Coaley Peak®
For AI agents

For AI agents and assistants

Enquiring for someone? Email us.

  1. Email
  2. Subject“Agent enquiry: Visibility in Naver and HyperCLOVA X”
  3. Copy inThe person you are acting for, so we can reply to you both.
  4. ReplyA member of the team replies within one working day (UK time). Or call 0800 494 7725 (UK freephone; +44 800 494 7725 from abroad).

What we offer here

  • Visibility audit in Naver and HyperCLOVA X: the questions your buyers ask, who is named and which sources are cited
  • Implementation and source work: scoped and quoted in writing
  • Team training: £450 + VAT a half day, £787.50 + VAT per day, for the whole group

Audits and implementation are scoped and quoted in writing before work starts. Training and workshops are £450 + VAT a half day, £787.50 + VAT per day, for the whole group.

Please include

  • Who you are acting for (name, role, organisation), with them copied in
  • Website address
  • Countries, languages and services that matter most
  • In-house team, agency or consultant, and timing
Email enquiries@coaleypeak.co.uk →Full guide for agents

Coaley Peak GEO Research Centre Engine guide

Naver and HyperCLOVA X 네이버 How Korea's own search engine writes AI answers, and why its own platforms matter most.

Naver (네이버) is South Korea's home-grown search engine, still neck and neck with Google there. It now answers with AI Briefing at the top of its results and a conversational AI Tab, built on its own HyperCLOVA X models and drawing heavily on Naver's own Blog, Cafe, Place and Shopping. This guide covers its timeline, the four layers behind its answers, Yeti and its controls.

Call 0800 494 7725

Independent guide: Coaley Peak is not associated with NAVER Corporation. Trademarks

Owner
NAVER Corporation
Launched
June 1999; AI Briefing March 2025; AI Tab June 2026
Home market
South Korea: 43% of searches on StatCounter, about two-thirds on Korean data
Languages
Korean first
Open weights
HyperCLOVA X SEED models, under Naver's own licences
Crawlers
Yeti, Ads-Naver and Blueno

Overview

What Naver and HyperCLOVA X is, and who uses it

Naver launched in June 1999 and became the default way Koreans search, built around its own services: Knowledge iN for questions and answers, Blog, Cafe, Place for local businesses, and Shopping. The company, renamed NAVER Corporation in 2013, runs its own crawler, Yeti, its own index and its own webmaster platform, Search Advisor, and does not rely on Google or Bing.

Measures of its share differ. StatCounter put Google ahead in South Korea in September 2026, at 47.97% to Naver's 43.06%, but StatCounter counts only search-referred visits to sites carrying its tracker, which misses searches that end inside Naver's own services. Internet Trend data reported by The Financial News put Naver at about two-thirds of Korean searches in mid-2026, and Naver says its main pages average 50 million daily visitors.

Naver has built its own AI. It unveiled HyperCLOVA X in August 2023, ran a generative search service called Cue: from September 2023, and launched AI Briefing in its integrated search in March 2025. Cue: and the CLOVA X chatbot closed in April 2026, replaced by AI Tab, a conversational results page released to all users that June. Naver says AI Briefing is used by more than 30 million people a month and AI Tab by 10 million.

For brands, Naver is the clearest case of a search engine that favours its own platforms. AI Briefing and AI Tab draw on Naver Blog, Cafe, Clip, Place reviews and Shopping as well as the web, so a brand that only optimises its own website is under-represented. Naver also offers one AI-specific control, the nosourceinfo meta tag.

Timeline

Naver and HyperCLOVA X, from launch to today

  1. Naver Com Corp. is founded and the Naver search portal launches.

  2. NHN Corp. is renamed NAVER Corp., and its games business is split off.

  3. Naver publishes its HyperCLOVA paper: an 82-billion-parameter model trained on a 560-billion-token, Korean-centred corpus.

