Hallucinated citations are polluting the scientific literature. What can be done?

2026-04-01 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Research Methodology & Innovation, Artificial Intelligence & Machine Learning · Depth: Intermediate, medium

Summary

A growing problem of AI-hallucinated citations is polluting academic literature, with analyses showing a sharp increase in non-existent references in scholarly papers. Computer scientist Guillaume Cabanac discovered a fabricated citation of his work in the *International Dental Journal*, prompting concerns about AI-generated inaccuracies. Surveys indicate researchers increasingly use large language models (LLMs) for literature searches and bibliography formatting, leading to these fabricated references. One analysis of nearly 18,000 computer science papers found 2.6% in 2025 had at least one potentially hallucinated citation, up from 0.3% in 2024. *Nature*'s news team, in collaboration with Grounded AI, estimates tens of thousands of 2025 publications, including journal papers and books, likely contain invalid AI-generated references. Publishers like Frontiers and Taylor & Francis are developing or exploring AI tools to screen submissions, with some editors rejecting a significant percentage of papers due to fake references.

Key takeaway

For research integrity officers and journal editors, the surge in AI-hallucinated citations necessitates immediate action. You should implement robust screening processes, potentially integrating AI-powered tools like Grounded AI's Veracity, and conduct manual verification for flagged references. This proactive approach is crucial to maintain the credibility of published research and prevent a "flood of fake references" from undermining academic trust.

Key insights

AI-generated hallucinated citations are rapidly increasing, polluting academic literature with fabricated and inaccurate references.

Principles

• AI-generated citations often combine fragments of genuine publications.
• Automated detection tools for invalid citations are not error-free.

Method

Grounded AI's Veracity tool checks citations against scholarly databases, flagging invalid, irrelevant, or retracted work, assigning a risk score based on major issues and AI generation likelihood.

In practice

• Manually check suspicious citations flagged by software.
• Develop in-house AI tools for flagging integrity issues.
• Screen submissions for problematic references using specialized tools.

Topics

• Hallucinated Citations
• Large Language Models
• Research Integrity
• Academic Publishing
• Grounded AI

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Scientist, Research Scientist, Policy Maker

Related on AIssential

• HalluCiteChecker: A Lightweight Toolkit for Hallucinated Citation Detection and Verification in the Era of AI Scientists — HalluCiteChecker detects and verifies AI-generated hallucinated citations to improve scientific integrity and reduce reviewer burden.
• South Africa yanks AI policy after AI-assisted drafting invents citations — AI-generated content, especially citations, requires rigorous human verification to maintain integrity.
• Imbad0202 / academic-research-skills — ARS augments human researchers with AI tools to enhance academic writing quality and integrity, avoiding common AI pitfalls.
• Researchers who use hallucinated references to face arXiv ban — Preprint servers are implementing strict policies to combat unchecked generative AI use and maintain scientific integrity.
• Where Fake Citations Are Made: Tracing Field-Level Hallucination to Specific Neurons in LLMs — LLMs hallucinate citation fields independently, with author names being the most frequent error source.
• i tested basically every AI tool i could find for med school research. most are useless unless you already know enough to catch them lying. — AI tools for professional research require extensive manual verification due to persistent accuracy and citation issues.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.