Answers are grounded in this paper and cite their sources.
Purpose The rapid proliferation of AI-generated content (AIGC) has reshaped the digital content ecosystem, with short-form video platforms emerging as key dissemination arenas. While content transparency is often assumed to enhance trust and improve user responses, we argue that revealing AI involvement may reduce user engagement by undermining perceived authenticity – an essential value driver in social media. Design/methodology/approach We exploit a platform-wide policy shock, the implementation of China’s Administrative Measures for Labeling AI-Generated Content on September 1, 2025, which requires all AI-generated content to be explicitly labeled. Leveraging a dataset of 18,311 short videos from 216 creators on Douyin, we employ a difference-in-differences design comparing creators with high versus low pre-policy AI content propensity. Findings We find that mandatory AI disclosure significantly reduces user engagement in the short-term, leading to substantial declines in likes, favorites and overall interactions. Notably, this negative effect is pronounced for personal content but largely absent for commercial content. Additional analyses show that the policy does not significantly affect creators’ posting behavior, suggesting that the observed engagement changes are not accompanied by significant changes in creators’ posting frequency. Originality/value These findings contribute to the literature on AI disclosure and digital content by showing that mandatory labeling yields unintended negative consequences. More broadly, our results highlight that the consequences of AI disclosure depend critically on content context, selectively reducing authenticity-driven engagement in social media ecosystems.
No takes yet. Share an insight, caveat, or question.
Han et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: