Generating Credibility in Crisis: Will an AI-Scripted Response Be Accepted?

Increasingly, practitioners are using artificial intelligence (AI) to strategically monitor and respond to crises. However, there is little evidence indicating whether a crisis response, disclosed as AI-scripted, will be accepted by stakeholders and what effects the disclosure may have on message credibility, attribution of responsibility, message acceptance, and organizational reputation. Using Situational Crisis Communication Theory (SCCT), this 2 (type of crisis) x 2 (presence or absence of AI label) online experiment explored how the disclosure of AI-generated content affects post-crisis organizational outcomes. Participants (n = 238) were randomly assigned to a vignette featuring a victim (e.g. shooting) or accidental (e.g. data breach) cluster crisis and a detailed response that either disclosed or did not disclose that the content was AI-generated. Results revealed no effect of disclosure on message credibility or attribution of responsibility. However, message acceptance served as a mediator between message credibility and attribution of responsibility on organizational reputation. This study contributes to SCCT and provides contextual evidence for practitioners who are considering AI for crisis responses. Ethical implications and future directions are discussed.