AI-Generated Fake Citations: A Growing Risk to Scientific Integrity
Introduction
AI-generated fake citations are becoming a growing threat to scientific integrity. When generative AI tools produce plausible but non-existent references, they can introduce errors into scholarly literature and weaken trust in scientific publishing.
A recent study published in The Lancet reveals a rapid increase in fabricated references appearing in peer-reviewed scientific publications. Once considered a marginal issue, this phenomenon is gaining momentum with the widespread adoption of generative artificial intelligence tools.
The findings suggest that traditional quality-control mechanisms are no longer always sufficient to detect these errors before publication.
What Are Fake Scientific Citations?
A fake citation is a bibliographic reference that points to a study that does not exist.
Such references may be introduced deliberately as part of fraudulent research practices, including the activities of paper mills, or unintentionally through the use of tools capable of automatically generating scientific content.
The issue becomes particularly concerning when a fabricated reference appears credible, featuring a realistic title, plausible authors, a recognized journal, and a consistent publication date.
At first glance, there is often no obvious way to distinguish a genuine reference from an invented one.
A Study of 2.5 Million Scientific Papers
To assess the scale of the problem, researchers analyzed:
- 2.5 million biomedical papers;
- More than 97 million verified references;
- Publications released between 2023 and 2026.
The results reveal a rapid increase in fabricated references across the scientific literature.
Key Findings
- More than 4,000 fabricated references were identified.
- Over 2,800 scientific papers contained at least one fake citation.
- The number of fabricated references increased more than twelvefold between 2023 and 2025.
- By early 2026, approximately one paper in every 277 contained at least one non-existent reference.
These figures suggest that fabricated citations are no longer isolated incidents but a measurable trend affecting scientific publishing.
Why Are Fake Citations a Major Concern?
Bibliographic references form the invisible infrastructure of scientific research.
When a citation points to a non-existent study, it becomes impossible to verify the information, reproduce the findings, or assess the strength of the evidence being presented.
The consequences affect multiple stakeholders:
- Researchers who rely on inaccurate literature;
- Clinicians who use evidence that cannot be verified;
- Policymakers whose decisions depend on reliable scientific evidence;
- Publishers whose credibility is directly at stake.
The study also found that 98% of papers containing fabricated references had neither been corrected nor retracted at the time of analysis.
This finding highlights the limitations of current reference-checking and quality-assurance processes.
The Growing Role of Artificial Intelligence
Large Language Models (LLMs) are capable of generating highly convincing bibliographic references.
When a model cannot identify a real source that matches a user’s request, it may generate a plausible but entirely fictitious citation. This phenomenon is commonly referred to as an “AI hallucination.”
As AI adoption expands across scientific writing workflows, this risk is becoming a practical concern for researchers, reviewers, and publishers alike.
The challenge lies in the fact that these fabricated references often look entirely legitimate and can easily escape detection during a routine manual review.
Towards Systematic Reference Verification
The rise of fabricated citations underscores the need for stronger bibliographic quality-control mechanisms.
Verifying the existence and accuracy of references can no longer be viewed as a purely administrative task. It is becoming a critical component of quality assurance in scientific publishing.
As AI-powered writing tools become increasingly common, the ability to quickly detect non-existent references will play a key role in preserving trust, transparency, and reproducibility across the research ecosystem.
Source study: “Fabricated citations: an audit across 2.5 million biomedical papers” published in The Lancet.
> Read more articles on the LibScience blog