Abstract
As organizations increasingly integrate artificial intelligence (AI) into routine operations, its use and disclosure in crisis responses often elicit negative public reactions, particularly concerning organizational credibility and accountability. This study examines how AI-generated crisis responses shape stakeholder evaluations through an online experiment employing a 3 (crisis-type-matched response [CTMR]) × 2 (source attribution: AI vs. human) × 2 (crisis severity: high vs. low) factorial design with 391 participants. Results reveal significant main effects of source attribution and CTMR on perceived organizational credibility, reputation, attitude, and negative word-of-mouth (NWOM) intentions, while crisis severity produced no significant main or interaction effects. Compared with human-authored messages, AI-generated crisis responses reduced perceived credibility, reputation, and attitude and increased NWOM intentions. Path analyses further indicate that credibility serves as the primary mediating mechanism linking CTMR to public attitudes and NWOM, exerting a stronger influence than reputation. Together, these findings demonstrate that public attributions of responsibility shape evaluations of organizational credibility, which in turn drive negative behavioral intentions. Although source attribution did not interact with CTMR, its robust main effect suggests that AI use in crisis communication remains detrimental regardless of crisis type or severity. Theoretical and practical implications for crisis communication and AI governance are discussed.

This publication has 32 references indexed in Scilit: