When AI Moves to the Forefront of Creativity: How Transparent Disclosure Shapes Consumer Acceptance Through Legitimacy Judgments

Generative artificial intelligence is increasingly used in brand content production, yet disclosing AI involvement can reduce consumer evaluations of creative work. Existing explanations primarily emphasize ability and reliability judgments, offering only a partial account of negative responses to AI‐related creator cues. Drawing on institutional legitimacy theory, this study develops a context‐specific AI disclosure evaluation framework integrating prior work on legitimacy, the associative–propositional evaluation model, and schema incongruity. The framework proposes that AI disclosure can reduce acceptance by lowering perceived legitimacy, elicit early differential attentional processing, and shape later evaluative integration depending on the work's emotional involvement. Study 1 uses a scenario experiment ( N = 300) and finds that an AI‐generated creative arrangement reduces acceptance through lower perceived legitimacy, alongside a countervailing positive direct effect. Study 2 uses an event‐related potential (ERP) experiment ( N = 27). ERPs are voltage changes time‐locked to stimulus presentation. The AI condition elicits a more negative N1, an early negative‐going component associated with differential attention allocation across work types. Larger P3/P300, a later positive component associated with resource allocation and evaluative updating, and lower explicit acceptance emerge only for highly emotionally involved work. Study 3 analyzes 11,572 Weibo texts and a regression subset of 9175 observations, identifying an analogous association between high emotional involvement and lower levels of acceptance in public discourse. Collectively, the studies shift attention from AI ability to the perceived legitimacy of disclosed creative arrangements and distinguish early cue processing from later context‐dependent evaluation. Perceived legitimacy is the behavioral mechanism tested in Study 1, whereas ERP measures characterize temporal processing differences. The findings suggest that managers should evaluate disclosed AI involvement in relation to the emotional significance of the content and clearly communicate human responsibility; the effectiveness of specific disclosure wording remains untested. Because experimental evidence comes primarily from younger Chinese participants and archival evidence from one Chinese social media platform, generalizability requires further testing across age groups, cultures, creative domains, and platforms.