A surface with 30 million monthly users became ad inventory today

Starting today, July 21, 2026, when you search on Naver, the AI-generated summary sitting at the very top of the results page can contain ads. Not ads in a labeled box below the summary. Ads inside it, between the paragraphs, written in the same register as the summary itself, by the same kind of system that wrote the summary.

Here's the deal: this date wasn't a leak. Naver put it in writing in its own July 16 press release, which refers to "the AI Briefing ADVoost advertising beginning on the 21st." Electronic Times flagged it back on June 18, and inews24, News1 and ZDNet Korea all confirmed the same date. Today isn't a rumor landing. It's a schedule executing on schedule.

Two numbers frame the whole story. AI Briefing has roughly 30 million monthly active users — reached in about a year, which in a country of ~52 million people means essentially everyone who searches. But the more important number is the one most coverage dropped: AI Briefing appears on about 27% of search results as of June, up from ~20% in May, and Naver's internal target is 40% by year-end. MAU tells you reach. The attach rate tells you surface area — and Naver is telling you it intends the summary format to cover close to half of its search results within six months.

That second number is also where the trap is. Every percentage point that AI Briefing gains at the top of the page pushes the traditional PowerLink ad block further down. Naver built a product that eats its own most profitable real estate. If the summary surface isn't monetized, then every gain in attach rate is a loss in ad revenue base. Read today's launch that way and it stops looking like a growth initiative and starts looking like what it actually is: Naver patching a hole it dug itself.

The players — a company that clawed back share, and an agent that writes your ads

Start with Naver Corp. Founded in June 1999 as Naver.com, merged with Hangame into NHN in the early 2000s, then spun off its games division (NHN Entertainment) in 2013 to become the Naver we know. Listed on KOSPI (035420). 2025 full-year revenue was 12.035 trillion won (roughly $8–9 billion depending on the rate you use). It's Korea's dominant search engine, but it spent years leaking query volume to YouTube, Google and Instagram. Per its own press release, Naver recovered its highest search share in eight years in 2026 — and the thing it credits for the recovery is AI Briefing.

AI Briefing launched in March 2025. It sits at the top of integrated search, analyzes what the user is actually trying to find out, and generates a summarized answer while citing sources from across Naver's ecosystem — blogs, cafés, shopping — and beyond. If you know Google's AI Overviews, you know the shape. Naver publishes two engagement metrics for it: long-tail queries running through AI Briefing were up more than 2.5x year over year as of March 2026, and clicks on follow-up questions are more than 10x their launch-period level. People are asking longer, weirder questions, and then digging one level deeper instead of bouncing.

ADVoost is the piece most non-Korean readers won't recognize, and it matters. It's Naver's AI ad-automation brand, introduced in 2024. As ADVoost Search expanded, every query in site search advertising (PowerLink) became eligible for ADVoost handling, and advertisers who flip on "expanded search" in advanced ad-group options can be shown against queries they never registered. In 2026 Naver even extended the brand into digital out-of-home with ADVoost Screen. So today's launch isn't a new product built from scratch — it's an existing automation engine grafted onto a new surface. That's why it shipped this fast.

Now the people. CEO Choi Soo-yeon set the frame on the Q1 2026 earnings call (April 30): "Naver is a uniquely positioned platform that holds search, commerce and payments infrastructure — the core competencies of the AI agent era — as a single connected flow," and said the company would build "a virtuous cycle running from user satisfaction to expanded monetization, centered on an 'executable AI' strategy." She also signaled that monetization would begin in earnest from Q3 and become a meaningful new revenue line by year-end. Today is the first physical instance of that Q3 promise.

The setup happened at the 2026 Naver Growth Summit, held July 14–15 at the Josun Palace hotel in Gangnam, Seoul, with 180 domestic and international brands across fashion, beauty, household goods, food and beverage, health supplements, consumer electronics, automotive, entertainment, games, finance and retail. Lee Jong-min, head of Naver's advertising business division, said the event was about "discussing new growth strategies with advertisers amid a shift in which AI is changing the user's search, discovery and purchase journey." Min Hyung-pil, who leads advertiser consulting, presented "Brand Discovery in the AI Era — Beyond Search to Action," arguing that "the role of advertising is expanding toward contributing to brand discovery within users' varied exploratory intents and contexts." Translation: Naver gathered 180 advertisers in a room to tell them the rules were about to change.

