Daum, the portal subsidiary of Korean internet giant Kakao, has launched a beta AI search service in partnership with Upstage, a domestic artificial-intelligence start-up. With a roadmap to expand into fully conversational search before the year's end, the collaboration has prompted speculation that South Korea's long-settled search market may at last be ripe for disruption.
A portal giant on the back foot
Daum's position in the domestic search market has been eroding for years. According to Internet Trend, Naver holds roughly 55–60% of South Korean search traffic as of 2025, while Daum has shrunk to the high single digits. Factor in Google's growing share of mobile search, and Daum's room to manoeuvre narrows further still.
Against that backdrop, the tie-up with Upstage reads less as a feature upgrade than as a survival strategy. Upstage is one of South Korea's most prominent AI start-ups, best known for its "Solar" large language model (LLM), which has ranked highly on the HuggingFace open LLM leaderboard — a widely watched benchmark for comparing open-source language models. The company's core strengths lie in document processing and retrieval-augmented generation (RAG), a technique that enables language models to draw on real-time external data when generating responses.
What conversational search actually means
Conversational search moves beyond the traditional model of typing keywords into a box. Instead, users pose questions in plain language, and an AI system interprets the context and delivers a synthesised answer. Prominent international examples include OpenAI's ChatGPT Search, Perplexity AI, and Google's AI Overviews feature.
Google began rolling out AI Overviews — which place an AI-generated summary above conventional search results — across major markets including the United States from 2024. Perplexity, meanwhile, has grown rapidly as an alternative to legacy portals, and is now valued at over $3bn (roughly 4 trillion won). In South Korea, Naver already operates its own conversational AI search service, known as "Cue:", while Kakao is advancing its AI capabilities through a proprietary LLM called Kanana.
The logic of the Upstage partnership
Upstage's Solar model has earned creditable scores on domestic and international benchmarks, with particular recognition for its handling of the Korean language and document comprehension. Analysts suggest that combining Daum's extensive Korean web-crawling data with Upstage's language model could yield a meaningful edge in Korean-language search quality.
RAG technology is central to this proposition. By enabling an LLM to retrieve live web data when formulating answers, it improves both accuracy and freshness while reducing the "hallucination" problem — instances in which AI models generate plausible-sounding but factually incorrect responses. Upstage's accumulated expertise in this area is expected to bolster the reliability of Daum's search offering.
The sceptics' case
Not everyone is persuaded. Industry observers argue that clawing back search share requires more than technical prowess — it demands a compelling platform ecosystem that keeps users engaged. Naver benefits from a powerful lock-in effect: its integrated services spanning shopping, news, maps, and community forums give users little reason to look elsewhere. Daum, by contrast, sits within Kakao's orbit and has access to KakaoTalk, South Korea's dominant messaging app — yet the connection between that platform and Daum's search product has historically been weak.
There is also the cost question. AI search services are computationally intensive, and even well-resourced players such as Google, Microsoft, and Naver are grappling with surging server costs since deploying AI search at scale. Whether the Daum–Upstage alliance can construct a sustainable business model remains unproven.
A threat to the advertising model
The rise of AI search poses a structural challenge to the advertising economics of the portal industry as a whole. Conventional keyword-based search advertising generates revenue when users click through to external pages. In a conversational search environment, where the AI synthesises an answer directly, the incentive to click elsewhere diminishes — potentially undermining the click-based advertising model at its foundations.
Perplexity initially championed a subscription model free of advertising before recently reversing course and introducing ads. Google, too, is watching closely for any impact on click-through rates since deploying AI Overviews. How Daum intends to monetise its AI search service remains unclear.
Niche challenger or market disruptor?
Analysts broadly agree that the Daum–Upstage partnership is unlikely to overturn Naver's dominance in the near term. The more pertinent question is whether it can claim a strategic foothold at a pivotal moment of technological transition. Search markets have, historically, proved capable of reshuffling during major platform shifts — as the move from desktop to mobile demonstrated. If AI search proves to be another such inflection point, the argument runs, the window for meaningful repositioning may be closing.
One scenario that analysts find plausible is integration with KakaoTalk: weaving conversational search into everyday messaging in a way that feels natural to users, potentially recasting Daum not as a legacy portal but as what its architects might call a "life AI platform." The formal launch of conversational search, promised before the end of the year, and the user response it receives, will constitute the first real test of that ambition.