Ecclesial Discernment in the Era of Generative Artificial Intelligence

This article offers an ecclesiological analysis of discernment under conditions shaped by generative artificial intelligence (AI). It examines how generative AI may reshape synodal listening by producing statistically central outputs that can be mistaken for ecclesial reception. After clarifying the theological role of the sensus fidei fidelium in discernment, the article analyses how large language models generate responses through probabilistic prediction, tending towards convergence and centrally weighted answers. Three tensions structure the argument: who the subject of discernment is, what kind of consensus is at stake, and how centre-periphery dynamics are affected by statistical centrality. An exploratory study reports three empirical patterns: output instability across prompts and settings, shifting registers of epistemic authority through generated voice, and the quantitative suppression of minority signals under aggregation and summarisation. The article concludes with theological-pastoral criteria that distinguish conditionally acceptable uses from imprudent ones, emphasising agency, traceability, and attention to the margins.