What actually happened — from the May 7 pilot to today's launch

First, a correction to something a lot of coverage got fuzzy. You'll read that Naver "ran a private beta from May." More precisely: the pilot started May 7, 2026 and ran roughly two months, it was limited to a subset of queries, and it was opt-in — only advertisers who wanted in participated. At the time, AI Briefing's attach rate on search results was about 20%. So the pre-launch evidence base is: some queries, on 20% of results, with self-selected advertisers, for eight weeks. That's the data behind today's general availability.

Second, today wasn't a single event. It was the second step of a two-stage rollout. On July 15, Naver opened a dedicated Advertiser Center for its AI ad products — a place for advertisers to see and configure the new inventory. Six days later, today, the actual AI Briefing placements went live in integrated search. Tools first, surface second. Generous read: Naver gave advertisers a week to prepare. Cynical read: the six-day gap is a formality in a system where the whole point is that you don't have to configure anything.

Now the mechanism, which is the real story. The eligible product is ADVoost search advertising — PowerLink. Naver Shopping product-linked ads are excluded from this launch, explicitly for user-experience reasons. That's a sensible call: drop a shopping card into a summary paragraph and the page reads like a home shopping channel in about half a second. Text ads first, carefully.

Here's how it runs. Naver's AI ad agent learns and analyzes the query, the content of the AI Briefing answer itself, the user's intent and the conversational flow, then selects the most appropriate ad on its own from the pool of ADVoost search ads. The copy is the part that should make you sit up: the AI generates the ad text by analyzing the advertiser's landing page. The example Electronic Times used is perfectly illustrative — a user searches "yellow stains on white clothes," and the AI writes something like "Remove set-in stains with ○○ ultra-concentrated detergent" and attaches it. That's not a creative the advertiser uploaded. That's a sentence the machine wrote for that specific moment. Placement is also the AI's call: it picks the optimal spot between briefing paragraphs or below them, keeping things text-first so the search context isn't wrecked.

Billing is CPC. But there is no auction. Pricing is set against "the average PowerLink price for AI Briefing-related queries," and delivery draws automatically from the advertiser's expanded-search budget. Twenty years of real-time bidding as the organizing principle of Korean search advertising, gone on this surface.

Item Detail
General availability July 21, 2026 (stated in Naver's official press release)
Prior stages May 7 limited pilot (~2 months, subset of queries, opt-in advertisers) → July 15 Advertiser Center opens
Eligible product ADVoost search advertising = PowerLink
Excluded Naver Shopping product-linked ads (user-experience reasons)
Ad selection Naver AI agent analyzes query, briefing content, user intent and conversation flow, then selects autonomously
Ad copy Auto-generated by AI from the advertiser's landing page
Placement Between briefing paragraphs or below; decided by the AI
Pricing CPC, no bidding — based on average PowerLink price for AI Briefing queries, drawn from expanded-search budget
Advertiser controls Timing only — hour of day, day of week
Not available Geo, gender, age or audience targeting; input on ad selection; input on ad copy
Category limits Pre-review categories such as medical are restricted; not every query carries an ad
AI Briefing MAU ~30 million
Attach rate on results ~20% (May) → 27% (June) → 40% target (year-end)

Look at that controls row again, because it's short for a reason. Advertisers can pick when — hour, day of week. That's it. No geographic, gender, age or audience targeting. No say in whether their ad gets picked. No say in what it says. The most accurate description going around is that the search ad platform has absorbed the ad agency's job: of the agency's three core functions — creative production, media selection, creative optimization — the first two now happen inside Naver.

There are limits. Categories subject to pre-approval review, such as medical advertising, are restricted, and not every query gets an ad. Which means the 27% attach rate is emphatically not 27% of monetizable inventory. The sellable slice is meaningfully narrower than the headline number.

