Naver’s regulatory filings say more about where its search is going than its press releases do. Electronic Times (etnews) went through the Q3 disclosures on DART, Korea’s electronic disclosure system, and found that Naver spent ₩1.8897 trillion on R&D in the first nine months of 2026. That’s up 16.2% from ₩1.6263 trillion a year earlier and the highest nine-month figure Naver has reported. The nine-month total is already above Naver’s R&D spend for the whole of 2024 (₩1.8579 trillion). R&D rose from 18.4% to 18.7% of revenue.
The project list in the filing is more telling than the totals.
The R&D project list
According to etnews, Naver’s R&D projects as of Q3 include:
- a browser-based AI agent system
- browser-based on-device AI
- a shopping AI agent
- an LLM-based DevOps agent
New in Q3 were medical AI projects, including a medical search and clinical decision-support agent and a medical foundation model. etnews links these to Naver’s statement that it will add a health AI agent to AI Tab before the end of the year, and describes the browser work as part of bringing AI into Whale (Naver’s browser) and Naver Plus Store.
Reading it as a search roadmap
Each item maps onto a search surface we already follow:
- Health. Naver has just rebuilt hospital Place search around symptoms and procedures. A dedicated medical search agent with clinical decision support is the next step: health queries answered by an agent drawing on structured medical sources, with hospital listings next to the answer. If you work in Korean healthcare marketing, expect the gap between “being cited by the agent” and “ranking a blog post” to widen in 2027.
- Shopping. A shopping AI agent fits the agentic, transaction-led direction set out in Naver’s Q3 results. Product data quality in Naver Plus Store will matter more than on-page copy.
- Browser agents. A browser-based agent in Whale means Naver could act on any website, not just its own services, in the same way Yandex’s Alice AI uses a cloud browser to fill in forms. That makes clean forms and simple checkouts on Korean sites an “agent-SEO” issue, not only a UX one.
Kakao goes the other way
etnews sets this against Kakao. Kakao’s nine-month R&D was ₩962.4 billion, down 0.03% year on year. It has been flat for three years (₩972bn in 2024, ₩962.7bn in 2025, ₩962.4bn in 2026), and its R&D share of revenue fell from 16.1% to 15.7%. Kakao told the paper it is “thoroughly managing cost efficiency so that AI investment leads to real profit growth”.
There is one search-specific detail. In Q3 Kakao reorganised AI Studio, its core AI unit, which CEO Chung Shin-a oversees directly. The unit formerly called “AI Search Agent” is now simply “AI Search”. “AI Agent Platform” became “AI Incubation”, and “AI Connect” became “AI Core Platform”. etnews reads the reshuffle as preparation for a new AI agent service Kakao is launching next month.
Kakao no longer owns Daum, but it is still developing Kanana search inside KakaoTalk, so the rename is worth noting. Dropping “Agent” from the search team’s name while launching a separate agent product suggests Kakao is treating search as a feature inside its agents rather than an agent in its own right. That is the opposite of Naver, which is building agents into search.
My take
Over the past month Naver has published one AI Tab statistic after another. This filing shows what it is actually spending on, and almost every project is an agent. With nine-month spend already above the whole of 2024, the AI Tab health agent and shopping agent look like 2027 products, not experiments.
For marketers in Korea the gap between the two platforms is now large. Naver has put close to ₩1.9 trillion into R&D in nine months, and most of the listed projects turn search into agents. Kakao is holding spending flat and concentrating on making AI pay. If you are deciding where to put Korean search effort next year, there’s no real contest: optimise for Naver’s agents, its Place, Plus Store and AI Tab data, and treat Kakao as a messaging and ads channel, not a search one.





Leave a Reply