When Artificial Intelligence Fabricates Sources: Why Training and Critical Thinking Are Indispensable

Artificial intelligence (AI) is rapidly reshaping how we conduct research, write, and share knowledge. Generative AI tools, such as large language models, can support researchers in drafting manuscripts, analyzing large datasets, and even suggesting ideas for future studies. Yet alongside these extraordinary opportunities comes a serious challenge: the risk of fabricated or inaccurate information.

A major concern is hallucination, where AI produces false yet plausible content, such as nonexistent bibliographic references. Recent studies indicate that fabricated citations and errors are not just isolated glitches but inherent limitations of current AI systems. This issue can undermine the reliability of scientific literature and damage trust in research.

AI: a potent yet imperfect instrument

Generative AI has unlocked remarkable opportunities for research, writing, and sharing knowledge. Whether drafting manuscripts or analysing vast datasets, its capabilities are clear. However, the evidence indicates that AI should not be regarded merely as an information-processing tool—it is mainly a language-processing system. This is why it can confidently generate seemingly credible yet incorrect citations or even invented results.

Training, Awareness, and Critical Thinking

How can we tackle this issue? The solution is rooted in education and cultural change. Teaching about how AI functions, how to interpret its results, and how to verify information should be a fundamental part of professional growth for researchers, clinicians, journalists, policymakers, and the public.However, technical training alone is insufficient. Equally important is fostering critical thinking—the capacity to:

• question the accuracy of AI outputs;
• carefully check sources;
• recognise biases (from both AI and ourselves);
• and understand the epistemological limits of automation.

In essence, AI should act as a co-pilot and enhancer, not a replacement for judgment and critical analysis.

AI: an opportunity for growth rather than mere delegation

Addressing fabricated citations (be they in books, articles, or AI-generated drafts) is more than just a technical error. As we incorporate AI into our daily workflows, it’s essential to do so carefully and responsibly. 
Responsible AI use involves, among others, developing our skills, enhancing critical thinking, and ensuring that innovation does not undermine scientific integrity. Only through this approach can we fully benefit from AI’s potential without sacrificing the quality and trustworthiness of scientific knowledge.