The rapid diffusion of generative artificial intelligence has enabled brands to deploy deepfake videos and fully synthetic (virtual) influencers as endorsers on social media, raising questions about how consumers evaluate the credibility of non-human or digitally manipulated sources. Grounded in Source Credibility Theory and the Elaboration Likelihood Model, this study empirically examines how perceived credibility of deepfake and synthetic influencer endorsements shapes consumer trust and, in turn, purchase intention. A quantitative, descriptive-cum-causal research design was employed. Data were collected through a structured questionnaire using a 5-point Likert scale administered to 350 social-media users who had been exposed to influencer marketing content. Purposive/convenience sampling was used to reach the target respondent group. Data were analysed using SPSS and SmartPLS, including descriptive statistics, reliability and validity testing, exploratory factor analysis, correlation, regression, and structural equation modelling (SEM), with mediation analysis used to test the indirect effect of trust. Findings indicate that perceived credibility exerts a significant positive effect on trust, and that trust significantly and partially mediates the credibility-purchase intention relationship, while authenticity concerns and disclosure of synthetic origin moderate these effects. The study contributes to the emerging literature on AI-mediated endorsements and offers practical guidance for marketers and regulators on the ethical use of synthetic endorsers.