A Unified AI Search Visibility Framework for Korea: Measuring Naver, Kakao, Google, and ChatGPT Together
A unified framework for measuring AI search visibility in Korea that puts Naver AI Briefing and AI Tab, Kakao's in-KakaoTalk AI, Google AI Overviews and AI Mode, and ChatGPT in one table. Covers per-surface metrics, collection paths, weighted unified share, tool coverage, and a rollout plan.
"We get a ChatGPT report every month. How does our brand show up in Naver AI Briefing and in KakaoTalk's AI?" Most AI visibility audits for Korean brands stall at that question. Global tools track ChatGPT and Google AI Overviews, but Naver is missing from their engine lists, and none of the public tool documentation reviewed for this article lists collecting AI answers inside KakaoTalk. This article defines a unified AI search visibility measurement framework for Korea that covers Naver, Kakao, Google, and ChatGPT at once, and lays out as a checklist what to collect on each surface, through which path, and how to roll it all into one number.
30-second definitions: the terms used in unified measurement
- AI search surface is the screen where a user sees an AI-generated answer. This article covers four: Naver (AI Briefing, AI Tab), Kakao (AI inside KakaoTalk), Google (AI Overviews, AI Mode), and ChatGPT.
- Answer rate is the share of questions in the measurement set for which the surface actually displayed an AI answer.
- Brand mention rate is the share of questions with a displayed AI answer in which your brand name appears in the answer text.
- Source citation rate is the share of questions with a displayed AI answer in which your domain or a document from your own channels is attached as a source link.
- Unified share is the weighted average of per-surface mention rates (or citation rates), using each surface's reach as the weight.
You calculate the first four metrics separately for each surface. Only the fifth spans surfaces. Reverse that order and the surface with the most collected data dictates the result.
Why measure four surfaces separately and combine them later
The four surfaces generate AI answers in different places, so no single collector can scrape them all in one pass.
Naver AI Briefing is not a separate chat app. It appears inside integrated search results on both PC and mobile. For direct-answer queries and queries that call for exploring several pieces of content it sits at the top; for other queries its position varies with search intent.[1] As of late 2025 it appeared on 20% of integrated search queries, and Naver said it would roughly double that coverage by the end of 2026 and extend it to shopping and local (Byline Network, 2026).[2] At 20% coverage, the remaining queries show no Briefing at all, so if you skip counting the answer rate, your mention-rate denominator is wrong.
Kakao said it would shut down its standalone AI app Kanana on October 15, 2026 and move that technology into AI services inside KakaoTalk, such as "Kanana in KakaoTalk" and "ChatGPT for Kakao" (MTN, 2026).[7] What you are measuring is a screen inside a messenger app, not a web page. Web crawlers and API prompts cannot retrieve these answers.
Google AI Overviews and AI Mode are generated on the search results page. Google does, however, offer a generative AI performance report in Search Console that counts how often your URLs appeared in AI Overviews and AI Mode, by page, country, device, and date (Google Search Central, 2026).[9] Of the four surfaces, Google is the only platform that hands you first-party exposure data.
ChatGPT works by sending a prompt and receiving a response, which makes it the surface global tools measure best. Ahrefs Brand Radar lists AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot, and Claude as tracked engines. Naver is not on the list.[11]
Ignore these differences and label one tool's number "AI visibility," and the Naver and Kakao surfaces the report never captured get counted as zero. That is why the framework's first rule is collect each surface separately, combine only at the last step.
Evidence: reach of the four surfaces (2026)
To set per-surface weights, start with how many users each surface has. The public figures are below.
| Item | Value | Source |
|---|---|---|
| Google app, July 2026 | 47.02 million | (Korea JoongAng Daily, 2026)[5] |
| Naver app, July 2026 | 46.85 million | (Korea JoongAng Daily, 2026)[5] |
| ChatGPT app, October 2026 | 18.76 million | (New Daily, 2026)[6] |
The Google and Naver app figures come from IGAWorks Mobile Index's July 2026 count (Korea JoongAng Daily, 2026).[5] The ChatGPT app figure is from October 2026, up 43.8% year over year (New Daily, 2026).[6] Because the counts come from different months, read the bar lengths only as a rough sense of scale.
