Qiao Dan Dan, from National University of Singapore, will present her research.

The Silicon Ghostwriter: How AI Usage Shapes Academic Influence

As large language model (LLM) writing tools diffuse across research practice, their relationship with scholarly outcomes remains empirically open. Using 24,454 arXiv papers published from 2021 to 2023 across 20 major categories, we provide observational evidence by linking detector-based AI-associated writing intensity to Google Scholar citation outcomes. Tobit models with field and publication year-month fixed effects show a positive but inverted-U-shaped relation between AI-associated writing intensity and annual citations. A two-channel mediation analysis reveals that AI shapes citations through textual conformity: AI elevates surface academic register, and it simultaneously compresses semantic distance from prior literature — and citation markets reward both. Heterogeneity analyses show stronger AI-citation associations for non-native English authors, developing-country authors, and corporate research labs. These findings characterize AI-associated writing as a socio-technical force shaping scholarly visibility and inequality.