Naver vs Google: Reading Korea Search Share

Ask who leads search in South Korea and you will get two confident answers, both sourced, both current, and pointing in opposite directions. InternetTrend puts Naver at 62.86% for full-year 2025 against Google at 29.55%. Statcounter puts Google at 49.54% against Naver at 40.90% for July 2026. Same country, same period, roughly a 30-point disagreement about who is winning.

Neither is lying. They measure different things, and the difference is the single most useful thing to understand before spending money in this market. Most agency pitch decks quietly pick whichever figure supports the story they are already telling. That is the part worth catching.

This article explains what each tracker counts, why the two diverge so sharply, which one to use for which decision, and what happened between January and July 2026 that changed the trend line on one of them. If you are building a Korea plan and someone has handed you a single market share number, this is the article to read before you accept it.

What the two trackers actually count

Start with the mechanics, because everything else follows from them.

InternetTrend analyses visit and query logs from major domestic Korean websites. It is a Korean tool built to answer a Korean question: when Korean people search, where do those searches happen? Korean media treat it as the reference series. Yonhap, Korea Times, Seoul Economic Daily and Financial News all cite it when reporting on this market, which is itself a signal about how the domestic industry reads its own numbers.

Statcounter Global Stats works from tracking code embedded across a large partner network of websites and counts page views by referring search engine. It is global, consistent across countries, free to inspect, and therefore the default source for any international comparison. It is also structurally biased toward whatever browsers and operating systems dominate general web browsing, which in practice means Chrome and Android weigh heavily.

That last sentence is the whole story. Consider what a page-view panel can and cannot see. If a Korean user opens the Naver app, searches for a restaurant, reads Naver Place reviews, and books a table without ever leaving Naver, how many measured page views on third-party partner sites did that generate? Close to none. The search happened. The commercial intent was satisfied. And a page-view panel barely registered it.

Now run the same journey on Google. Query, click, land on an external website that may well carry a partner tracking script, page view counted, referrer recorded as Google. The behaviour that Google drives is exactly the behaviour a page-view panel is built to observe. The behaviour Naver drives is largely invisible to it.

This is not a criticism of Statcounter. It measures what it says it measures. The error is in reading it as a proxy for Korean search intent when it is closer to a proxy for referral traffic reaching the open web.

The number that resolves most of the contradiction

Here is the detail almost nobody quotes, and it comes from Statcounter itself.

Cut Statcounter's own Korean data to mobile and tablet only, for December 2025, and Naver sits at 60.30% against Google at 36.61%, with Bing at 0.21%. That is not a small adjustment. It flips the leader inside the same dataset, using the same methodology, changing only the device filter.

So a large share of the apparent conflict between the two trackers is a device story rather than a methodology war. Desktop browsing in Korea leans toward Chrome and toward Google. Mobile search behaviour, which is where most Korean commercial discovery happens, leans heavily toward Naver. Aggregate the two and you get a headline that describes neither.

What follows for planning is straightforward. If your customers reach you on a phone, and in Korea most consumer categories do, then the mobile cut is the relevant series and Naver is your primary surface. If you sell into a desktop-heavy professional workflow, developer tooling, enterprise software, technical research, then Google carries real weight and may carry most of it. This is a segmentation question, not a market share question, and it is answerable from your own analytics rather than from anyone's panel.

What changed between January and July 2026

Something moved in 2026, and it moved on the domestic tracker rather than the global one.

InternetTrend recorded Naver averaging 63.82% from 1 January to 26 April 2026. From 27 April to 17 June, after Naver launched the beta of its conversational AI Tab, that average rose to 66.34%, a gain of roughly 2.5 percentage points. On 24 May the daily figure touched 81.34%. Across roughly six weeks following the AI Tab beta, Naver's daily share exceeded 70% on fourteen separate days.

Widen the lens to half-year averages and the pattern holds. Naver took 64.28% for January to June 2026, up 2.46 points from 61.82% in the second half of 2025. Google slipped to 28.37%, down 2.32 points. The gap between them widened from 31.13 points to 35.91 points inside six months.

Korean reporting attributes the move to Naver's AI search rollout. That attribution is reported rather than proven, and worth holding loosely, since correlation over a six-week window is not causation. But the direction is not ambiguous, and the timing is hard to ignore.

