When Cow Urine Cures Constipation on YouTube: Limits of LLMs in Detecting Culture-specific Health Misinformation

2026-04-23

AI Summary

This study addresses the systematic limitations of large language models (LLMs) in identifying health misinformation rooted in non-Western cultural contexts—particularly when such content intertwines religious narratives, gendered rhetoric, and pseudoscientific claims, as exemplified by YouTube videos from India promoting cow urine as a therapeutic agent. Through LLM-assisted discourse analysis of 30 multilingual video transcripts, the research evaluates the performance of GPT-4o, Gemini 2.5 Pro, and DeepSeek-V3.1 under varied prompting strategies. Findings reveal that current LLMs, biased toward Western-centric training data, struggle to parse the distinctive rhetorical structures of culturally embedded misinformation. The study demonstrates that prompt engineering alone cannot overcome this deficit and argues that LLMs must develop intrinsic cultural competence rather than relying on post-hoc remediation.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Application Domains: Misinformation & Fake News

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Social media platforms have become primary channels for health information in the Global South. Using gomutra (cow urine) discourse on YouTube in India as a case study, we present a post-facto Large Language Model (LLM)-assisted discourse analysis of 30 multilingual transcripts showing that promotional content blends sacred traditional language with pseudo-scientific claims in ways that sophisticated debunking content itself mirrors, creating a rhetorical register that LLMs, trained predominantly on Western corpora, are systematically ill-equipped to analyse. Varying prompt tone across three LLMs (GPT-4o, Gemini 2.5 Pro, DeepSeek-V3.1), we find that culturally embedded health misinformation does not look like ordinary misinformation, and this cultural obfuscation extends to gendered rhetoric and prompt design, compounding analytical unreliability. Our findings argue that cultural competency in LLM-assisted discourse analysis cannot be retrofitted through prompt engineering alone.

Problem

• health misinformation
• cultural competency
• large language models
• discourse analysis
• Global South

Innovation

• culture-specific misinformation
• LLM limitations
• discourse analysis
• prompt engineering
• health misinformation