Bin Wang, Yingying Zhang, Qianru Li, Bingbing Li, Juan Wang
Narrative integration, transforming fragmented cognitive content into coherent form, is theorised as a core cognitive process in emotion regulation, yet its active components have rarely been isolated experimentally. The present study used a between-subjects experiment (N = 87) to disentangle two cognitive components hypothesised to underlie narrative-based emotion regulation: self-referential processing (personalised content) and cognitive integration (narrative coherence). AI text generation technology was employed to manipulate these components across three experimental conditions: a Complete Condition containing both components, an Integration-Only Condition (narrative structure without personalization), and a Self-Referential-Only Condition (personalised content without narrative structure). Participants completed pre- and post-exposure assessments of state anxiety and affect. Significant Time × Group interactions emerged for both state anxiety and positive affect. Only the Complete Condition reduced state anxiety and enhanced positive affect; neither single-component condition produced beneficial changes, and the Self-Referential-Only Condition decreased positive affect. These findings provide initial experimental evidence that self-referential processing and cognitive integration can operate synergistically in narrative-based emotion regulation, suggesting that neither component alone may be sufficient. (Trial registration: ChiCTR2500114344).