The AI Eye: LLMs vs Human Perception in Visual Marketing

Authors

Yi-Lin Tsai (Presenter), The University of Melbourne

Yung-Hsiu Lai (National Taiwan University)

Abstract

This research investigates how large language models (LLMs) replicate human perceptions of visual marketing content. Using image stimuli from four papers from leading marketing journals, the study examines OpenAI's ChatGPT and Anthropic Claude's interpretations of visual elements in advertising and logo designs and the corresponding product perception. Out of the four studies, LLMs successfully mimic human behaviors in two, show a null association in one, and demonstrate an opposite association in another. LLM’s qualitative responses are analyzed to rule out AI hallucination. While the results are robust among simulations over time, the responses can be heterogeneous across LLM providers. As marketers increasingly rely on Gen AI for visual content creation, our findings highlight the need for careful application and human oversight in AI-assisted marketing strategies.