Hallucinated AI Citations Are Polluting the Scientific Literature at Scale

Nature has reported that hallucinated citations generated by AI tools are increasingly appearing in published scientific literature, creating a growing pollution problem that threatens research integrity.

The Problem

AI language models sometimes generate plausible-looking but entirely fictional citations — complete with realistic author names, journal titles, and DOIs. These fabricated references are making their way into:

• Submitted manuscripts under peer review
• Published papers in some journals
• Literature reviews compiled by researchers using AI assistants
• Grant applications and thesis documents

Why It's Hard to Detect

• Hallucinated citations often reference real journals with realistic formatting
• Some cite real authors but with fabricated paper titles
• DOI numbers may not exist in databases but appear valid
• Overworked reviewers may not verify every reference

The Scale

The problem is growing as AI writing assistants become ubiquitous in academic work:

• Researchers use AI for literature reviews and background sections
• Time pressure leads to insufficient citation verification
• Some journals lack resources for thorough reference checking

Responses

• Journal policies: Some now require AI disclosure
• Automated tools: Citation verification software is being developed
• Community awareness: Calls for better training on AI limitations

Broader Context

This is part of a larger trend of AI-generated content degrading information quality:

• AI-generated research papers submitted to conferences
• Synthetic data contaminating training datasets
• AI-generated reviews on platforms like Amazon and PubMed

The citation pollution problem requires systemic solutions at the intersection of technology, publishing practices, and academic culture.