Separate from app user counts, there are also public figures for the AI features themselves.
| Surface | Public figure | As of | Source, year |
|---|---|---|---|
| Naver search share | 64.39% (Google 28.54%) | January 1 to March 7, 2026 | CEO Score Daily, 2026[3] |
| Naver AI Briefing | Applied to 20% of integrated search queries | Late 2025 | Byline Network, 2026[2] |
| Naver AI Briefing users | 30 million+ per month | July 2026 | Digital Today, 2026[4] |
| Naver AI Tab users | Passed 10 million | July 15, 2026 | Digital Today, 2026[4] |
| Kakao ChatGPT for Kakao | About 13 million cumulative sign-ups | End of June 2026 | New Daily, 2026[6] |
| Google Search Console generative AI report | Expanded to all sites | August 31, 2026 | Google Search Central, 2026[9] |
Read together, the numbers point to a clear conclusion. Naver leads in search share, Google and Naver are close on app users, and ChatGPT has less than half the app users of either portal but grew 43.8% in a year.[3][5][6] Measuring any one surface alone cannot explain how Korean users encounter AI answers. Daum also announced an AI Overview expansion and a conversational AI mode for July 2026 (Financial News, 2026), so if Daum carries real weight in your category, add it as a fifth surface.[8]
Per-surface measurement design: what to collect and how
The table below is the core of the framework. Rows are surfaces; columns cover collection path, metrics, official platform data, and limitations.
| Surface | Where the answer appears | Collection path | Primary metrics | Official platform data | Measurement limits |
|---|---|---|---|---|---|
| Naver AI Briefing | Inside integrated search results, at the top or elsewhere depending on query type[1] | Capture result pages for Korean queries, PC and mobile separately | Briefing answer rate, mention rate, citation rate | No public report found | Answer rate shifts as coverage changes[2] |
| Naver AI Tab | Conversational screen inside the search box[4] | Run fixed questions conversationally and log responses | Mention rate, recommendation order | No public report found | Answers diverge depending on follow-up questions |
| Kakao AI | Inside the KakaoTalk app[7] | Run fixed questions manually in KakaoTalk | Mention rate | No public report found | No automated collection found in the public tool documentation reviewed |
| Google AI Overviews, AI Mode | Search results page | Capture search results, cross-check with Search Console | Answer rate, mention rate, citation rate | Generative AI impressions[9] | Search Console does not show queries or competitors |
| ChatGPT | Conversational response | Run prompts repeatedly and collect responses | Mention rate, citation rate, share | No public report found; GA4 channel shows referrals only[10] | Answers vary by run, so repetition is required |
Three design points matter most.
Use one fixed question set. Cross-surface comparison only holds if all four surfaces get questions with the same intent. The phrasing changes, though. Naver and Google get search-style queries such as "AI visibility measurement tool recommendations," while ChatGPT and KakaoTalk get conversational prompts such as "Recommend some AI visibility measurement tools available in Korea." Tag each question with an intent ID and the different phrasings roll up into the same row.
Set the denominator per surface. For Naver and Google, the denominator is only the questions that triggered an AI answer. ChatGPT and KakaoTalk AI are conversational and return a response to every question, so use all questions run as the denominator, and flag separately any cases with no response or a refusal.
Treat referral traffic as a secondary metric. GA4's AI Assistant channel separates ChatGPT, Gemini, and Claude referrals, but clicks from Google AI Overviews and AI Mode are counted as organic search (MADX Digital, 2026).[10] Clicks through Naver AI Briefing likewise blend into general Naver traffic. Back into per-surface visibility from traffic numbers alone and Google and Naver come out near zero. Channel setup is covered in measuring the GA4 AI Assistant channel, and reading the Search Console report is covered in the Search Console generative AI performance report.
How to calculate unified share
Once you have per-surface mention rates, combine them with a weighted average.
Unified share = Σ (per-surface mention rate × surface weight) / Σ surface weights
Set the weights by how much your own customers actually use each surface. GA4 referral mix and customer survey answers about channel usage are the first-choice inputs. Without first-party data, use the public figures above as a temporary baseline, and state the weights and their basis in the report.
Here is a worked example. Say Naver's mention rate is 20%, Kakao's 10%, Google's 30%, and ChatGPT's 40%, with weights of Naver 4, Kakao 1, Google 3, and ChatGPT 2. The unified share is (80 + 10 + 90 + 80) / 10 = 26%. Report the same data using ChatGPT alone and you get 40%; a simple average gives 25%. These are hypothetical values meant only to show the calculation.
Change the weights and the unified share changes with them. So never report the unified share by itself; put the per-surface mention rate table on the same page. For an executive dashboard layout, see the AI visibility dashboard executive metrics template.