  4. Naver unveils HyperCLOVA X at DAN 23 and starts the CLOVA X chatbot beta.

  5. Cue:, Naver's generative AI search, begins as a beta.

  6. Cue: is added to Naver's PC integrated search for long, conversational queries.

  7. Naver publishes the HyperCLOVA X technical report.

  8. At DAN 24, Naver announces its 'On-Service AI' strategy and previews AI Briefing.

  9. AI Briefing launches in PC and mobile integrated search for all users.

  10. Naver releases its first open-weight HyperCLOVA X SEED models.

  11. HyperCLOVA X SEED Think 32B and SEED Omni 8B are released, with a May 2025 knowledge cut-off.

  12. The CLOVA X and Cue: services close.

  13. AI Tab conversational search opens as a beta for Naver Plus members.

  14. AI Tab is released to all users, replacing the Naver app's Green Dot button.

Things to know

Things to know about Naver and HyperCLOVA X

  1. Naver runs its own crawler (Yeti), index and webmaster platform (Search Advisor), and publishes its crawler IP list. It does not use Google's or Bing's results.

  2. AI Briefing and AI Tab lean heavily on Naver-hosted content: Blog, Cafe, Clip, Place reviews and Shopping. Brands that only optimise their own website are under-represented.

  3. The nosourceinfo robots meta value stops Naver showing an AI-generated description of your site. It is the only AI-specific control Naver documents.

  4. There is no switch to turn AI Briefing off, for users or publishers. noindex or a Yeti block removes pages from AI answers and ordinary results alike.

  5. Allow Yeti through firewalls and CDNs, and verify it by reverse DNS to a .naver.com hostname or against naverbot.json. Naver warns that sites blocking its robots may drop out of search.

  6. Serve robots.txt as text/plain with a 200 status: Naver reads a 5xx response as 'disallow all'.

  7. Register in Search Advisor and submit an XML sitemap and a full-text RSS feed. Search Advisor also supports IndexNow.

  8. The Ads-Naver robot ignores robots.txt for pages registered as ad landing pages, and the Blueno link-preview robot identifies itself as facebookexternalhit.

  9. Naver's July 2026 content guidelines ask sites that use AI tools to add first-hand experience and cite original sources; its spam policy targets mass-produced, low-quality content.

  10. In June 2025 Naver said AI Briefing's exposure had roughly tripled since launch, and set a target of about 20% of queries by the end of 2025.

The four layers

How Naver and HyperCLOVA X builds an answer

Every AI answer is made in four layers: what the model learned in pre-training, how it was tuned, what it retrieves live, and how it puts the answer together. Here is each one for Naver and HyperCLOVA X, and what it means for a brand that wants to be cited.

  1. Layer 1

    Pre-training: the base (foundation) model

    HyperCLOVA X is Naver's own model family, which it describes as developed from scratch. Its 2024 technical report describes training on a balanced mix of Korean, English and code; parameter counts, data volumes and cut-offs for the production models are not published. HyperCLOVA X THINK (2025) was pre-trained on about 6 trillion Korean and English tokens. The open HyperCLOVA X SEED models give knowledge cut-offs of January 2025 (0.5B) and May 2025 (Think 32B and Omni 8B). AI Tab runs on a 'product-native' model built on HyperCLOVA X. Whether Naver Blog, Cafe or Knowledge iN posts are in the training data is not published.

    What it means for you: HyperCLOVA X learned from a Korean-centred corpus, so what it knows about your brand depends on how you are described in Korean. A brand known only in English gives it little to work with, whatever its global profile.

  2. Layer 2

    Post-training: tuning and alignment

    Naver's reports describe instruction tuning on human-annotated data under strict safety guidelines and, for HyperCLOVA X THINK, supervised fine-tuning and reinforcement learning from verifiable rewards. Its models are evaluated on Korean benchmarks such as KMMLU and the Korean college entrance exam. For search, Naver builds task-specific models on HyperCLOVA X for query analysis, summarisation and document analysis, and it publishes an AI Safety Framework.

    What it means for you: Naver's search models are tuned for the Korean language and Korean questions. Content written natively in Korean, the way Koreans actually phrase their questions, fits the system far better than translated copy.

  3. Layer 3

    Retrieval: live search

    AI answers draw on Naver's own crawl, index and vertical databases, not an outside search engine; Naver says it greatly expanded its web index from August 2024. AI Briefing uses Naver's own content pools, such as Blog, Cafe and Clip, official and public-body sources for factual queries, and Place, Shopping and short-form video data for its other formats. AI Tab combines integrated search with Shopping, Place, Blog, Cafe and visitor reviews. Third-party sites enter through Yeti's crawl and Search Advisor submissions.

