![[2602.22221] Misinformation Exposure in the Chinese Web: A Cross-System Evaluation of Search Engines, LLMs, and AI Overviews](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.22221] Misinformation Exposure in the Chinese Web: A Cross-System Evaluation of Search Engines, LLMs, and AI Overviews
Summary
This article evaluates misinformation exposure on the Chinese web by comparing traditional search engines, LLMs, and AI-generated overviews using a new dataset of real user queries.
Why It Matters
Understanding how misinformation spreads in non-English web ecosystems is crucial for developing reliable AI tools. This study highlights the structural risks associated with AI-mediated search, emphasizing the need for improved accuracy and transparency in information access.
Key Takeaways
- The study introduces a fact-checking dataset of 12,161 Chinese Yes/No questions.
- Significant differences in factual accuracy were found across search engines, LLMs, and AI overviews.
- The research estimates the potential misinformation exposure of Chinese users across different regions.
- AI-mediated search poses structural risks that need addressing for better information reliability.
- The findings call for more transparent and reliable information-access tools.
Computer Science > Information Retrieval arXiv:2602.22221 (cs) [Submitted on 15 Dec 2025] Title:Misinformation Exposure in the Chinese Web: A Cross-System Evaluation of Search Engines, LLMs, and AI Overviews Authors:Geng Liu, Junjie Mu, Li Feng, Mengxiao Zhu, Francesco Pierri View a PDF of the paper titled Misinformation Exposure in the Chinese Web: A Cross-System Evaluation of Search Engines, LLMs, and AI Overviews, by Geng Liu and 4 other authors View PDF HTML (experimental) Abstract:Large Language Models (LLMs) are increasingly integrated into search services, providing direct answers that can reduce users' reliance on traditional result pages. Yet their factual reliability in non-English web ecosystems remains poorly understood, particularly when answering real user queries. We introduce a fact-checking dataset of 12~161 Chinese Yes/No questions derived from real-world online search logs and develop a unified evaluation pipeline to compare three information-access paradigms: traditional search engines, standalone LLMs, and AI-generated overview modules. Our analysis reveals substantial differences in factual accuracy and topic-level variability across systems. By combining this performance with real-world Baidu Index statistics, we further estimate potential exposure to incorrect factual information of Chinese users across regions. These findings highlight structural risks in AI-mediated search and underscore the need for more reliable and transparent information-acces...