What each side gets — and why the timing wasn't optional

Naver's motive is legible in its own financials. Q1 2026 consolidated revenue hit 3.2411 trillion won, up 16.3% year over year — a record first quarter. Operating profit was 541.8 billion won, up 7.2%. Fine so far. Then: net income fell 31.3% to 291.0 billion won. The cause is cost. Q1 operating expenses hit 2.6993 trillion won, up 18.3%, and AI infrastructure costs including GPU procurement were up 32.5% year over year. Naver is spending enormously on AI and had, until today, nothing to point to that the spending had produced in revenue terms.

The advertising line is where it stings most. Q1 ad revenue was 1.3945 trillion won, up 9.3% — short of double digits. And in the same breath the company disclosed that more than 50% of the ad revenue growth came from AI contributions. Put those two facts next to each other and the picture is uncomfortable: AI is already generating over half the growth, and the total is still only 9.3%. Strip the AI contribution out and the underlying business was growing at something closer to 4–5%. AI advertising isn't an upside option for Naver. It's already load-bearing.

By segment: Naver Platform at 1.8398 trillion won (+14.7%), Financial Platform at 459.7 billion won, Global Ventures at 941.6 billion won. For reference, Q4 2025 was 3.1951 trillion won of revenue and 610.6 billion won of operating profit. The body keeps growing; margin is getting squeezed by the AI build-out. This is the classic phase where a company has to prove the capex converts.

For advertisers, the outcome splits sharply by size. Small advertisers do well here. A business with no copywriter and no keyword-bidding strategy can now get placed on premium new inventory as long as its landing page is decent — the barrier to entry basically evaporates. Large brands with brand managers, tone-of-voice guidelines and legal review cycles get something closer to a nightmare: sentences go out under their brand name that nobody at the company wrote or approved. For regulated industries with mandatory copy sign-off, this isn't an inconvenience, it's a compliance problem.

Agencies get their job description rewritten rather than eliminated. Creative production and bid optimization on this surface are gone. What replaces them is landing-page information architecture — making sure the AI reads your brand the way you want it read. Call it the search era's SEO migrating into "how does the generative layer describe us." The catch: nobody has a methodology yet, because Naver hasn't disclosed what signals the AI uses to write copy. For the next few quarters this will be educated guesswork dressed up as strategy.

Users get a genuine gray zone. Optimistically, a relevant product appears right where you were already reading, and a step of the discovery journey disappears. Pessimistically, the line between editorial summary and paid placement dissolves. That's the next section.

The skeptic's case — zero-click, disclosure law, and revenue nobody has verified

Several objections deserve real weight here, so let's take them in order.

One: the zero-click problem. There's a documented critique that rising AI Briefing usage hasn't translated into longer session time. If users read the summary and leave, ad inventory value erodes across the whole platform. And notice the specific trap this creates for this product: billing is CPC. Fewer clicks, less revenue. The better the AI answers the question, the less money the ad makes. That structural tension is baked into the design, and no amount of placement optimization resolves it — it's a conflict between answer quality and monetization at the level of the business model.

Two: ads and information become hard to tell apart. An AI writing copy tuned to the surrounding briefing text and dropping it between paragraphs is, by construction, harder to distinguish from editorial content than a labeled ad block. Korea's Display and Advertising Act and the Fair Trade Commission's guidelines require that advertising be clearly identifiable. Whether an ad whose prose style was machine-matched to the surrounding article satisfies that standard has never been adjudicated. Solution News flagged this as the leading risk, and I'd agree — this is the most likely place for the first real regulatory challenge to land.

Three: the advertiser black box. There's no explanation of why your ad was selected, or why it wasn't. With no auction, there's no "spend more, show more" lever either. And the sharper version of the problem is legal: if the AI-written copy constitutes an exaggerated or misleading claim, who is liable? The advertiser never saw the sentence. Naver wrote it. There is no published answer to this question, and "we'll figure it out" is not a compliance posture for a pharmaceutical, financial or food-supplement brand.

Four: none of the revenue impact is verified. Naver's official position is exactly this and nothing more: "We expect revenue from AI Briefing advertising to begin in the second half, when it formally launches." No CTR. No CPC level. No revenue contribution. No published results from the two-month pilot. And here's the conflation to watch for: 30 million MAU is the AI Briefing user count, not an ad impression count. With pre-review categories excluded and ads not attaching to every query, the actual monetizable surface is far smaller than the headline suggests.