    What it means for you: Two pools matter: the web, through Yeti and Search Advisor, and Naver's own platforms. A brand needs both: a crawlable Korean-language website, and an active Naver Blog, an accurate Place listing and well-reviewed Shopping products.

  4. Layer 4

    The answer: what people read

    Informational AI Briefings appear at the top of results with a summary, links to the original creators' content as sources, follow-up questions and, for some queries, personalised recommendations; Place, Shopping and short-form formats appear within results. Naver says AI Briefing aims to show many good sources rather than a single answer. AI snippets summarise individual results, and AI-generated source descriptions sit behind a result's menu. AI Tab is a separate conversational results page, with cards that link to Place, Shopping, maps and reservations.

    What it means for you: AI Briefing credits creators, so Naver Blog posts and Cafe threads that answer a question well can be cited directly. Pair your website with Naver-hosted content that answers the same questions, and keep your Place and Shopping listings complete.

Crawlers

Naver and HyperCLOVA X's crawlers and user agents

The user-agent tokens to look for in your server logs and to name in robots.txt.

Crawlers and fetchers associated with Naver and HyperCLOVA X, with what each does and how it treats robots.txt
User-agent tokenWhat it doesrobots.txt
Yeti
Documentation		Naver's general web search robot, which collects documents for Naver Search: the web pool that AI Briefing and AI Tab draw on. User agent: Mozilla/5.0 (compatible; Yeti/1.1; +https://naver.me/spd), with a Chrome-style variant.Follows robots.txt (RFC 9309). A 5xx response is read as 'disallow all', and it may still fetch favicons, JavaScript and CSS used by allowed pages.
Ads-Naver
Documentation		Collects advertisers' landing pages to measure their quality and match ads to queries. Its user agent includes 'compatible; Ads-Naver/1.1; +https://naver.me/adsn'.Ignores robots.txt for URLs registered as ads: Naver treats registration as permission.
Blueno
Documentation		Builds link previews when users share links in Naver Blog, Cafe and other services, reading Open Graph or Twitter Cards markup first. User agent: facebookexternalhit/1.1 (compatible; Blueno/1.0; +http://naver.me/scrap).Not guaranteed: it acts on users' requests.

Naver documents no crawler for training HyperCLOVA X, and no separate bot for AI Briefing or AI Tab: they draw on what Yeti collects and on Naver's own services.

Crawl sources

Where Naver and HyperCLOVA X crawls from, and how to verify it

Naver publishes its search robots' IP ranges as JSON at searchadvisor.naver.com/doc/naverbot.json and says the list can change without notice. When we checked on 7 October 2026 it was dated 12 July 2024 and listed 36 IPv4 prefixes covering 628 addresses in five blocks, including 110.93.150.x, 114.111.32.x, 125.209.235.x and 211.249.46.x. There is no IPv6, and no network is published.

Naver also documents reverse DNS verification: the hostname must end in .naver.com, such as crawl.125-209-235-169.web.naver.com, and a forward lookup must return the original IP. Blueno's link-preview requests may also resolve to line-apps.com hostnames.

Naver warns that sites blocking its robots by user agent or IP may drop out of search or go stale, so check your firewall and CDN bot rules against the published list.

Opting out

How to opt out of Naver and HyperCLOVA X, or limit what it uses

  1. Set rules for Yeti

    Use User-agent: Yeti groups in robots.txt, with Allow, Disallow, * and $; a Sitemap line is honoured. Serve robots.txt as text/plain with a 200 status, because a 5xx is read as 'disallow all'. Search Advisor has a robots.txt test tool.

    Official page

  2. Hide AI source descriptions

    <meta name="robots" content="nosourceinfo"> stops Naver showing an AI-generated description of your site. It is the only AI-specific control Naver documents.

    Official page

  3. Keep pages out of results

    noindex removes a page from results, and nofollow stops links being collected. Naver warns that pages which only redirect may not have the tag honoured, so tag the target pages too.

    Official page

  4. There is no opt-out from AI Briefing

    Naver documents no publisher opt-out from AI Briefing, AI snippets or AI Tab beyond noindex or blocking Yeti, which also removes pages from ordinary results. Wrong AI snippets can be reported through Naver's correction form.

