When AI admits its flaws: authenticity and novelty pathways of flaw disclosure in AI-generated advertising

Abstract

As generative AI becomes embedded in advertising production and mandatory AI disclosure requirements expand globally, brands face a structural challenge: disclosure of AI authorship can erode perceived authenticity. This study examined whether voluntary flaw disclosure counteracts this deficit in AI-generated advertising, and whether its effectiveness is conditioned by consumers' perceived risk. A between-subjects experiment ( N = 474) crossed three flaw disclosure conditions (direct/indirect/none; all carrying an identical platform-standard AI disclosure label) with message appeal (factual/humor) and tested a dual-pathway mediation model focused on the disclosure factor. Direct disclosure most strongly elevated perceived brand authenticity, whereas indirect disclosure most strongly enhanced perceived advertising novelty (η 2 = 0.13 and 0.21). Both pathways significantly and independently mediated the effect of flaw disclosure on attitude toward the ad (authenticity: IE = 0.46, 95% CI [0.31, 0.64]; novelty: IE = 0.40, 95% CI [0.24, 0.56]), and carried serially through brand attitude to purchase intention. A negative residual direct effect indicated a competitive suppression structure: the benefit of admitting a flaw was entirely mediation-dependent. The hypothesized moderating role of perceived risk on the authenticity pathway was not supported (index of moderated mediation 95% CI [−0.09, 0.23]); the dual-pathway mechanism appeared relatively stable across the observed range of perceived risk, although broader contexts require examination. The findings extend the minor flaw effect to AI advertising disclosure and identify authenticity and novelty as parallel, complementary routes for restoring consumer response.