A Ranking That Hadn't Moved in Five Years Just Flipped

Here's the deal: in July, Google's app recorded 47,022,854 monthly active users in Korea. Naver recorded 46,845,017. The gap is 177,837 users. It's the first time Google has been ahead since IGAWorks' Mobile Index began tracking this in March 2021.

In the overall app ranking, Google moved to third behind YouTube and KakaoTalk, and Naver dropped to fourth. "Naver isn't number one" is a sentence Korean internet watchers have essentially never had to write.

The speed of the swing is the striking part. In May, Naver led by about 400,000. In June the gap narrowed to 250,000. In July it reversed. Nearly 600,000 users of movement in two months is unusual at this scale.

The Seat Naver Held for Twenty Years

To weigh this properly, you need to know what Naver was defending.

Korea is effectively the only open market where Google never won search. China is a withdrawal, Russia has Yandex operating under policy protection, and Japan's Yahoo! Japan runs on Google's engine anyway. Korea is close to the only country that held Google off through straight competition.

Naver did it with content, not ranking algorithms. Knowledge iN, launched in 2002, was decisive. At the time there simply wasn't enough Korean-language web text — crawling perfectly still left most questions unanswered. So Naver stopped crawling for answers and asked people instead, then turned those Q&As into search results. Blogs and Cafés extended the same model.

The key detail: that content lived only inside Naver. However good Google's search was, it could not index Naver Café posts. If you can't reach the document containing the answer, you can't improve the answer. That was the moat for twenty years.

AI summaries route around it. When answering shifts from "send you to the document with the answer" to "read several documents and compose the answer," walling content inside one platform buys less. And because a model's general knowledge alone answers a large share of questions, plenty of queries no longer need Korean web documents at all.

What Drove the Swing

The Google app isn't only search. It bundles search, the Discover feed, Lens, and more recently AI Overviews and AI Mode — summary-on-top-of-results and conversational handling of complex, follow-up questions respectively. Those features are converting it from a search tool into a question tool. As people move from typing keywords to asking full sentences, Google's long-standing weakness — thin Korean-language web corpus — matters relatively less.

Gemini is a separate app but part of the same current. Its Korean MAU reached 9.47 million in July, second among AI chatbots behind ChatGPT's 23.67 million, per WiseApp·Retail figures published on August 11. That's a record high for Gemini in Korea since launch.

Globally, Gemini crossed 1 billion MAU in the same window — announced by Google on August 11 and described as the fastest growth in the company's history. It was at 400 million in May 2025, so it roughly two-and-a-half-x'd in fifteen months.

The Google app's own trajectory matters too. It first passed 40 million Korean MAU in May 2025 and crossed 46 million in May 2026 — six million added in a year, overlapping precisely with when AI features started landing.

Naver is flat. Its MAU didn't collapse; it barely moved while Google climbed past. Naver has been expanding AI Briefing and deepening HyperCLOVA X, and it hasn't translated into user growth.

Flat is arguably the harder position. When users leave you can diagnose why and plug the leak. When your users stay and the competitor grows past you, there's no obvious lever — especially with Korean smartphone penetration saturated, which caps new inflow. Google still has room via Android defaults, YouTube integration and global product velocity.

The Numbers Side by Side

Item Google Naver
July 2026 MAU 47,022,854 46,845,017
Overall app rank 3rd 4th
May gap ~400K behind ahead
June gap ~250K behind ahead
July result ahead by 177,837 behind
Korean AI chatbot MAU (July) Figure
ChatGPT 23.67M
Gemini 9.47M

Read one caveat carefully: passing on MAU is not passing on search share. Google app MAU includes people reading Discover, using Lens, or hitting the Android home-screen search widget. Industry consensus is that Naver still leads on actual query share in Korea.

The metric matters anyway because of direction. When app opens start slipping, time-on-app usually follows, and query share after that. And once a habit relocates, it comes back slowly if at all.

What It Means for Each Party

Google gains symbolic ground. Korea was one of a handful of markets where Google didn't lead search — and unlike China or Russia, it lost there on competition rather than policy. Flipping the ranking reads as evidence that Google's AI strategy travels across language and culture barriers, and Mountain View now has a case study to apply elsewhere.

Naver faces a structural question. Its strength was Korean-language content assets — Cafés, blogs, Knowledge iN — that kept people inside because the answers were inside. Once AI reads across sources and composes answers, "where the content lives" matters less than "who summarizes well." The moat erodes.

Naver's counter-move is to tighten the moat further: block outside AI from training on or citing its content and reserve it for its own models. Plenty of platforms have gone that way. But that only works on people already inside your app — if they're opening a different one, what's inside yours is irrelevant.

