.
Max Rivera: Wait — say that again. Feasibility over importance? Like, a scientist commits fraud in, I don't know, a hard-to-explain corner of epidemiology, and a reporter just... doesn't chase it because the story won't land with a general audience?
Clara Bennett: That's exactly the mechanism. And the editorial independence piece matters here too — Fleerackers documents adequate staffing and editorial protection as prerequisites for watchdog function. Remove either one, and the scrutiny collapses to whatever is explainable in 800 words.
Max Rivera: So the fraud that's hardest to understand is also the fraud most likely to go uncovered. That's — I mean, that's almost a perfect inversion of what you'd want from a quality-control layer.
Clara Bennett: And it creates a trust asymmetry that's worth naming. Public trust in science rests on competence and integrity. Trust in science journalism rests on selection credibility — did they pick the right story — and accuracy. Those are different contracts. So when an under-resourced reporter does challenge a well-funded lab, the public isn't sure whose credibility to weight.
Max Rivera: Right — but the same dynamic runs in reverse too. A journalist covers a promising finding, it spreads, and then the retraction comes two years later and nobody sees it. Journalism is cure and disease at the same time. That's not a metaphor — that's a logistics failure.
Clara Bennett: Now add AI into that. If you solve the resource problem by using an AI system as the first interpreter of biomedical research — summarizing trial results, flagging significance — you've introduced generalization bias. These systems over-extend findings beyond their actual scope. And no existing watchdog structure, not peer review, not journalism, was built to audit that.
Max Rivera: So the watchdog needs a watchdog and nobody built one.
Clara Bennett: In practice — no. The accountability structures we have were designed for human actors making legible decisions. An AI flattening a clinical trial's scope limitations into a shareable summary doesn't look like fraud. It looks like simplification. And that's the part that won't show up in Retraction Watch.
Max Rivera: And that's — I mean, that's the reality. Nobody voted for this. Nobody sat down and said: journalists will be the fourth pillar of scientific integrity. It just... became true. The Guardian investigation, the Retraction Watch numbers, cold fusion, the vaccine-autism case — all of it accumulated into a job that journalism is doing without ever having been hired for it.
Clara Bennett: The paper mill messages are still coming in. Someone is reading them. It's just sitting with me — the system catching industrialized fraud is a reporter checking a WhatsApp notification.
Max Rivera: Yeah. And I don't know if that's enough. I genuinely don't. The question isn't whether watchdog science journalism is doing something real — it is. It's whether anyone actually understands that this is the job it's doing now. Structurally fragile, never designed for this scale, and holding anyway.