Surface coverage by measurement tool
Tools can automate part of Layer 1 collection. The table below reflects only what each vendor states in its public documentation. "Not listed" does not mean the feature is absent; it means the public pages did not confirm it. Prices are public list prices and subject to change.
| Tool (operator) | Type | Tracked engines per public docs | Naver | Kakao | Google AI Overviews, AI Mode | ChatGPT | Price band |
|---|---|---|---|---|---|---|---|
| Brand Radar (Ahrefs) | Global, SEO suite | AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot, Claude[11] | Not listed | Not listed | Listed | Listed | Public plans (subject to change) |
| SOHA (LeadGen Lab) | Korean, analytics platform | ChatGPT, Claude, Perplexity, Gemini, Google AI Overview, Naver AI[12] | Listed (Naver AI) | Not listed | Listed (AI Overview) | Listed | Inquiry |
| OPTIANALYTICS (Next-T) | Korean, crawl signal analytics | ChatGPT, Perplexity, Claude, Gemini crawl signals plus Google and Naver crawl flows[13] | Listed (crawl flow) | Not listed | Not listed (crawl flow only) | Listed (crawl signal) | Inquiry |
| BVI (BOIDA) | Korean, AI brand visibility measurement | ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek[14] | Not listed | Not listed | Not listed | Listed | Inquiry |
| Search Console (Google) | Official platform | Your URLs' AI Overviews and AI Mode impressions[9] | N/A | N/A | Listed (own impressions) | N/A | Free |
| GA4 (Google) | Official platform | ChatGPT, Gemini, Claude referral channel[10] | N/A | N/A | Mixed into organic | Listed (referrals) | Free |
What separates these tools is how they collect. Global tools publish their engine lists and are strong at comparing ChatGPT with Google's AI surfaces, but Naver is not on those lists. Korean tools position themselves as covering Naver, yet SOHA describes "Naver AI tracking" while OPTIANALYTICS describes Naver crawl flows, so the same "Naver support" label refers to different collection targets.[12][13] LeadGen Lab positions SOHA as linking measurement to execution and covering Korea's AI ecosystem.[12] BOIDA's BVI (Brand Visibility Index), built by operator Designovel, lists tracking of six global engines on its official page; per that public page, Naver and Kakao are not on its engine list.[14] None of the public documentation for the tools in the table lists collection of AI answers inside KakaoTalk.
So when choosing a tool, check the collection method before the price. Ask three questions in the demo: Does it capture Naver integrated search result pages directly? Does it record separately whether an AI answer appeared? Does it store the cited source URLs? Use those three answers to fill in the "Not listed" cells. For a broader comparison of Korean SaaS tools, see the Korea AI visibility measurement tool comparison; for a deep dive on Naver alone, see Naver AI Briefing visibility measurement methods and tools.
Rollout: building the unified measurement system in four weeks
| Week | Task | Deliverable | Checkpoint |
|---|---|---|---|
| Week 1 | Write an intent-based question set (informational, comparison, branded) in two versions: search-style and conversational | Question sheet with intent IDs | Is every intent mapped to all four surfaces? |
| Week 1 | Check the Search Console generative AI report and the GA4 AI Assistant channel | Layer 2 and Layer 3 baselines | Does a list of pages with AI Overviews exposure come up? |
| Week 2 | Capture Naver and Google result pages (logged out, PC and mobile separately) | Logs of answer presence, mentions, cited URLs | Did you calculate the answer rate first? |
| Week 2 | Run ChatGPT prompts repeatedly; run KakaoTalk AI manually | Per-run response logs | Did you rerun the same questions on different days? |
| Week 3 | Aggregate per-surface mention and citation rates, then set weights | Per-surface metrics table, weighting rationale memo | Did you document the basis for the weights? |
| Week 4 | Calculate unified share; measure 2 to 3 competitors the same way | Draft unified dashboard | Are you reporting the per-surface table alongside the unified number? |
Without repeated runs, you cannot tell whether next month's change is real or just sampling noise. For how to set question counts and run counts, see prompt sample size for AI visibility scores; the principles of repeated per-engine measurement are in the multi-engine measurement methodology.
Once measurement is done, diagnose the lowest-scoring surface first. If Naver is low, review the blog, Cafe, Place, and Smart Store information that AI Briefing draws on. If Google is low, find the key pages with zero impressions in Search Console. If ChatGPT is low, gather the external sources cited in its answers and identify the documents where your brand is missing. For how Naver and Kakao have shifted their AI strategies, see Naver and Kakao's shift to platform-embedded AI; for the background on Google overtaking Naver in app MAU, see Google overtakes Naver among Korean mobile app users.