Korea's AI industry gets a warning. Efforts like the Upstage-Daum rollout are pushing domestic models into search, and users are drifting to Google in the meantime. Technological self-sufficiency and market share are different problems.

Advertisers face a budget question. Korean digital ad spend has centered on Naver for two decades. Once user metrics flip, allocation debates follow — though search advertising tracks query share more directly than MAU, so this will move slower than the headline.

When Search Rankings Shifted Before

Yahoo Korea is the first case. A leader in the early 2000s, it lost as Naver accumulated Korean-language Q&A data through Knowledge iN, and it exited Korea in 2012. The decider wasn't search technology; it was the volume of Korean content.

Daum followed a similar path. It had a large base through Hanmail and Cafés but lost search to Naver. Content attracts people, people attract more content, and Naver won that loop — which also demonstrated how hard the loop is to reverse once it tilts.

Baidu in China is the inverse. It held the market after Google withdrew, then bled traffic to WeChat and ByteDance during the mobile shift. Baidu never lost search — people just stopped searching as much.

What's happening in Korea now blends all three. Naver's twenty-year content moat is still real, while the interface through which people get information is changing underneath it. The very logic that beat Yahoo is now working against Naver.

The Board From Here

Naver's counters are expanded AI Briefing, a stronger HyperCLOVA X, and bundling with commerce, payments and maps. On search alone it's outgunned; the Korea-only service bundle is still formidable. The problem this month's numbers reveal is that the bundle didn't defend MAU — partly because commerce and payments live in separate apps that don't feed the Naver app's count.

Google owns the default. Most Korean phones run Android, and the home-screen widget and Chrome's default search are both Google. Every AI feature added compounds the value of that default.

OpenAI has more than twice Gemini's Korean MAU at 23.67 million, but ChatGPT functions as a separate tool rather than a search replacement, so it isn't disturbing the search ad market directly. Now that OpenAI has begun introducing ads, that could change — but not yet.

Kakao is rebuilding its AI strategy after Daum's operator moved to Upstage. As the portal fight shifts from Naver-vs-Daum to Naver-vs-Google, where Daum lands is unresolved.

Regulation is a wildcard. Korean platform regulation debates have largely targeted domestic operators. A foreign company at number one changes the framing, and network fees, cross-border data transfer and taxation could resurface. Worth noting that regulation has rarely reversed a ranking anywhere.

So What Actually Changes

For everyday users, this is confirmation of something already happening. If you've been opening Google instead of Naver more often, the statistics caught up. Getting answers directly from AI summaries is a hard habit to unwind. That said, Naver remains strong for local information, restaurants and real estate, where Korean community data is the substance — so split usage by task will persist for a while.

For content creators, distribution strategy needs a rethink. Optimizing for Naver blogs and Cafés was the correct answer for a long time; if the axis of search traffic moves, that premise wobbles. Being citable in Google search and AI summaries now matters alongside it — and owning your own domain gains value again, because content trapped inside a platform sinks with the platform.

For marketers, reset the dashboard. Budgets and KPIs built around Naver search ads should sit next to Google's. But MAU isn't query share, so decide on conversion data, not headlines — pull six months of your own channel data and check the trend directly, because it varies enormously by vertical.

For Korean AI companies, the market got harder. Building a domestic model and getting users onto a service that contains it are separate achievements. This number says distribution, not capability, is the binding constraint — plenty of teams can build a model; almost none have a channel that puts it in front of tens of millions of people daily.

For investors, watch the follow-through. This is one month of MAU. Whether it hardens into a trend needs August and September, and the real evidence is whether lagging indicators — query share, ad revenue — move with it. Naver's Search Platform revenue growth rate is the most direct signal.

🥄 Three Things You're Probably Wondering

— So Google is number one in Korean search now? No. MAU counts anyone who opened the app; search share counts actual queries. Google app MAU includes Discover and Lens users. On query share, industry consensus still puts Naver ahead.

— Isn't 178,000 within the noise? It's 0.4% of 47 million, so yes, it's small in absolute terms. The signal is the direction: 400,000 behind in May to 177,000 ahead in July is a two-month swing. Whether it's a one-month blip needs the August figures to answer.

— Is Gemini definitely the cause? Isolating the contribution is hard. Gemini's Korean MAU did hit a record 9.47 million, and AI Overviews and AI Mode did expand inside the Google app. But YouTube integration and Android defaults are also in play, so "Gemini alone" overstates it. Several factors pushing the same direction is the accurate reading.

References

Numbers and criteria are as of measurement and may change.