Recognising and mitigating LLM Pollution in online behavioural research.

Rilla, Raluca; Werner, Tobias; Yakura, Hiromu; Rahwan, Iyad; Nussberger, Anne-Marie · Nat Commun · 2026

Where this comes from

Abstract

Online behavioural research faces a growing methodological and epistemic threat as participants increasingly rely on large language models: <i>LLM Pollution</i>. Amid accumulating empirical evidence of contamination, we introduce a conceptual framework that distinguishes three variants — <i>Partial LLM Mediation</i>, <i>Full LLM Delegation</i>, and <i>LLM Spillover</i>. Their interaction distorts samples, biases inferences, and fuels an escalating methodological arms race. We outline mitigation strategies spanning researcher practices, platform accountability, and community adaptation.