Graph-Augmented Source Credibility Modeling for Influencer Advertising in Product Discovery Environments

• Minh Hoang

  Department of Information Systems, Vietnam National University of Agriculture, Trau Quy Town, Gia Lam District, Hanoi 131000, Vietnam

  Author

• Quang Nguyen

  Department of Business Administration, Hung Yen University of Technology and Education, Dan Tien Commune, Khoai Chau District, Hung Yen Province 170000, Vietnam

  Author

Digital product discovery increasingly occurs through hybrid environments where consumers encounter influencer endorsements, recommendation panels, product bundles, and algorithmically ordered content within the same browsing session. This paper presents an empirical study of how influencer source type and graph-based product affinity signals jointly shape consumer attention, product consideration, and purchase intention in lifestyle e-commerce advertising. A controlled online experiment was conducted with 1,536 adult participants, 6,144 influencer-ad exposures, and 49,152 product-panel observations. Participants viewed mobile shopping feeds containing sponsored endorsements by either human influencers or AI-generated virtual influencers, followed by product recommendation panels whose ordering was generated from experimentally varied affinity-graph conditions. The study estimates source effects, product-affinity effects, and their interaction using mixed-effects regression, graph-regularized prediction, and structural response models. Results show that human influencers produce higher trust and authenticity, while virtual influencers produce higher novelty and initial attention. Product-affinity alignment improves product recognition, consideration, and purchase intention, especially when the endorsed item appears within a coherent cross-product neighborhood. A graph-augmented gradient boosting model predicts high-intention impressions with a ROC-AUC of 0.862, outperforming source-only and product-only baselines. Interaction tests indicate that virtual influencers are less penalized when recommendation panels provide coherent product-neighborhood evidence. The findings suggest that endorsement effectiveness in digital commerce depends not only on who presents the product, but also on whether the surrounding product graph makes the recommendation appear contextually meaningful.

• PDF

Copyright (c) 2026 International Journal of Science, Engineering and Applied Technologies

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.