Period and source	Naver	Google	What it tells you
2014, Nielsen KoreanClick	76.69%	2.06%	Historical baseline. Naver's dominance was once near total
Full-year 2024, InternetTrend	58.14%	not stated in the same report	The low point of the recent decline
Full-year 2025, InternetTrend	62.86%	29.55%	Back above 60% for the first time since 2022. Bing 3.12%, Daum 2.94%
December 2025, Statcounter mobile only	60.30%	36.61%	The device cut that explains most of the tracker gap
March 2026, InternetTrend	63.8%	28.7%	Naver had briefly passed 70% on 28 February and 1 March
H1 2026, InternetTrend	64.28%	28.37%	Gap widened from 31.13 to 35.91 points versus H2 2025
July 2026, Statcounter all platforms	40.90%	49.54%	Bing 5.44%, CocCoc 1.35%, Yandex 1.22%, Daum 0.95%

Read the table as two parallel series rather than one timeline. Comparing an InternetTrend figure to a Statcounter figure produces a number with no meaning. Comparing InternetTrend to InternetTrend, or Statcounter mobile to Statcounter mobile, produces a trend you can act on.

The third measurement almost nobody quotes

There is a series that neither tracker covers, and it produced the most striking single data point of 2026.

On 11 August 2026, Korean reporting citing Mobile Index data from IGAWorks put the Google app at 47.02 million monthly active users against the Naver app at 46.84 million. Google's app had overtaken Naver's app for the first time since the tracker began measuring in March 2021.

Sit with how oddly that fits alongside everything above. On InternetTrend, Naver's lead over Google widened through the first half of 2026. On app-level MAU, Google pulled ahead in the same period. Both can be true, and the reason they can be true is that they measure different objects. Opening an app is not the same act as running a search. The Google app on an Android phone is a default surface reached constantly and often incidentally, including for things that are not search at all. Naver app sessions tend to be deliberate and, when they are searches, they tend to be commercial.

Which is a useful corrective in both directions. If you were about to conclude that Google is losing Korea, app MAU says otherwise. If you were about to conclude that app MAU means Google now owns Korean search intent, InternetTrend says otherwise. Three series, three different questions, and a plan that only reads one of them is a plan with a blind spot.

One more figure worth having, since it shows how volatile the daily numbers are. Before the AI Tab period, Naver had already briefly passed 70% on 28 February and 1 March 2026, reaching 70.6% and 70.4%. So the 81.34% peak in May was an extreme rather than an anomaly of a kind that had never occurred. Daily figures in this market swing far more than annual averages suggest, which is another argument for never quoting a single day as evidence of anything.

Naver disputes the global figure, and that matters less than it sounds

Naver has publicly pushed back on the Statcounter numbers. You would expect that, and it does not settle anything either way. A platform is not a neutral referee on its own market share.

What is more useful is noticing that both series can be true at once without contradiction. Naver can hold a commanding share of Korean search intent, most of it on mobile, most of it resolved inside its own properties, while Google simultaneously accounts for a larger share of the referral traffic that actually lands on external websites. Those are two different business realities and both have budget implications.

If anything, holding both figures in view is more useful than resolving them. A brand that is invisible on Naver has lost the domestic discovery layer. A brand that is invisible on Google has lost the technical, academic, English-language and international-buyer layer, plus a substantial share of its own website referral traffic. Neither gap is survivable in a serious Korea plan.

What about Bing and Daum?

Bing sits between roughly 3% and 5.5% depending on which series you read, InternetTrend at 3.12% for 2025 and Statcounter at 5.44% for July 2026. Small, but not zero, and worth noting because Bing feeds Copilot. If your category has meaningful enterprise desktop usage, a Bing check costs almost nothing.

Daum is the more interesting case, and the answer is that it has essentially stopped being a search decision. The measurements do not agree on the exact number, which is itself instructive: InternetTrend 2.94% for 2025, Statcounter 0.95% for July 2026, Statista 1.39% for 2024, and Korea Herald reporting 3.29% for August 2024, down from 4.41% a year earlier. Four sources, four methods, four windows. What they agree on is the direction and the order of magnitude. Below 3%, falling.

Kakao still matters enormously in Korea. Just not through Daum as a general search engine. It matters through KakaoTalk, its channels and its advertising ecosystem, which is a different workstream with different mechanics.

How to use this in a pitch, a plan or a board paper

Four habits, and they are cheap to adopt.

Name the source and the window every time. Not "Naver has 64% of Korean search" but "InternetTrend recorded Naver at 64.28% for the first half of 2026." The second version survives being challenged. The first one does not.

Never average across methods. The temptation is real, because a single number looks decisive in a slide. Resist it. An average of 64.28% and 40.90% is 52.59%, and there is no population anywhere in Korea for which that figure is true.

Pick one series per trend line. Switching sources mid-chart is how a flat market becomes a growth story or a decline. If your baseline is InternetTrend, stay on InternetTrend.

