As users favor authentic, relatable content, fast-food brands have adopted user-generated-content-like (UGC-like) aesthetics that may obscure persuasive intent. Drawing on brand authenticity and engagement research, this project examines whether UGC-like aesthetics generate greater engagement than corporate aesthetics. The study analyzes 10,000 Meta-Content-Library Instagram photo posts from 25 brands, using like-to-view ratios to measure engagement. It contributes a theory-informed human–AI framework classifying aesthetic style along corporate-commercial and user-like/platform-native dimensions using 12 visual characteristics (e.g., human presence, product centrality, staging, polish, etc). Domain patterns and coder judgments classify images as corporate, UGC-like, hybrid, or neutral/other. Multilevel statistical models will compare engagement across those types while accounting for brand, campaign, format, timing, promotions, and disclosures. Current efforts establish coding reliability. Three coders classify 100 McDonald’s images through blind rounds using Krippendorff’s alpha, disagreement review, and retesting. Findings show moderate agreement for human presence (α=.68), product centrality (α=.62), and brand prominence (α=.62), while meme/platform-native cues and staging remain less consistent. Once reliability is achieved, coders will label 20 images per remaining brand, creating a 580-image cross-brand gold-standard dataset. A 60/20/20 train/validation/test split will be used to train a multimodal CLIP–BERT classifier with MLP to classify aesthetic styles. Performance will be evaluated using macro-F1 and confusion matrices. The classifier will label remaining images, with low-confidence predictions reviewed by human coders. The dataset will enable us to examine whether UGC-like aesthetics generate greater engagement than corporate styles, providing insight into brands' incentives to adopt peer-like advertising and informing transparency and disclosure standards for digital food marketing.

Authors:
Yu Wang, University of Georgia, United States
Amelie Wei, South Eugene High School, United States
Nicole Ngo, University of Oregon, United States

About the Presenter(s)
Yu Wang is an Assistant Professor at the University of Georgia. He leads the KIND Lab and studies machine learning, agentic AI, LLMs, and graph-structured data for effective and trustworthy scientific and infrastructure applications.
https://www.linkedin.com/in/yuwanggraphml/

Amelie Wei is a senior student at South Eugene High School, in Eugene, Oregon, USA. Her research interest lies in visual analysis for marketing of online businesses.
https://linkedin.com/in/ameliewei/

Connect on Linkedin
https://www.linkedin.com/in/yuwanggraphml/

Connect on ResearchGate
https://www.researchgate.net/profile/Yu-Wang-623

Additional website of interest
https://yuwang0103.github.io/

See this presentation on the full schedule – On Demand Schedule

Bookmark