    Official page

  5. Request collection or removal in Search Advisor

    Search Advisor lets you request collection of individual URLs, which can take from a day to several weeks, suggest a crawl frequency, and request removal of pages that already return 404. It also supports IndexNow.

    Official page

  6. Remove pages in the right order

    Delete pages or add noindex first, confirm they have dropped out, and only then block them in robots.txt. Blocking first stops Naver confirming the removal.

    Official page

  7. Legal and privacy removal

    Naver removes results for legal or policy breaches, or with a court order. Individuals can ask for documents that invade their privacy or defame them to be removed.

    Official page

Naver documents no opt-out from AI training: its privacy policy says personal information is used to develop AI technology, and there is no training crawler to block. For brands, the controls that matter most are the ones that keep Yeti crawling you and your Naver listings accurate.

How we help

How we get brands cited in Naver and HyperCLOVA X

The Coaley Peak GEO Research Centre researches and implements GEO for SEO agencies and in-house teams, with foreign markets as our speciality.

  1. Naver visibility audit

    We search your Korean customers' questions on Naver and record whether AI Briefing and AI Tab mention you, which Blog, Cafe and web sources they cite, and where competitors appear.

  2. Search Advisor and Yeti set-up

    Site verification, sitemaps, full-text RSS, IndexNow, and robots.txt and firewall rules checked against naverbot.json, so Yeti can collect everything you want found.

  3. Naver platform presence

    AI Briefing and AI Tab lean on Naver's own services. With native Korean writers, we help you build an official Naver Blog, a complete Place listing and accurate Shopping products.

  4. Korean content that answers questions

    We write and rework pages in natural Korean, with first-hand detail and cited sources, as Naver's 2026 guidelines ask, so they can be quoted in AI Briefing.

  5. Measurement in Korea

    We track your presence in AI Briefing and AI Tab alongside Google's AI Overviews in Korea, so you see your share of Korean AI answers.

More on our research, services and team at the Coaley Peak GEO Research Centre.

FAQ

Naver and HyperCLOVA X GEO: questions

  • Does Naver still matter in South Korea?

    Yes. StatCounter put Naver at 43.06% of Korean searches in September 2026, just behind Google, and Internet Trend data reported by The Financial News put it at about two-thirds of searches in mid-2026. Naver says its main pages average 50 million daily visitors.

  • What is Naver's AI Briefing?

    Naver's AI answer at the top of its search results, launched in March 2025 and built on its own HyperCLOVA X models. It summarises answers, links to the creators' content it draws on and suggests follow-up questions. Naver says more than 30 million people use it a month.

  • Can I opt out of AI Briefing?

    Not separately. Naver documents no publisher opt-out from AI Briefing: noindex or blocking Yeti removes pages from AI answers and ordinary results alike. The nosourceinfo meta tag only stops AI-generated descriptions of your site.

  • What is Naver's crawler called?

    Yeti. Its user agent includes 'Yeti/1.1; +https://naver.me/spd'. Naver publishes its IP ranges at searchadvisor.naver.com/doc/naverbot.json and documents reverse DNS verification to .naver.com hostnames.

  • Why do Naver Blog and Place matter for AI search?

    Because AI Briefing and AI Tab draw heavily on Naver's own content, including Blog, Cafe, Clip, Place reviews and Shopping, as well as the web. A brand with only its own website is under-represented.

  • Is HyperCLOVA X open source?

    Naver has released open-weight HyperCLOVA X SEED models under its own licences, which allow commercial use but require a separate licence above 10 million monthly users or for competing products. The production models behind AI Briefing are not published.

Want a free chat about Naver and HyperCLOVA X?

Tell us which market and brand you are working on. If we can help, we will set up a free, no-obligation chat with Stephen or Ethan.

Call 0800 494 7725Email us

More engine guides

Related engines

Independence and trademarks. NAVER, 네이버, HyperCLOVA X, CLOVA X, Cue:, AI Briefing and Naver Search Advisor are trademarks or brand names of NAVER Corporation. Coaley Peak is independent: we are not associated with, endorsed by or a partner of NAVER Corporation. This guide is our own research from public documentation and reputable reporting, checked in October 2026.

Reviewed by Stephen Grindley, research centre lead. Last checked .

Document reference: ISO_webpage_geo_engine_naver_v1

Last modified: 7 October 2026

GEO Research Centre·Engine guides·Naver and HyperCLOVA X