Naver AI Shopping Agent Sees 82% Transaction Surge, Adds Delivery-Based Recommendations

Naver's AI shopping agent recorded an 82% surge in transactions and has added delivery-based recommendations, according to a report by Chosun Ilbo. The numbers point to a conversational shopping assistant that is now moving real orders inside Naver's commerce business, rather than only improving how products are browsed.

Delivery-based recommendations suggest fulfilment is being folded into the ranking logic: when arrival time and delivery coverage become part of what users see, price and popularity are no longer the only signals on screen. For fresh food and everyday essentials, where timing often decides the purchase, that dimension sits close to the actual moment of decision.

The significance of the 82% figure is that it measures transactions rather than traffic. Driving more clicks with an AI assistant is not hard; turning a conversation into a completed order is. By disclosing transaction volume, Naver frames its agent as an entry point for purchases, not just an advice tool.

Naver's position in Korea makes that framing easier to sustain. It combines search, shopping and payments, so an AI agent can be grafted onto an existing purchase path. Embedding models into a ready-made catalogue of products, prices and delivery data shortens the loop between recommendation quality and conversion, an advantage pure model providers find hard to replicate.

For the wider e-commerce industry, shopping assistants are shifting from demo features to channels that carry revenue. Transaction volume is harder to dress up than conversation counts or active users, and it says more about whether recommendations actually improve matching. Naver's choice of metric could push competitors to publish figures closer to business outcomes.

What to watch next is whether Naver breaks out more detail, such as retention driven by the agent, changes in basket size, and what fulfilment-aware recommendations cost in delivery terms. The 82% growth will only matter if it lowers the effort of choosing, rather than moving orders that would have happened anyway into an AI entry point.