Naver, Daum ramp up AI search race with agent-focused strategies

Screenshots of Naver's new AI Tab search feature / Courtesy of Naver

Naver and Daum are intensifying competition in artificial intelligence (AI)-powered search, moving beyond chatbot-style responses toward agent-driven systems that leverage their proprietary data and service ecosystems to execute real-world tasks.

Naver recently unveiled the road map for its conversational AI search service AI Tab, which launched for all users in late June, positioning it as a gateway to AI agents capable of handling tasks such as reservations and shopping.

The company identified a product-native large language model (LLM), harness engineering and multimodal AI as three core technologies behind the strategy.

Unlike conventional foundation models built for general-purpose reasoning, Naver said its product-native LLM is tailored for real-world services. The model is designed to handle long-context, multiturn interactions while selecting appropriate tools to complete tasks across Naver’s ecosystem, including search, shopping, maps, blogs and community platforms.

Harness engineering connects the AI to Naver’s real-time data sources, including blogs, maps, shopping and news, enabling it to generate grounded, up-to-date responses.

Naver said it has improved efficiency by adopting a distributed architecture that uses multiple smaller models instead of a single large system, which doubled response speed while reducing operating costs by up to threefold.

The company added that it uses reinforcement learning, a trial-and-error training method, so the model asks follow-up questions when a query is unclear and cuts hallucinations by up to 30 percent than its previous HyperCLOVA X model in benchmark tests.

Naver is set to integrate AI Briefing, Smart Lens visual search features and real estate services into AI Tab this quarter, followed by dedicated AI agents for its Whale browser and health care later this year.

Upstage CEO Kim Sung-hoon speaks during the company's media day in Seoul, June 16. Courtesy of Upstage

Meanwhile, Daum is accelerating its own AI search strategy under Upstage, launching a beta AI Summary service powered by its proprietary LLM, Solar, last week, with plans for a full rollout later this year.

The feature automatically analyzes web documents and generates summarized search results with cited sources across categories including finance, health, entertainment and trending topics.

The AI Summary service marks the first major integration of Upstage’s AI technology into Daum since the startup acquired the portal’s operator AXZ earlier this year.

Upstage also plans to roll out conversational AI Mode later this year, enabling conversational search that allows users to explore topics through multiturn dialogue. The company argues that tighter integration between its model and platform will allow faster iteration and improved user relevance.

“Daum AI Summary marks a starting point for demonstrating how AI models can drive meaningful change when they intersect with everyday life,” Upstage CEO Kim Sung-hoon, said. “Going forward, we will add Upstage's AI across Daum's services to usher in an era of AI for everyone, where anyone can naturally enjoy the benefits of AI in everyday life.”