Building the oracle: power, culture, and the organizational domestication of In-house AI in Chinese newsrooms

While Chinese media outlets increasingly deploy in-house large language models (LLMs) to execute a range of tasks in the newsroom, the practical outcomes vary drastically. To unpack these disparities, this study challenges the monolithic view of technology adoption by proposing an Intraorganizational Domestication Dynamics Framework and conceptualizing the newsroom as a contested domestication field. Through a comparative case study of two major media outlets, we reveal fundamentally divergent trajectories. On the one hand, an application-driven approach in the first organization fostered voluntary appropriation, embedding the LLM as an empowering “efficiency assistant.” On the other hand, a top-down mandate in the other organization provoked forced appropriation, causing frontline staff to objectify this form of artificial intelligence as a “grave digger,” which threatens their expertise and triggers multilayered de-domestication resistance. This study advances domestication theory by demonstrating how complex internal sociotechnical struggles and power dynamics determine a technology’s fate.