Abstract:  Online technological controversies often involve both ordinary disagreement and counterspeech, yet these forms of opposition are rarely compared within the same discussion environment. Drawing on argumentation theories, this study compares disagreement replies to non-hateful comments with counterspeech replies to hateful comments in Douyin discussions of China’s AI controversies, where hate speech may include stigmatizing generalizations about social groups. Using 9,407 oppositional parent–response dyads, we compared like counts, toxicity, subsequent toxic replies, and the distribution and reception of six response types. Counterspeech and disagreement did not differ significantly in like counts, but counterspeech exhibited higher toxicity and was related to greater odds of toxicity among direct replies posted within seven days. Despite sharing the same six response types, the two response contexts varied in their type distributions and associations with likes. These findings situate counterspeech within a broader spectrum of online opposition, revealing contextual differences in its argumentative expression and subsequent interaction across hateful and non-hateful contexts.

Keywords:  artificial intelligence; China; counterspeech; disagreement; social media; stigmatization; toxicity

Published:   12 November 2026

Issue:   Vol 14 (2026): The Role of AI for Counter Speech: Detection, Intervention, and Risks (In Progress)