Fact-Checking from Human to AI: Transformations in Accuracy, Trust, and Agenda-Setting Effects

This theoretical essay examines the shift from human fact-checking to AI fact-checking from the perspective of independent fact-checking organizations. While AI fact-checking offers clear advantages in speed, scalability, and operational efficiency, it also introduces new concerns regarding epistemic dependence, transparency, trust, and agenda-setting effects. This paper reviews the key limitations of human fact-checking and explains how AI appears to address them. It further proposes six propositions linking perceived accuracy, transparency, perceived selection bias, trust, and agenda-setting effects in AI fact-checking. It concludes by presenting a conceptual model that summarizes these propositions and discussing hybrid and multi-model approaches as possible responses.