Can AI-assisted apologies foster trust and forgiveness? The role of warmth and competence in crisis communication tone

This study examines how perceived authorship and relational tone jointly influence audience responses to AI-assisted crisis apologies. Using a 4 (perceived authorship: AI, human, mixed, control) × 2 (relational tone: competence, warmth) factorial design, we investigate whether warmth-focused language can moderate the relationship between AI authorship and perceived sincerity, which in turn influences audience responses to crisis apologies. Results show that apologies perceived as human-authored are evaluated as more sincere, leading to higher trust and forgiveness intentions. In contrast, apologies perceived as AI-authored increase machine heuristic perceptions, which lower perceived sincerity, and subsequently, trust and forgiveness intentions. Warmth moderates the relationship between perceived authorship and sincerity for apologies perceived as human-authored but fails to attenuate the negative indirect effects of machine heuristic processing for apologies perceived as AI-authored. We ground this asymmetric pattern in need for cognitive closure theory, arguing that an unambiguous AI authorship cue triggers cognitive closure on the communicator’s nature, rendering subsequent warmth-focused cues incongruent with the already-formed judgment and therefore discounted. The SCM provides a complementary account of why the AI stereotype resists message-level updating while the human category remains open to elaboration. These findings advance theoretical understanding of divergent audience processing in AI-mediated crisis communication and carry practical implications for how organizations deploy AI in high-stakes apology contexts where authenticity is closely scrutinized.