Naver’s AI Shopping Agent Now Sells Before You Search

Naver’s AI Shopping Agent will now start the sales pitch on its own. On June 1 the South Korean internet giant updated the tool inside its Naver Shopping app so it reads a user’s clicks, saved items and cart additions, then floats a shopping suggestion on the home screen before anyone types a single query. Three months after the agent went live in beta, the search box has started talking back.

That sounds like a minor convenience upgrade. It arrives in the middle of a global race to rebuild online retail around AI agents that push products at shoppers rather than wait to be asked, and Naver is moving to defend the most valuable real estate in Korean commerce before a rival agent reaches its customers first.

Naver’s Shopping Agent Starts the Conversation First

Until this update, the agent waited for a prompt. A user typed “sofa,” and the tool summarized buying tips by room size, material and popularity, then pointed to suitable brands. Useful, but reactive. The June 1 release flips that order by analyzing a shopper’s recent behavior alongside current trends, then opening with a suggestion the person never requested.

The examples Naver gave are deliberately mundane. A user living alone who has been searching meal kits might be greeted with a prompt offering to find single-serving kits among items they recently viewed. Someone who dropped a moisturizer into their cart could see a nudge to recommend skincare products that pair well together. The agent also folds complex conditions into its openers, the kind shoppers rarely bother to type.

What the proactive version now does on its own:

• Reads clicks, saved items and cart additions as live signals of intent, not just typed keywords
• Opens conversation prompts on the app home screen immediately after the update installs
• Bundles fiddly filters into one suggestion, such as meal kits that cook in under 10 minutes or a three-pack priced around 10,000 won (about $7)
• Targets life-stage context, like solo households, rather than treating every shopper as identical

Push Beats Pull in the New Commerce Playbook

For two decades, online shopping ran on pull. You knew roughly what you wanted, you typed it, the engine returned a wall of options, and the retailer’s job was to rank them. The search bar was the front door, and whoever owned the most-used bar owned the funnel.

Agentic commerce inverts that. A capable agent watches behavior, infers a need and proposes the purchase, sometimes before the shopper has consciously decided to buy. The trigger moves from the customer to the software. Lee Jung-tae, head of Shopping Search and AI at Naver, framed the update as the agent evolving into an execution-oriented system that not only understands a shopper’s context but suggests the next action to take.

That shift is the part most coverage skipped past. The headline is a feature note; the substance is Naver retraining its biggest commercial muscle to initiate rather than respond. eMarketer’s analysis of how AI is reshaping retail operations puts the same pivot at the center of the industry’s near-term roadmap, with discovery and recommendation increasingly handled by software that acts on a shopper’s behalf.

The prize is obvious. A shopper who follows a prompt has effectively outsourced the first decision, which compresses the path from idle browsing to checkout. The risk is just as obvious, and it sits at the other end of this article.

The Conversion Math Behind the Bet

Naver has reason to lean into push. Early engagement on the agent-powered app has run well ahead of its older web shopping experience, and the company has been blunt about wanting that gap to widen as the agent learns to close sales rather than just guide them.

The numbers from the agent’s first months on the Naver Plus Store app, where the tool launched in beta on February 26:

• 8.38 million monthly active users (MAU, a count of unique monthly visitors) on the app in April, lifting Naver Plus Store past AliExpress to third among Korean general shopping apps
• 84% higher purchase conversion in the app than on the web, which Naver attributes to longer sessions and more frequent visits inside the agent experience
• 28% quarter-on-quarter growth in first-quarter app transaction value
• First time in its 27-year history that commerce is set to overtake search as Naver’s largest segment, against group revenue of 3.24 trillion won (about $2.2 billion) in the first quarter, per NAVER’s first-quarter 2026 results

Amazon and Walmart Are Running the Same Race

Naver is not improvising. The same instinct is visible across the world’s largest retailers, each trying to become the layer that decides what a shopper sees and buys.

Amazon retired its standalone Rufus chatbot in May and folded the technology into a relaunched assistant for shopping that it says reaches more than 250 million users. Walmart took a different route with Sparky, its goal-oriented agent that plans around requests like “host a cookout,” then pushed the same assistant out into ChatGPT and Google’s Gemini rather than keeping it walled inside one app.

The strategies split along a clear line. Naver and Amazon are reinforcing their own storefronts as the place the agent lives. Walmart is hedging that shoppers will increasingly start inside someone else’s AI, and is planting itself there too.

The money behind the trend explains the urgency. AI platforms are on track to drive roughly $20.9 billion of US retail e-commerce in 2026, nearly quadruple the prior year, according to the 2026 forecast for AI-driven retail sales. Accenture has estimated that more than 30% of online commerce could route through AI agents by 2030.

Whoever owns the prompt owns a growing slice of that flow. That is why a Korean app tweak and an Amazon relaunch belong in the same sentence.

Coupang, the Breach, and Korea’s Two-Horse Lead

At home, Naver’s bet plays out against one rival above all. Coupang and Naver have traded the top of Korean e-commerce for years, with Naver’s commerce gross merchandise value (GMV, the total value of goods sold across the platform) sitting in the low 50 trillion won range, neck and neck with its competitor.

The standings shifted this spring. Coupang still commands far more raw traffic, around 33.1 million monthly users, but that base edged down 0.2% while Naver’s shopping app climbed. A Coupang data breach disclosed in March sent some shoppers looking elsewhere, and Naver Shopping’s February transaction volume rose 28% year over year, outpacing both the broader market and Coupang’s growth. Statista’s tracking of South Korea’s e-commerce market share by platform shows how tightly the leaders are bunched.

The proactive agent is Naver’s lever to convert that momentary advantage into habit. A shopper who opens the app each morning to a tailored suggestion has a reason to return that a search bar alone never gave them, and habit is the one thing a data breach cannot easily hand a competitor.

Where Proactive Selling Meets the Trust Line

The upside is real, and so is the catch. An agent that suggests before you ask is also an agent making commercial choices on your behalf, and the line between a helpful nudge and a manipulative one is thin. A prompt steering a solo diner toward a pricier three-pack looks like convenience until the shopper notices the cheaper option never surfaced.

There is a structural risk too. The more shoppers learn to follow agent prompts, the less attached they become to any one storefront, because a prompt from ChatGPT or Gemini reads much the same as a prompt from Naver. By teaching its users to trust push commerce, Naver is also training them for a world where the agent that reaches the shopper first wins the sale, even if that agent belongs to someone else.

Naver’s answer is to bind the agent ever tighter to assets a rival cannot copy: its membership benefits, its delivery network and a commerce-specific language model trained on years of Korean pricing, reviews and buying behavior. The company plans to turn the guide into a full transaction agent that handles checkout, not just discovery.

If that bundle keeps app conversion running ahead of the web through the second half of the year, Naver converts a search habit into a buying habit it controls. If shoppers drift toward general-purpose agents that treat every store as interchangeable, the proactive prompt becomes a feature Naver built to train customers it may not keep.