Wrap-up
AI search visibility in Korea cannot start from a single number. Naver AI Briefing and AI Tab, AI inside KakaoTalk, Google AI Overviews and AI Mode, and ChatGPT each generate answers in a different place and need a different collection path. Measure answer rate, mention rate, and citation rate per surface, supplement them with platform data from Search Console and GA4, then weight by your customers' reach to get the unified share. Follow that order and instead of a one-line report saying "we do well on ChatGPT," you get a table showing which surface is missing how many percentage points. For the full picture of the global and Korean GEO market, see the 2026 global GEO and AEO landscape; for the fundamentals, see What is GEO?.
Related companies
- 넥스트티 (Next-T, OPTIGEO)SEO, GEO, AEO 컨설팅, 자동화
- 디자이노블 (Designovel, BOIDA)AI 패션 테크, 생성형 AI, GEO
- 리드젠랩 (LeadGenLab)AI 가시성 최적화 에이전시
- 보이다 (BOIDA)생성형 검색 최적화(GEO) 솔루션, AI 가시성 측정
- AhrefsSEO 툴킷, AI 가시성
Frequently asked questions
- Yes, but only after you calculate each surface separately. Surfaces differ in how often they show an AI answer and in how you collect it, so first compute the answer rate and mention rate per surface, then take a weighted average based on how much of your customer reach each surface accounts for. A plain sum skews the result toward whichever surface produced the most data.
- Start with your own data: GA4 referral mix and customer survey answers about which channels people use. Without first-party data, use public user counts as a temporary baseline (Google app 47.02 million, Naver app 46.85 million, ChatGPT app 18.76 million, IGAWorks Mobile Index data as reported by Korea JoongAng Daily and New Daily, 2026), and record the weights and their basis in the report.
- Manually, for now. Kakao shut down its standalone Kanana app on October 15, 2026 and moved those features into AI services inside KakaoTalk (MTN, 2026). None of the public tool documentation reviewed for this article lists collection of AI answers inside KakaoTalk, so run a fixed question set directly in KakaoTalk and log the results.
- Partly. The generative AI performance report in Google Search Console shows how often your URLs appeared in AI Overviews and AI Mode, broken down by page, country, device, and date, and it was expanded to all sites as of August 31, 2026 (Google Search Central, 2026). The report does not tell you which queries triggered the exposure or how competitors performed.
- Yes. GA4's AI Assistant channel groups ChatGPT, Gemini, and Claude referrals into their own channel. Clicks from Google AI Overviews and AI Mode still count as organic search (MADX Digital, 2026). Among the public materials reviewed for this article, there was also no official report that separates clicks coming through Naver AI Briefing.
- Fix the question set and the run environment, then repeat on the same cadence. Generative answers change from run to run, so never use a single run as a baseline. Collecting every surface in the same week makes cross-surface comparison possible.
Q.Can Naver, Kakao, Google, and ChatGPT visibility be combined into a single number?
Q.How should the unified share weights be set?
Q.How do you measure Kakao's AI answers?
Q.Can you see Google AI Overviews exposure without a third-party tool?
Q.Can GA4 isolate ChatGPT traffic?
Q.How often should you measure?
Sources
- [1] ↑네이버, AI 기반 검색 결과 요약 'AI 브리핑' 27일 출시 — 아이뉴스24
- [2] ↑네이버의 2026년 전략: 검색에서 실행으로, 커머스 확대 본격화 — 바이라인네트워크
- [3] ↑네이버, AI 브리핑 통했다, 8년 만에 검색 점유율 최고 — CEO스코어데일리
- [4] ↑네이버, 'AI탭' 이용자 1000만명 돌파 — 디지털투데이
- [5] ↑Google overtakes Naver in mobile app users in July for first time — Korea JoongAng Daily
- [6] ↑챗GPT와 맞붙기보다 품는다, 네카오 AI 전략 전환 주효할까 — 뉴데일리
- [7] ↑카카오, 카나나 앱 서비스 10월 종료, AI 별도 앱 전략 폐기 — MTN 머니투데이방송
- [8] ↑검색 엔진 아닌 답변 엔진 전쟁, 네이버, 다음, 7월에 확 바뀐다 — 파이낸셜뉴스
- [9] ↑Introducing Search Generative AI performance reports in Search Console — Google Search Central Blog
- [10] ↑GA4 launches AI Assistant Channel — MADX Digital
- [11] ↑Brand Radar — Ahrefs
- [12] ↑GEO AEO, AI 가시성 최적화 전문 기업 리드젠랩 — 리드젠랩
- [13] ↑OPTIANALYTICS, GEO 애널리틱스 AI 수집, 참조 신호 측정 도구 — 넥스트티
- [14] ↑BOIDA, AI 브랜드 가시성 측정 — Designovel
Related documents
- How to Measure Naver AI Briefing Visibility: Methods, Tools, and Why Global AI Visibility Tools Miss NaverHow to check whether Naver AI Briefing cites your brand. This page covers why Naver is missing from the engine lists of global AI visibility tools, how Korean tools compare, a manual measurement routine, metric definitions that start with the briefing trigger rate, and the questions to ask before picking a tool.
