When AI ads violate expectations: the interplay of AI character, product type, and advertising appeal of AI ads on consumer attitudes

Generative AI is rapidly reshaping video advertising production. Using expectancy violations theory as an interpretive lens, this study develops a framework in which ad appeal (rational vs. emotional), product type (high-tech vs. low-tech), and AI character embodiment (no character vs. animal-like vs. human-like) jointly relate to ad attitude and to two appraisals: perceived novelty and perceived uncanniness. This study tests the framework in five between-subjects experiments with consumers recruited from an online panel. Results show that rational appeals outperform emotional appeals in AI-generated ads, especially for low-tech products. Novelty is positively and uncanniness negatively associated with ad attitude, and the indirect effects through both appraisals are consistent with the proposed process. Human-like AI characters amplify uncanniness in emotional low-tech executions. In one rational low-tech execution, an animal-like character lowered uncanniness relative to a no-character baseline, but ad attitude did not differ significantly. The findings extend expectancy-violation theorizing in AI advertising and offer configuration-level design guidance for brands deploying disclosed AI-generated advertising.