Scientifically Speaking: Fake scientific research on the rise, says study

A new study in Proceedings of the National Academy of Sciences maps the scale of scientific fraud and the networks behind fake research

At 11.47 pm, a scientist in California compares two cancer research papers. They come from different labs and claim to study different things, but one image in each is identical. She checks it against her database of suspicious cases and finds the same image—flipped and relabelled—in four more papers. All of them have been published in peer-reviewed scientific journals.

The authors estimate that the number of suspected paper-mill products is doubling every 18 months, compared with 15 years for the scientific enterprise as a whole. (Illustration: Ben Denzer)

A new study in Proceedings of the National Academy of Sciences, led by Reese Richardson at Northwestern University with Luís Amaral and Jennifer Byrne, maps the scale of what these fraud detectives are up against. The team analysed millions of journal records, retractions, and post-publication critiques to detect statistical “footprints” of scientific fraud. These telltale signs range from editors linked to unusually high numbers of retracted papers, to networks of authors who funnel work to those editors, to batches of papers with reused images.

The picture that emerges is not of lone cheats occasionally fooling the system. It is of organised networks—some tied to brokers or so-called paper mills—producing fake research in bulk and getting it into the same venues that publish legitimate science.

The authors estimate that the number of suspected paper-mill products is doubling every 18 months, compared with 15 years for the scientific enterprise as a whole. If those trends hold, they say, fraudulent research could soon outnumber all the papers that are retracted or publicly challenged combined.

That projection assumes the current rate continues (and some healthy skepticism is warranted), but the pace they document is alarming nonetheless.

These fake papers don’t just gather digital dust. They are indexed in major databases. They get cited. They can influence grant decisions, sneak into meta-analyses, or—most dangerously—shape guidelines that impact health or engineering standards.

Some of the machinery is depressingly efficient. In certain journals that list their handling editors, a tiny fraction of editors seems to be behind a large share of later-retracted papers. Some handle each other’s submissions in tight loops, turning peer review into a closed circuit. Fraudulent papers often arrive in batches—sometimes dozens at a time in the same issue.

The study documented one broker, the Academic Research and Development Association, that grew from listing 14 journals in 2018 to 86 journals by 2024. The study even noted that this broker “guarantees” publication in a stable of journals, swapping in new titles when old ones are blacklisted.

The incentives for publishing more—even at the cost of fraud—aren’t mysterious. Publishing equates to prestige, more grant points, and greater career advancement. The picture that is emerging is that, since the system rewards quantity, someone will supply it—for a fee.

The researchers also uncovered networks of 2,213 articles connected through shared fabricated images, yet only 34% of these obviously fraudulent papers have actually been retracted. One wonders how many papers out there are fraudulent that haven’t been detected yet!

A few caveats are in order. The authors are clear about the limits of their work. They’re looking at known cases of retractions, flagged papers, and curated watchlists, so they may be missing large swaths of fraud—or they may be overcounting papers that turn out to be genuine. Some fields and journals are over-represented simply because their data are more open. And statistical patterns, while revealing, don’t prove intent in every case.

That being said, the study offers a dismal glimpse into the future. Artificial intelligence (AI) is making it easier to generate convincing fake images from scratch and to rewrite text so it slips past plagiarism detectors.

The picture that emerges is one of science under siege. In short, the people making and selling fake science are faster and better organised than the systems designed to stop them. A handful of unpaid sleuths cannot protect the record on their own. In recent years, some scientific publishers have instituted formal research integrity and ethics offices, but fraudsters are quick to identify and infiltrate journals with weak processes. There’s hope in the fact that fraud leaves patterns—and patterns can be detected.

What is clear is that science, which for a long time relied on the honour system, must strengthen the guardrails, invest in detection, and change the incentives that make fake papers worth buying. Protecting the scientific literature matters far beyond academic incentives, because fake science has real-world consequences.

(Anirban Mahapatra is a scientist and author, most recently of the popular science book, When The Drugs Don’t Work: The Hidden Pandemic That Could End Medicine. The views expressed are personal.)