Naver's robots.txt explicitly prohibits bots that collect content for AI training or retrieval, and it names crawlers such as GPTBot, OAI-SearchBot and ClaudeBot. If your brand's Korean content lives only on Naver Blog or Naver Cafe, ChatGPT and Perplexity cannot read it, so they cannot cite it.
Korean consumers now look for brands in two ways. Many still search on Naver. A growing number ask ChatGPT, Perplexity or Google AI in plain conversational questions. The catch is that these two worlds read different sources, and Naver has drawn a clear line between them.
Naver put the block in writing
A common assumption among teams entering Korea goes like this: “We have plenty of posts and reviews on Naver Blog and Cafe, so AI tools will pick those up too.” They won't. Naver explicitly blocks outside AI systems from collecting its content.
Websites use a file called robots.txt to tell crawlers what they may access. The Internet Engineering Task Force (IETF) formally standardized this convention in 2022 as RFC 9309, the Robots Exclusion Protocol. OpenAI, Google and Anthropic state in their official documentation that their crawlers follow it.
Open the robots.txt files for Naver Blog and Naver Cafe and the top of each carries this line, in English:
BOT ACCESS FOR THE PURPOSES OF AI TRAINING AND RETRIEVAL-AUGMENTED GENERATION (RAG) IS STRICTLY PROHIBITED.
The bots Naver names
Naver doesn't stop at a statement. It lists the crawlers it blocks by name, including these:
- GPTBot. OpenAI's training crawler. OpenAI's documentation describes it as collecting content that may be used to train its generative AI foundation models.
- OAI-SearchBot. The bot behind ChatGPT's search answers. OpenAI documents it separately from GPTBot.
- Google-Extended. Google's access token for training models such as Gemini.
- ClaudeBot and Claude-SearchBot. Anthropic's crawlers for Claude.
- PerplexityBot and CCBot. Perplexity's crawler, and the Common Crawl bot whose data is used to train many AI models.
The key detail: Naver blocks not only training bots but also the AI search bots that fetch pages in real time to answer a user's question. Inside Naver, it is effectively a walled garden.
Why blocked content can't earn AI recommendations
For an AI assistant to mention a brand in an answer, it needs a source it can actually read about that brand. If all of that evidence sits inside Naver Blog or Cafe, then from the blocked assistant's point of view the information does not exist.
This affects search visibility directly. OpenAI's own documentation spells out what happens when a site blocks its search bot:
Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links.
Put simply, Naver keeps its data for its own services and closes it to outside AI. What ChatGPT, Google and Perplexity can read is the open web. For a brand, the part of the open web you control is your own website.
How other Korean platforms compare
Naver is not the only platform making this call. We checked the robots.txt files of 15 major Korean platforms on July 27, 2026, and rechecked all 15 at the bot-name level on August 27, 2026. The summary below reflects what the files said at the time of checking.
| Platform | AI training bots | AI search bots | Note |
|---|---|---|---|
| Naver Blog | Blocked | Blocked | Training and search bots blocked by name, plus the RAG prohibition notice |
| Naver Cafe | Blocked | Blocked | All bots blocked, including Google and Bing search crawlers |
| Naver SmartStore | Blocked | Blocked | AI bots blocked by name, everything else blocked too |
| Naver main site | Blocked | Blocked | Everything beyond the front page blocked |
| Kakao Map | Blocked | Blocked | Uses the same RAG prohibition notice as Naver |
| Brunch (Kakao) | Blocked | Allowed | Training bots blocked; AI search bots allowed on par with regular search |
| Tistory (Kakao) | Allowed | Allowed | No AI bot restrictions; only admin paths blocked |
| DiningCode | Blocked | Partial | Training bots blocked; ChatGPT search bot allowed, Perplexity blocked |
| Siksin | Allowed | Allowed | No AI bot restrictions |
| Catch Table | Blocked | Blocked | Allowlist approach; AI bots are not on the list |
| Baemin | Blocked | Blocked | Only Googlebot and Naver's bot allowed |
| Yogiyo | Allowed | Allowed | No AI bot restrictions; only payment and similar paths blocked |
| Daangn | Blocked | Blocked | 46 AI and scraping bots blocked on Korean content: the broadest list |
| Blocked | Partial | Training bots and PerplexityBot blocked; other AI search bots not mentioned | |
| YouTube | Allowed | Allowed | Watch pages open, but only text such as titles and descriptions is read |
Naver Place is not in the table because its robots.txt could not be retrieved (the request was rate limited with a 429 response). Robots.txt files can change at any time, so treat this as a snapshot. The Brunch file, for example, carries an update note dated April 22, 2026.
Two caveats apply. Robots.txt is a request, not a technical barrier, but because major AI companies publicly commit to following it, the practical assumption should be that blocked means not citable. And blocking does not erase data that was already used for training before the block.
What a foreign brand should do
Each engine leans on different sources. Naver's AI draws on Naver Place and reviews, while ChatGPT and Google look mainly at websites on the open web. For a fuller picture of how the two ecosystems differ, see our breakdown of Naver vs Google in Korea.
- Own a Korean-language site on your own domain. This is the channel that AI assistants outside Naver can read, and the only one whose crawler policy you control. A platform that is open today can close tomorrow.
- Make sure AI crawlers can actually reach it. A site can load perfectly for people and still be invisible to AI if a firewall or robots.txt setting blocks crawlers. We have seen a site that never appeared in ChatGPT for exactly this reason: a default firewall setting left in place by the site builder.
- Structure it for machines. Structured data and question-based content help AI systems understand what the page is about and cite it.
- Keep your Naver presence. Naver Blog, Naver Place and Naver's own AI Briefing still shape what Korean users see inside Naver. Treat Naver and your own site as two channels with different jobs, not as substitutes.
The more Naver raises its walls, the more valuable an open, well-structured site outside those walls becomes. If you are planning a launch, our guide to Korea market entry looks at how the channels fit together, or you can talk to our team in Seoul about your brand's current setup.
Frequently asked questions
Q1.If our brand publishes a lot on Naver Blog, will ChatGPT cite it?
In most cases, no. The robots.txt files for Naver Blog and Naver Cafe name generative AI crawlers such as GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended and block them. Crawlers that follow the robots.txt standard do not collect those posts, so they are unlikely to be used as sources in AI answers outside Naver.
Q2.Does that make Naver marketing pointless for a foreign brand?
No. Naver Blog, Naver Place and Naver's own AI search still matter inside Naver. That value stays inside Naver, though. To appear in AI answers outside Naver, such as ChatGPT, Google AI and Perplexity, your brand needs a separate website that is open to their crawlers. Each engine draws on different sources.
Q3.Is having a Korean website enough to show up in AI answers?
No. If a security firewall or a robots.txt setting blocks AI crawlers, the site effectively does not exist for AI, even though it loads normally for people. The site has to be reachable by AI crawlers and structured so machines can understand it, for example with structured data and question-based content.
Q4.What is the difference between blocking AI training bots and AI search bots?
Training bots such as GPTBot, ClaudeBot, Google-Extended and CCBot collect data to train AI models. AI search bots such as OAI-SearchBot, ChatGPT-User, Claude-SearchBot and PerplexityBot read pages in real time when a user asks a question and cite them in the answer. A site that blocks only training bots can still be cited. A site that also blocks search bots cannot. Naver blocks both.
Korean version: 네이버가 AI 접근을 막았습니다
