South Korean retailers are using AI-generated shoppers to test products, prices and store layouts before making changes for real customers. The experiments could help businesses spot weak ideas and costly friction earlier, but the most striking accuracy figure comes from a company selling the technology.
Watch Desk analysis
What happened
The Straits Times reports that Lotte Members is developing an AI “twin consumer” using purchasing data from 45 million L.Point users, alongside social-media data. Coupang and Naver are testing virtual shoppers on online shopping journeys, while CU is exploring simulations of product displays, customer traffic and merchandise choices in physical stores.
AI research firm Intellicia says it has completed 120 proof-of-concept projects with 50 companies. It claims its synthetic consumers achieved a 91.2% correlation with a survey of 500 real consumers. That is the company’s benchmark claim, not a guarantee that a virtual panel will predict what shoppers do in every market.
Why it matters
These tools are being pitched as a quicker, cheaper way to test product concepts and customer experiences than repeatedly recruiting survey panels. Retailers could use simulated journeys to find where people hesitate or abandon a purchase, or test store layouts before moving shelves in the real world.
The approach also raises a question about what “representative” means when virtual shoppers are built from consumer data and then used to stand in for consumers. A strong match on one benchmark is useful evidence, but it does not settle how well a model will capture a new product, a changing trend or shoppers missing from its data.
Our read
This is a practical application of generative AI, not just another chatbot in a new hat. But synthetic shoppers should be treated as a way to narrow options and sharpen questions, not as a replacement for asking real customers. A virtual panel can save time; it cannot make its own assumptions disappear.
What to watch
- Whether Lotte Members says how its AI twin is built and how consumer data is protected.
- How Intellicia’s 91.2% correlation was calculated, and whether it holds up beyond its benchmark.
- Whether retailers publish examples where synthetic-consumer tests changed a product, price or store layout.
Discussion spark: Should retailers use AI-generated shoppers to decide which products and store changes reach the public, or should real-customer testing remain essential?
Sources and evidence
- Source update (26 October 2026, 06:55 UTC)
- Source update (26 October 2026, 06:55 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.