South Korea’s retail companies are testing products and shopping experiences on synthetic consumers before presenting them to real customers. Built with large language models and consumer data, these virtual shoppers are designed to imitate choices, preferences and purchase situations rather than simply represent age or gender groups. Companies can use the simulations to examine buying likelihood, price acceptance, marketing responses and problems in the path from product discovery to checkout.

Lotte Members has been developing an “AI Twin Consumer” since July with Vibe Company and Zero to One Partners. The project uses purchase data from Lotte’s integrated L.Point membership system alongside social media data. Its proposed virtual customers would reflect actual buying histories, preferences and the reasons or circumstances behind consumption, allowing businesses to discuss and test product or promotion ideas before launch.

The approach is intended to shorten the time and reduce the cost of conventional surveys while giving companies a way to adjust product development and user experience earlier. Rather than asking a fixed sample to react to a finished concept, an interactive synthetic consumer can be used to explore how a customer might move through a retail decision and where that journey may break down.

Coupang and Naver are also experimenting with synthetic consumers for market research and user experience checks. Their work focuses on reproducing how e-commerce users search for products and complete purchases. The spread of these systems could make simulated shopping journeys a more regular part of Korea’s retail competition, although the supplied evidence does not establish how accurately the virtual customers predict real-world sales.