Five: cannibalization could exceed the net add. If AI Briefing occupies 27% of results now and 40% by year-end, traditional PowerLink clicks decline by roughly that displacement. There is currently zero published evidence that new AI ad revenue exceeds the loss. Best case, it's incremental. Worst case, it's a migration to lower-yield inventory — and note that the no-auction, average-price mechanism could actively suppress unit prices on exactly the high-demand keywords where competitive bidding used to drive them up. A CFO looking at this should want the like-for-like comparison, and Naver hasn't offered one.

Precedents — AdWords worked, Advantage+ worked, Enhanced Campaigns got reversed

Success case one: Google AdWords. Launched in 2000, and in 2002 Google introduced CPC auctions and Quality Score, handing ad ranking over to an algorithm. Advertisers hated it at first — "I paid more, why is that guy above me?" It also turned Google into a company with hundreds of billions in revenue. Naver's no-bid AI matching reads as the logical next rung on that automation ladder. AdWords automated rank. Naver has automated rank plus selection plus the copy itself, in one move.

Success case two: Meta's Advantage+. Launched in 2022, it delegates targeting, creative selection and budget allocation wholesale to AI. After Apple's ATT gutted Meta's targeting signal, Advantage+ absorbed much of the loss and drove the 2023–2025 ad revenue recovery. It is the canonical case of reduced advertiser control being justified by results. This is precisely the outcome Naver is betting on: take the control, pay it back in performance.

The failure case: Google Enhanced Campaigns. In 2013 Google forcibly merged desktop and mobile campaigns, removing advertisers' ability to budget separately by device. Advertiser and agency backlash was severe, and in 2016 Google restored device-level bid adjustments. The lesson is blunt: when a platform takes control away and can't prove better outcomes, it gives the control back. Naver has taken the control first and disclosed no performance data at all. Against this precedent, that's the wrong order of operations.

The uglier failure case: Yahoo. Through the 2010s Yahoo outsourced search advertising to Microsoft's Bing and leaned on automation, failed to differentiate, and watched its search ad share collapse. Automation is a mechanism, not a moat. If all Naver gets from AI Briefing ads is parity — "Google does this, so do we" — then what's left is the cannibalization without the differentiation.

How rivals counter — Google, OpenAI, Kakao

Google unveiled ads inside AI Mode at Google Marketing Live 2026, currently testing on US English queries only. Eligible campaign types are Performance Max, AI Max, Shopping and broad match, inserted as recommendations inside list-format answers. Three conditions gate delivery: commercial intent, contextual relevance, and high ad quality. Google is consolidating AI Overviews and AI Mode, and European expansion is expected 6–12 months out. The interesting fact here is that Naver is ahead of Google on shipping. Google is still in English-language testing; Naver went nationwide general availability. That's the advantage of a single-market operator with a simpler regulatory perimeter and no global rollout committee.

OpenAI took the opposite approach. It announced ChatGPT advertising on January 16, 2026, began US testing February 9, and in May outlined expansion to Korea, Japan, the UK, Mexico and Brazil. Ads run on Free and Go tiers, appear below the answer with a "sponsored" label, and OpenAI states explicitly that ads don't influence the content of the answer. Advertisers get no access to conversations, memory or personal data — only aggregate metrics like views and clicks. Pro, Business and Enterprise tiers carry no ads. As of June 2026 it's live in the US, Canada, Australia, New Zealand and the UK. (Caveat: OpenAI's own announcement page and CNBC's coverage returned HTTP 403 on direct fetch, so these details come from search-result summaries rather than verified primary text — worth checking the originals before you build a plan on them.)

The contrast is the sharpest thing in this whole story. OpenAI separates the ad from the answer. Naver fuses them, and has the AI write the ad in the answer's voice. Nobody knows yet which approach is correct. But it's obvious which one carries more user-trust risk, and it isn't OpenAI's. It's also worth noting that OpenAI's stated Korea expansion is real competitive pressure — defending domestic search demand before it leaks to ChatGPT is the most direct explanation for why today, and not next quarter.