Treat your own data as the tiebreaker. Panels describe the market. Your Search Console, Naver Search Advisor and Naver Analytics describe your business. Where they disagree, yours wins for budgeting purposes, because it is a census of your traffic rather than a sample of someone else's.

What this does not tell you

A market share figure is an input, not a strategy, and there are things it genuinely cannot answer.

It cannot tell you how the language split works. We looked specifically for credible public data quantifying how much of Korean querying happens in Korean versus English, and how query syntax differs between the two engines. It does not exist in any form we would put in front of a client. The qualitative direction is consistent across sources, Google skewing toward technical, academic and English-language queries and Naver toward conversational, review-driven, everyday commercial ones. The magnitude is unmeasured, so we report it as unmeasured.

It also cannot tell you what share of product discovery happens inside apps rather than in general web search, in Coupang, YouTube or KakaoTalk. Multiple sources flag that this share is significant. None of them quantify it, and the ones that appear to are usually citing marketing blogs without methodology.

And it says nothing about the harder question, which is how much of a Naver results page you can realistically occupy. That depends on Smart Block composition per query, on whether an AI Briefing appears above the links, and on which Naver properties your content lives in. Share of engine is not share of page.

For that last question, and for how Naver's AI answer layer changes the click behaviour underneath it, see our SEO service page for South Korea. Where the question becomes visibility inside generative answers rather than inside search results, it moves over to our GEO service for South Korea.

---

Frequently Asked Questions

Which Korean search market share figure should I use?

It depends on the decision. For where Korean search intent happens, use InternetTrend, which recorded Naver at 62.86% for full-year 2025 and 64.28% for the first half of 2026, cross-checked against Statcounter's mobile-only cut showing Naver at 60.30% for December 2025. For international comparison or for understanding referral traffic reaching external websites, use Statcounter, which recorded Google at 49.54% against Naver at 40.90% for July 2026. Name the source and the period whenever you quote either, and never average the two.

Why do InternetTrend and Statcounter disagree so much about Korea?

Because they count different populations. InternetTrend analyses visit and query logs from major domestic Korean sites, so it captures searches that begin and end inside Naver. Statcounter counts page views across a partner network of websites, which weights heavily toward Chrome and Android browsing and barely registers a journey that starts in the Naver app and finishes with a Naver Place booking. Statcounter's own mobile-only cut for December 2025 shows Naver at 60.30% against Google at 36.61%, which suggests much of the gap is a device effect rather than a methodological dispute.

Is Naver's share in Korea rising or falling?

Rising on the domestic tracker. InternetTrend recorded Naver at 58.14% for 2024, 62.86% for 2025, and 64.28% for the first half of 2026, while Google slipped to 28.37%. The gap between them widened from 31.13 points in the second half of 2025 to 35.91 points in the first half of 2026. Naver averaged 63.82% before the April 2026 beta of its conversational AI Tab and 66.34% afterwards, touching 81.34% on 24 May 2026. Korean reporting attributes the move to the AI search rollout, which is a reported attribution rather than a proven causal finding.

Should I still optimise for Daum?

Only as a low-cost test, and only if your own data supports it. Daum sits below 3% on every 2024 to 2026 measurement we reviewed, with InternetTrend at 2.94% for 2025, Statcounter at 0.95% for July 2026, and Statista at 1.39% for 2024. The sources disagree on the exact figure but agree on the direction. Kakao remains commercially important in Korea, through KakaoTalk, its channels and its advertising ecosystem rather than through Daum as a general search engine.

Does Bing matter in South Korea?

Marginally, and mainly where enterprise desktop usage is significant in your category. InternetTrend put Bing at 3.12% for 2025 and Statcounter at 5.44% for July 2026. The reason to check rather than ignore it is that Bing feeds Microsoft Copilot, so a small search share can carry disproportionate weight for AI-assisted discovery among professional users. It does not justify a separate workstream on its own.

Can I quote a single Korean market share number in a client deck?

You can, but you should quote it as a sourced figure with its window and its method attached rather than as the market share of Korean search. An unattributed single number invites exactly one question from anyone who knows this market, which is which tracker it came from. Presenting both series with their methods stated takes one extra line and makes the analysis harder to dismiss.

What data is missing that I should stop asking for?

Two things, and no credible source has them. First, the language split: there is no reliable public data quantifying how much Korean querying happens in Korean versus English or how syntax differs between the engines, only a consistent qualitative direction. Second, the share of product discovery happening inside apps such as Coupang, YouTube or KakaoTalk rather than in general web search. Figures circulate for both, and the ones we traced led back to marketing blogs without stated methodology.

---