Deepfake endorsement advertisements that attach the faces and voices of well known Indonesian figures to online gambling sites and investment schemes have become a recurring form of digital fraud. This study examines how such advertisements construct credibility for an endorsement that never took place. Using a qualitative multiple case design with multimodal semiotic analysis, it combines the meaning transfer model of celebrity endorsement, the three dimensions of source credibility, and the grammar of visual design. Seven cases from 2024 to 2026, each declared synthetic by fact checkers, by the impersonated figure, or by the institution concerned, were purposively selected and analysed through the documentation in verification reports. The analysis shows that synthetic credibility is assembled in three cumulative layers. The first layer appropriates the credibility dimension already attached to a public persona: expertise for financial officials and business leaders, trustworthiness for journalists and news presenters, and attractiveness for entertainers. The second layer fabricates the fit between endorser and product through scripts of ownership, generosity, insider access, and success testimony, so that a figure with no relation to gambling appears to have a reason to recommend it. The third layer converts that credibility into action through direct address or overheard testimony, concrete financial promises, and exit routes to external links or private messaging groups. The study concludes that the decisive weakness of these advertisements lies in their fabricated endorsement relation rather than in their technical realism, and it proposes endorser consent verification and fit literacy as complements to technical detection in design education and platform governance
Copyrights © 2026