Kakao declared 2026 its "first year of AI monetization" (2025 full-year revenue: 8 trillion won). In the second half it plans agentic commerce built on KakaoTalk conversation context, and it's converting commerce-surface ads — historically sold only as guaranteed inventory — into auto-bidding performance inventory on the KakaoTalk ad platform. The stated goal is to lift ad revenue as a share of commerce GMV to roughly 4x the level at the start of the year by year-end. Same direction as Naver: automate advertising with AI to raise yield on existing surfaces. But Kakao's inventory is conversation, not search, and that changes the risk profile entirely. Users tolerate ads mixed into search results far better than ads mixed into a private messaging context.

Zoom out and every major platform on earth is running the same experiment right now: how do you put advertising inside an AI-generated answer? Google's answer is labeling plus quality gates. OpenAI's is strict separation. Naver's is full integration. Over the next two or three years, one of these will turn out to have preserved both user trust and revenue — and the other two will be case studies.

So what actually changes

If you're an advertiser or marketer — there is exactly one thing to fix today, and it's your landing page. That's the input the AI reads when it writes your ad. If your product name, core benefit, differentiator and price range aren't stated in clear declarative sentences on the page, the AI will either write vague copy or skip you entirely. Get the information architecture clean and you can win placement on brand-new inventory with zero creative production. Second: go check your expanded-search budget. This surface draws from it automatically, so if you never look, you'll eventually find spend in your report against a placement you didn't know existed. Third, if you're in a regulated category or have a mandatory copy approval workflow, get legal in a room now and settle who owns liability for AI-written copy. Nobody has answered that publicly, which means you're negotiating it, not inheriting it.

If you're an investor — split the confirmed from the unconfirmed. Confirmed: July 21 general availability, ~30M MAU, 27% attach rate with a 40% year-end target, Q1 ad revenue of 1.3945 trillion won (+9.3%), and 50%+ of ad growth attributed to AI. Unconfirmed: CTR, CPC, and revenue contribution for AI Briefing ads — all of it. The company's only statement is "we expect revenue in the second half." So the Q3 print has two things worth watching: whether ad revenue growth crosses from 9.3% back into double digits, and whether the AI infrastructure cost growth rate (+32.5% in Q1) decelerates. If Naver stays in the zone where revenue climbs while net income falls 31.3%, the mere existence of AI ad inventory won't carry the stock.

If you're a regular user — the practical takeaway is that from today, the summary at the top of your Naver results may contain advertising, and it may not look like advertising. PowerLink has always been a separate labeled block. AI Briefing ads sit between summary paragraphs in a voice matched to the summary. Exactly how they're disclosed is something you'll have to see in the wild, and it's where the regulatory fight will start. Two habits help in the meantime: click through to the cited sources under the summary, and occasionally ask yourself why a particular sentence is in there at all.

If you're a developer or run a site — structured information just got a lot more commercially valuable. When an AI reads your landing page to write your ad, your page's semantic structure, metadata and sentence clarity translate directly into ad performance. Classic SEO was about being indexed well. The objective now is being summarized and cited well by a generative system — the thing people are calling GEO. Today is the first case in the Korean market where that discipline maps directly onto ad revenue rather than organic traffic, which means it's about to get budget, tooling and a lot of bad advice very quickly.

🥄 Three Things You're Probably Wondering

— So what does this mean for me? If you use Naver search, it means the summary at the top of your results can now contain AI-written ad copy. Ads used to live in a clearly separate block; now they sit inside the summary in a similar voice. It's worth pausing on any sentence that reads like a product recommendation before you treat it as neutral information.

— Why now, specifically? Because Naver needed it. Q1 ad revenue grew only 9.3% while AI infrastructure costs rose 32.5%, pushing net income down 31.3%. On top of that, AI Briefing now occupies 27% of search results and is pushing Naver's own PowerLink inventory down the page. Leaving the summary surface unmonetized meant watching its ad revenue base shrink from the inside.

— Is Naver ahead of Google and OpenAI here? On shipping speed, yes. Google is still testing inside AI Mode on US English queries only, and OpenAI's ads are live in a handful of countries. But calling it a win is premature. Naver picked the most aggressive design — the AI writes the copy — while publishing zero performance data and leaving the ad-versus-information disclosure question untested. The Q3 results are the first real read.

Sources

Numbers are as of announcement and may change.