AI-Assisted Crisis Communication: Effects of Crisis Responsibility and AI Usage Transparency on Credibility and Organizational Reputation

Sammanfattning

As organizations increasingly rely on AI in their communication workflows, questions arise about how AI usage shapes stakeholder trust in high-stakes contexts like crisis communication. This study integrates two previously disconnected frameworks: Situational Crisis Communication Theory (SCCT), explaining how crisis responsibility drives reputational threat, and AI-mediated communication (AI-MC) research, examining how algorithmic authorship influences perceptions of trustworthiness and authenticity. Yet no prior experiment has simultaneously contrasted low and high crisis responsibility, examined AI-assisted rather than fully AI-generated authorship, and incorporated AI acceptance as a theoretically relevant factor. The study investigates how crisis responsibility and AI-assisted authorship disclosure jointly influence stakeholder evaluations of message credibility, source credibility, and organizational reputation. A 2x2 between-subjects online experiment (N = 266) exposed participants to a fictional environmental crisis response varying in crisis type and authorship disclosure. Regression analyses revealed that AI-assisted labeling significantly reduced all three outcome variables, with the strongest effect on message credibility. Crisis responsibility significantly predicted only source credibility, and no interaction effect was found, indicating that AI disclosure exerted a consistent negative influence regardless of crisis type. General AI acceptance positively predicted all outcomes but did not moderate the labeling effect. These findings reveal a transparency paradox: disclosing AI assistance, though ethically warranted, reduces the very credibility organizations seek to maintain in a crisis, with implications for how AI-assisted crisis communication should be implemented and disclosed.