Is the paper you're reading even real?

I read a lot of papers. Lately I catch myself doing something I never used to do. Before I trust a result, I check whether the paper is real: is the journal a real journal, are the authors real people, does the data look like data or like something a machine dreamed up on request. That check used to feel paranoid. Now it feels like basic hygiene, and I don’t like that it does.

That unease has a name and a number now. In November 2025, Bernhard Sabel and Dan Larhammar published a Perspective in Royal Society Open Science called “Reformation of science publishing: the Stockholm Declaration”. Larhammar is a past president of the Royal Swedish Academy of Sciences, so this is not a blog rant, it is a formal shot across the bow. Their message, stated flatly: fake and low-quality science is now being produced at industrial scale, and they call it “arguably the largest science crisis of all time, threatening to erode people’s trust in research.”

What the paper claims, in plain numbers

The authors name three threats to the literature. First, for-profit publishers, who charge steep fees for open access and take, by the authors’ own figure, between 25% and 35% profit from academia. Second, predatory journals, which publish a flood of low-quality papers with no real peer review. Third, and newest, paper mills.

The headline number is the number of fake papers: “hundreds of thousands per year,” which the paper attributes to two earlier studies. If that estimate is anywhere near right, the counterfeit is no longer a rounding error in the scientific record. It is a meaningful fraction of it.

What a paper mill actually is

A paper mill is a business. It manufactures articles with fabricated data, increasingly using AI-generated text, tables, and images, and then, in the paper’s own words, it “sells authorships to scientists under pressure to publish and bribes editors to publish these papers.”

The buyer is not a cartoon villain. It is usually a researcher who needs one more line on a CV to keep a job, land a grant, or finish a degree, working inside a “publish or perish” culture that counts papers instead of reading them. The paper’s point: the incentive comes first, the fraud follows. The metrics we chose, publication count, impact factor, h-index, get gamed, because any scoreboard eventually does.

What the Stockholm Declaration wants

The declaration was drafted at a conference at the Royal Swedish Academy of Sciences on 9-10 June 2025, and signed by more than two dozen researchers, including the image-integrity sleuth Elisabeth Bik and a board member of PubPeer. It asks for four things:

• Academia takes publishing back. Move to non-profit models (diamond open access, authors keep copyright) instead of paying for-profit publishers.
• Reward quality, not quantity. Stop scoring people by how many papers they produce, in hiring, tenure, and funding.
• Independent fraud detection. Detection run by “researcher-controlled” bodies that are not funded by publishers, able to “fraud tag” fake articles and the journals that print them.
• Legislation. Legal definitions and penalties for industry-scale fraud, so that paying a paper mill carries an actual cost.

My take: the diagnosis is sharper than the cure

I believe the diagnosis. The cure I’m less sure about.

The scary number does a lot of work, and it arrives as an estimate carried across a citation, not something I can audit from the page. I would have liked a table. The four recommendations, meanwhile, are the ones the reform movement has been making for years. The paper knows this: it prints its own table of predecessors, Leiden 2015, DORA 2012, Plan S 2018, and more. Good ideas usually don’t fail for lack of being restated one more time. There is even a reply already in the same journal, titled “Why the Stockholm Declaration will never work”, which argues the plan rests on assumptions that are too optimistic. Read it next to the declaration, not instead of it.

Two smaller notes, for honesty. The lead author founds and holds a stake in a science-integrity foundation, disclosed in the paper. That doesn’t make him wrong, it is context. And the paper carries a line I keep re-reading: “We have not used AI-assisted technologies in creating this article.” You have to declare that now. That’s the world we write in.

Why this is your problem, even if your work is clean

If you’re a researcher trying to be found and cited, and your work is honest, you might file this under someone else’s problem. It isn’t.

Fakes don’t only add noise, they tax trust, and the tax is flat. When a reader, a reviewer, or an AI answer engine can’t cheaply tell your real result from a manufactured one, the cheap move is to discount all of it. Skepticism gets applied to the whole field, and the honest pay it right alongside the frauds. The entire reason to put real work where people can find it is that being found is supposed to mean being believed. Erode the trust base and visibility stops paying: you can be the most discoverable paper in your subfield and still lose, if the currency of the field turns out to be counterfeit.

So maybe the check I started doing, is this paper real, isn’t paranoia anymore. Maybe it is just what reading looks like now. I would rather it went back to being paranoid. Do you think it’ll?