The rapid advancement of generative AI technology is fundamentally transforming advertising production methods. However, existing research has predominantly focused on virtual influencers or static images, relying solely on survey-based purchase inten...
The rapid advancement of generative AI technology is fundamentally transforming advertising production methods. However, existing research has predominantly focused on virtual influencers or static images, relying solely on survey-based purchase intention measures while overlooking the gap between stated intentions and actual behavior. Empirical studies examining consumer responses to AI-generated content and the effects of AI disclosure in cosmetic video advertising remain particularly scarce. This study aimed to empirically investigate the effects of generative AI-based cosmetic video advertising on consumer responses and the impact of AI disclosure.
A mixed-methods approach was employed across three phases. Phase 1 involved developing measurement items through focus group interviews with 9 participants. Phase 2 consisted of an online survey with 490 female respondents aged 19-49, randomly assigned to either an AI non-disclosure group (n=248) or an AI disclosure group (n=242). Phase 3 involved actual advertising campaigns on Meta (Instagram, Facebook), YouTube, and TikTok platforms, with subsequent collection of behavioral data.
The key findings are as follows. First, AI-generated videos received significantly higher ratings than filmed videos across all consumer response measures, including cognitive response, affective response, attitude toward the advertisement, and purchase intention. Second, sequential mediation effects were confirmed, wherein cognitive and affective responses influenced purchase intention through attitude toward the advertisement, with affective response demonstrating full mediation. Third, while in-video AI disclosure had no significant effect on initial evaluations, post-viewing disclosure triggered a backlash effect that significantly decreased all response indicators. Fourth, the negative impact of post-viewing disclosure was significantly mitigated by consumers' positive attitudes toward AI technology. Fifth, actual platform data revealed that AI-generated videos achieved a click-through rate (10.51%) approximately 1.78 times higher than filmed videos (5.89%). This difference was particularly pronounced on TikTok, where AI-generated videos showed approximately 2.52 times higher click-through rates than filmed videos, confirming a strong interaction effect between platform characteristics and production method.
These findings offer several implications. First, AI-generated videos can serve as a viable alternative for product categories where emotional appeal is important, such as cosmetics, enabling reduced production costs and time while achieving superior advertising effectiveness. Second, AI disclosure strategies require careful consideration of timing and salience rather than mere presence of disclosure; pre-disclosure does not negatively affect consumer responses, whereas post-disclosure triggers strong backlash effects, suggesting that clear upfront disclosure is advisable. Third, as AI content effectiveness varies by platform characteristics, a differentiated strategy that concentrates AI content resources on platforms characterized by rapid content consumption, such as TikTok, would be more efficient. Fourth, since psychological and attitudinal receptivity serves as a more important buffer than technical proficiency in AI advertising acceptance, communication strategies that position AI technology as an innovative tool are necessary. These findings also provide empirical evidence for developing disclosure guidelines under Korea's AI Basic Act, scheduled for implementation in 2026.
This study advances beyond prior research that relied exclusively on survey data by combining psychological response measurement through surveys with actual platform behavioral data collection, thereby providing empirical evidence for AI-generated advertising effectiveness. Furthermore, by identifying the differential effects of AI disclosure timing, this study offers insights for future policy discussions on AI advertising transparency. However, as this study was limited to Korean women aged 19-49 and utilized video stimuli from a single brand, future research should extend to diverse demographic groups, product categories, and cultural contexts, while tracking longitudinal changes in consumer responses to AI advertising.