Consumers’ Adoption Process of AI-Powered Visual Search Engines: An Application of the Diffusion of Innovation Theory

Abstract

As an increasingly common feature in online shopping, AI-powered visual search engines (AI-VSEs) allow consumers to search for products using images instead of text. Despite their growing presence, little is known about what leads consumers to adopt and continue using them. The study addressed this gap by building and testing an integrated model of the AI-VSE adoption journey based on the innovation-decision process model, theory of reasoned action, and cognitive dissonance theory. Specifically, it examined how consumers’ knowledge of AI-VSEs shapes their perceived characteristics of AI-VSEs, which in turn influence their attitudes, use intentions, perceived accuracy, and reuse intention of AI-VSEs. Online survey data were collected from a quota sample of 400 U.S. consumers recruited via Prolific. The measurements were culled from existing literature with modifications or newly developed and validated through confirmatory factor analysis. Results from structural equation modeling showed that consumers’ how-to and principles knowledge significantly influenced their perceptions of AI-VSEs. Among perceived characteristics, relative advantage and compatibility were the strongest drivers of attitudes, which in turn, significantly predicted use intention. Use intention subsequently influenced reuse intention for a specific AI-VSE platform both directly and indirectly through perceived accuracy. Findings of this study extend established theories to the emerging AI-VSE technology adoption context and provide validated measures of consumer knowledge and perceived innovation characteristics for future research. This study is among the first to test the full consumer adoption journey for AI-VSEs in a single integrated model. Practically, the study offers guidance for retailers and technology providers on educating consumers, focusing on how to use AI-VSEs, how AI-VSEs work, and their relative advantage, compatibility, and accuracy.