Exemption or backfire? The double-edged effect of AI attribution strategy on brand trust in digital content moral crises

Human-AI collaboration has become central to brand content production, but AI's black-box nature obscures responsibility in digital content moral crises. When such content is perceived as violating widely shared social values, brands increasingly attribute such incidents to AI errors. Prior research has mainly emphasized the responsibility-exemption role of AI attribution strategy, offering limited insight into its moral risks, mixed consequences, and underlying mechanisms. Drawing on the fact-value model, this study conducts five scenario-based experiments to examine how AI attribution strategy affects brand trust in this crisis context. The findings reveal a double-edged effect. At the factual level, this strategy increases consumers' perceived responsibility reduction, thereby producing an exemption effect that supports trust repair. At the value level, it also triggers perceived moral indifference, thereby producing a backfire effect that damages trust. This double-edged effect is moderated by public opinion and consumer traits. High negative electronic word-of-mouth (eWOM) dispersion strengthens the perceived responsibility reduction path, whereas low negative eWOM dispersion strengthens the perceived moral indifference path. AI literacy strengthens the exemption effect by helping consumers understand AI-related technical limitations, whereas hostile attribution bias strengthens the backfire effect by leading consumers to question the moral motives behind the brand's technical explanation. This study advances crisis attribution research by examining AI attribution as a proactive brand communication strategy under attribution ambiguity and shows how it can reduce factual blame while raising value-based moral concerns. The findings offer guidance for using AI attribution carefully and reducing ethical communication risks.