Key points
- Daum is writing new rules for AI crawlers
- Naver already prohibits several major AI crawlers
- An analyst questions whether crawler rules will be enough
Daum (다음), one of Korea's two big web portals, is preparing new rules on how outside AI services such as ChatGPT can collect its content, the Korea Economic Daily (한국경제) reported on September 28. Naver (네이버, 035420), the country's largest portal, already prohibits several major AI crawlers from accessing its blogs and online communities.
The review covers Daum's portal, its Tistory (티스토리) blogging service, and Daum Cafe. Outside AI services currently collect content from those sites to generate answers, according to the newspaper. The report did not say when the rules would take effect or which crawlers they would cover.
Upstage's plans for Daum
Daum is operated by AXZ, which Kakao (카카오) agreed to sell to Korean AI startup Upstage (업스테이지) in May in exchange for Upstage shares. Upstage plans to combine its Solar AI model with Daum's search engine and content to build a new AI portal, Electronic Times (전자신문) reported.
At the time of that report, the deal still needed Fair Trade Commission approval. One question before the regulator was whether other AI companies would retain access to Daum's data for training. Upstage is also preparing to list on the Kosdaq.
What Naver's crawler rules say
Naver banned outside AI services from collecting its data without permission in June last year, according to the Korea Economic Daily. "We're evolving our own technology to stop attacks that try to take data by going around the policy," a Naver official told the newspaper.
A website tells automated crawlers what they may read in a public file called robots.txt. Naver Blog's file names nine AI crawlers one by one and disallows each of them from the whole site. The list includes OpenAI's GPTBot and OAI-SearchBot, Google-Extended, PerplexityBot, Anthropic's ClaudeBot, and Meta's crawler. The section opens with a line in capital letters.
"BOT ACCESS FOR THE PURPOSES OF AI TRAINING AND RETRIEVAL-AUGMENTED GENERATION (RAG) IS STRICTLY PROHIBITED," the file says.
Google-Extended is the setting that controls whether Google can use a site's content for its Gemini AI models. The file doesn't block Google's regular search crawler from Naver Blog. Naver Cafe's file goes further and disallows almost every crawler, including the regular search bots from Google and Microsoft's Bing. Daum Cafe's file, by comparison, doesn't prohibit crawling of its posts, as of September 29.
Why the content matters to advertising
The portals see the Korean-language posts that users have put on their sites over decades as their main strength against global AI, the Korea Economic Daily reported. Blog posts, cafe posts, and question-and-answer pages hold local information, product reviews, and everyday tips that general AI training data doesn't cover well.
A Hyundai Motor Securities (현대차증권) report makes the same point about Naver, EBN reported. Place reviews, shopping information, and blog reviews that people use for buying decisions are concentrated on Naver, the report said. Because Naver restricts outside crawling, global AI models have limited access to that data, according to the report. The report said this is why an outside AI chatbot has a harder time with a question like "Tell me good restaurants in Gangnam" than Naver search does.
Naver's search advertising has held up better than other local portals. Naver's ad sales grew 6 percent in 2024 and 9 percent last year, and the report expects 8 percent growth this year. China's Baidu has seen its ad sales shrink since the second quarter of 2024, and Yahoo Japan's search business has shrunk for five straight quarters.
"The spread of AI search will keep hurting traffic to legacy search," said Choi Seung-ho, an analyst at Hyundai Motor Securities.
"Among local search portals, the reason Naver is holding up relatively well is likely the difference in its mix of business queries and informational queries," he continued.
Google's app passed Naver's in July
App usage offers another measure of the competition. In July, the Google app had 47.02 million monthly users in Korea, about 178,000 more than the Naver app's 46.85 million, according to Mobileindex (모바일인덱스) data cited by EBN. It was the first time Google's app had overtaken Naver's in the dataset, which began in March 2021. The Korea Economic Daily tied the change to the Google app adding Gemini, and reported that ChatGPT has also passed Daum in monthly users.
App users alone don't show how many people have stopped using Naver search, or how much AI has replaced it, EBN noted.
Naver is building its own AI search at the same time. Its AI Briefing feature, which summarizes search results, has passed 30 million monthly users. AI Tab, a search assistant people can talk to, has passed 10 million users in total, EBN reported.
The limits of blocking crawlers
The Hyundai Motor Securities report also questioned whether crawler restrictions would be enough as browser-based AI agents become more common. It described agents operating through a user's browser and login, reading pages and clicking through the site much as a person would.
"It can't block data collected that way," Choi said.
Agents are still slow and costly for collecting data in bulk, since each page takes time and computing work, the report said. The report named a different key risk. If people only read the answer an AI has put together, they see fewer of Naver's search pages and the ads on them, so ad sales can suffer even when the data stays locked.
Meanwhile, Naver shares closed down 0.7 percent at 194,900 won ($144) on Tuesday.
The comments from Naver and Hyundai Motor Securities are translated from Korean. Won figures are converted at 1,354 won to the dollar, the Yahoo Finance rate on September 29, 2026.
