Citation Hallucination Rates in LLM-Generated Research: A 2025 Benchmark Across 10 Models

July 2, 2026Open Access

View Full Paper

Ask AI

Bookmark

Share

Key Points

This research aims to evaluate the rates of citation hallucination in research generated by large language models (LLMs).
Conducted a benchmark evaluation across 10 different LLMs for citation accuracy.
Analyzed the instances of citation hallucinations identified in the generated texts.
Utilized various metrics to quantify and compare the citation accuracy among models.
The rate of citation hallucinations varied significantly across the evaluated models.
Model A showed a hallucination rate of 12% (95% CI: 10%-14%, p<0.01), while Model B had a rate of 5% (95% CI: 4%-6%, p<0.01).
Overall, 7 out of 10 models exceeded a hallucination rate of 10%, raising concerns about their reliability.

Evaluates citation hallucination rates in LLM-generated research, highlighting accuracy concerns.

Research article: Citation Hallucination Rates in LLM-Generated Research: A 2025 Benchmark Across 10 Models

No takes yet. Share an insight, caveat, or question.

Ивченко et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

Share