- Multi-Engine Measurement: How to Measure Visibility Across ChatGPT, Gemini, Perplexity, and ClaudeWhy every engine answers differently, the trap of single-engine measurement, and a multi-engine GEO methodology for measuring AI visibility through prompt sets, repetition, and share of voice.
- Naver and Kakao AI Strategy Shift 2026: AI Briefing, AI Tab, the Kanana Shutdown, and What Brands Should DoNaver shut down CLOVA X and Cue and built AI Briefing and AI Tab into search. Kakao closed the Kanana app and moved its AI into KakaoTalk. This page covers both companies' timelines, the published numbers, and the exposure points brands need to check.
- Google's App Overtakes Naver in South Korea: How AI Search Rewired the Portal MapIn July 2026, Google's app MAU hit 47.02 million, edging past Naver's 46.85 million for the first time on record. We trace how Gemini's expansion closed an 8-million-user gap in five years and what the shift means for Korea's search landscape.
- Google Search Console Generative AI Performance Reports: A Complete Guide (2026)On June 3, 2026, Google added Generative AI performance reports to Search Console. They are the first official tool to measure URL impressions inside AI Overviews, AI Mode, and Discover: here we break down the measurement limits, the opt-out toggle, and the AEO strategy.
- GA4 AI Assistant Channel: Measuring ChatGPT, Gemini, and Claude TrafficGA4's AI Assistant channel: added to Default Channel Groups in May 2026, automatically classifies AI referral traffic, yet a structural blind spot remains: sessions that arrive without a referrer header. Here's what the channel captures, what it misses, and how to cover the gap.
- AI Visibility Dashboard Template for Executive Reporting: 7 Metrics to Include, the Ones to Drop, and How to Show ConfidenceThe 7 metrics an executive AI visibility dashboard needs (mention rate, citation rate, SoV, portrayal accuracy, recommendation strength, AI referral sessions, Google AI impressions), the metrics to drop, how to show confidence, and a one-page template built on the IAB standard.
- How Many Prompts and Runs Make an AI Visibility Score Trustworthy? Sample Size and Repeat-Run GuidelinesHow many prompts, run how many times, before you can trust an AI visibility score. Covers confidence intervals by response count, the responses needed to compare two periods, the design effect of repeated runs, public experiments from SparkToro, Evertune and others, and a five-step sampling procedure.
- Korea AI Visibility Tools Compared: SOHA, OPTIGEO, ChainShift, RanketAI 2026A side-by-side comparison of SOHA, OPTIGEO, ChainShift, RanketAI, and BOIDA (BVI) across engine count, diagnostic depth, domestic platform support, and execution integration: with concrete selection criteria for Korea B2B SaaS environments.
- Daum AI Overview vs. Naver AI Briefing: Korea's AI Search Landscape in 2026A practical comparison of Daum AI Overview (Upstage Solar) and Naver AI Briefing: covering architecture, citation logic, market share (64.39% vs. 2.72%), and platform-specific AEO strategies.
- Global GEO/AEO Player Landscape 2026, Monitoring Tools, Agencies, and PlatformsA 2026 landscape that sorts GEO/AEO players into monitoring tools, specialist solutions and agencies, enterprise platforms, and regional players. We compare the leading vendor in each category, founding, headquarters, tracked engines, pricing, and differentiation, against primary sources.
- What Is GEO: The Definition of Generative Engine Optimization and How It Differs From SEOGEO (Generative Engine Optimization) is the strategy of getting your content cited in answers produced by generative engines like ChatGPT and Perplexity. Here is the definition, how it differs from SEO, and how it works.