A Growing Threat to Public Trust and Economic Progress

Scientific research, long considered the bedrock of innovation and public policy, is facing an unprecedented crisis. Mass-produced, fraudulent, and AI-generated research papers are flooding academic journals, eroding trust in science and posing serious risks to public health and economic development.

A tragic case illustrates the human cost: a woman suffering from cancer placed her trust in an experimental treatment called GcMAF, which she discovered through seemingly credible online studies. The research backing the drug was fabricated. She died, a victim of fake science.

According to Bernhard Sabel, Professor of Medical Psychology at Otto-von-Guericke University in Magdeburg, and Dan Larhammar, Professor of Molecular Cell Biology at Uppsala University, this is not an isolated incident. In a recent article in “Royal Society Open Science”, they describe the situation as “the biggest scientific crisis ever.”

The Scale of the Problem

The proliferation of fake research is no longer a niche issue—it is systemic. Estimates suggest that between 5.8% and 15.3% of biomedical papers may be fraudulent or suspicious. That translates to roughly 100,000 fake biomedical articles published annually. Globally, the economic toll is staggering: mis-investments stemming from fake research may reach €4 billion per year, with broader R&D losses potentially hitting €145 billion.

The problem is exacerbated by the rise of "paper mills"—commercial operations that produce and sell fake research—and predatory journals that publish low-quality or fabricated studies for a fee. These entities exploit the pressure researchers face in a “publish-or-perish” culture, where publication counts often determine career advancement and funding.

AI: A Double-Edged Sword

While artificial intelligence has the potential to revolutionise research, it also introduces new risks. Recent studies have shown that AI models like ChatGPT can generate highly convincing but entirely fabricated scientific articles in under an hour. These papers often include plausible data, citations, and formatting, making them difficult to distinguish from legitimate research.

In one experiment, researchers used ChatGPT to create a fraudulent neurosurgery article. The result was a 1,992-word paper with 17 citations that fooled even some expert reviewers. Although the references contained errors, the overall structure and language were convincingly scientific. Detection tools failed to flag the article as AI-generated, highlighting the limitations of current safeguards.

The Stockholm Declaration: A Call for Reform

In response to this crisis, Sabel and Larhammar convened a conference at the Royal Swedish Academy of Sciences in June 2025. The outcome was the “Stockholm Declaration”, a comprehensive framework for reforming scientific publishing.

Key Proposals Include:

·        Shifting from quantity to quality in research evaluation, allowing scholars to submit a limited number of their best works for assessment.

·        Promoting non-commercial, community-owned publishing models, including “diamond” open access (free to publish, free to read).

·        Establishing independent bodies to detect and penalise fraud, including legal frameworks to sanction authors, publishers, and paper mills.

·        Creating public integrity scoreboards for journals and publishers to increase transparency and accountability.

A Broken Publishing Model

The current academic publishing system is deeply flawed. Researchers conduct studies, write articles, and peer-review manuscripts—often without compensation. Yet, commercial publishers charge exorbitant fees, sometimes exceeding €10,000 per article, while retaining copyright and reaping massive profits.

As Larhammar notes, “If someone had described this system in a satirical magazine, it would have been laughed at as absurd.”

Even well-intentioned policies like open access have been co-opted. While designed to democratise knowledge, open access has often become a revenue stream for publishers, who charge authors to make their work publicly available. This has led to a surge in low-quality journals that publish anything for a fee.

The Role of AI and Misattribution

AI-generated content is not only being used to create fake articles—it’s also being falsely attributed to real researchers. In one case, a paper on business strategy was published under the name of a prominent academic without his knowledge. The article was clearly AI-generated, yet it appeared in a journal with a valid DOI and was indexed by Google Scholar.

Such incidents damage reputations, mislead hiring committees, and pollute the scientific record. If left unchecked, AI-generated content could be used to train future AI models, leading to what researchers call “model collapse”—a degradation in AI performance due to recursive training on low-quality data.

Probing the future

Despite the gravity of the situation, Sabel and Larhammar emphasise that science itself is not the enemy. “We must not distrust all science,” they write. “But we must openly talk about the problems in order to find solutions.”

The Stockholm Declaration offers a roadmap. It calls for coordinated international action, legal reforms, and a cultural shift in how research is evaluated and disseminated. The European Research Council and other funding bodies are already moving toward quality-based assessment models.

However, implementing these changes will require global cooperation. Without it, fraudsters can simply relocate to jurisdictions with lax oversight, continuing to exploit the system.

To conclude, the integrity of science is under threat. Fake research, AI-generated articles, and predatory publishing are not just academic concerns—they undermine public health, distort policy decisions, and waste billions in research funding. The Stockholm Declaration provides a clear, actionable response. What remains to be seen is whether the global scientific community—and the governments and institutions that support it—will act decisively before the damage becomes irreversible.

By Ganye Kwah Driscole (PhD)