Google used its I/O 2026 developer conference in May to make the most aggressive statement yet about the future of online search: the company’s AI Mode experience has crossed more than 1 billion monthly users, queries are more than doubling every quarter, and the entire product now runs on a new default model called Gemini 3.5 Flash. With the rollout reaching nearly 200 countries and territories across 98 languages at no subscription cost, Google AI Mode has gone from a US-only experiment to the largest deployment of agentic search on the planet in roughly twelve months.
This is not an incremental feature update. Google is converting its core money-making product – the search box that just posted an all-time-high query volume last quarter – into an agent platform that can book a karaoke room, build a custom interface on the fly, and run background “information agents” 24/7. For a business that earns the majority of Alphabet’s revenue from search advertising, that is a high-stakes bet. This analysis breaks down what Google announced, what the numbers actually mean, how it compares with OpenAI’s ChatGPT and Perplexity, and what it signals for publishers, advertisers, and investors in the back half of 2026.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Google AI Mode Passes 1 Billion Monthly Users
The headline figure from Google’s I/O 2026 Search announcement is scale. Google first tested the feature in Search Labs running on a custom version of Gemini 2.0 in March 2025, per Google’s own blog, before Google AI Mode – the dedicated conversational tab inside Search – opened to all U.S. users just two-and-a-half months later on May 20, 2025, at that year’s Google I/O, retiring the Search Labs sign-up requirement that had previously gated the feature, a timeline The Guardian also documented. It now serves more than 1 billion monthly users, a threshold Search Engine Journal tied directly to AI Mode’s one-year anniversary in May 2026, just twelve months after that 2025 debut. The ramp to get there was steep: Digital Applied estimated AI Mode was still at roughly 75 million daily users in early 2026 before the growth curve accelerated into the billion-user milestone a few months later. Under the hood, AI Mode still leans on the “query fan-out” technique Google introduced at that 2025 launch, firing off multiple queries simultaneously so it can stitch together a single response to complex, multi-part questions. To put that in perspective, AI Mode reached the billion-user threshold faster than almost any consumer product in Google’s history, and it did so as an opt-in surface layered on top of classic Search rather than as a standalone app.
Two supporting data points matter just as much as the raw user count. First, Google said AI Mode queries are more than doubling every quarter since launch – a compounding growth curve that Google VP Nick Fox had already previewed in January 2026, when he told Quantumrun that AI Mode was handling 1 billion monthly queries on top of 100 million active users across the US and India, months before the platform’s overall monthly-user count itself crossed the billion mark; tracking firm WebOpTech corroborated that trajectory, reporting that by June 2026 AI Mode’s own monthly query volume had exceeded 1 billion. Second, Google stressed that overall Search queries hit an all-time high in the last quarter. That second statistic is a direct rebuttal to the long-standing bear thesis that generative AI would cannibalize search volume. According to Google’s own data, the opposite is happening: AI features are expanding the total number of questions people ask.
The expansion is genuinely global, and it built on a base that was already large before I/O 2026. 9to5Google reported that AI Mode had already crossed 100 million monthly active users across the US and India by July 2025 – the same month AI Mode launched in the UK running a custom version of Gemini 2.5 as a dedicated tab in Search and the Google app, according to the Google UK Blog. That two-market footprint expanded fast: the Google Blog confirmed AI Mode’s country and territory support had grown past 180 by September 2025, and the company then scaled it further still: according to the Google Blog, the company completed the rollout of Personal Intelligence in AI Mode to nearly 200 countries and territories across 98 languages in May 2026, with no subscription required; users can opt in to connect Gmail and other Google apps so AI Mode has personalized context to work with, and Google has said it plans to eventually draw on a user’s search history as well for more tailored recommendations. Separately, Google Research reported that the underlying Gemini system now spans more than 70 languages across more than 230 countries, which the company describes as making Gemini the most widely available system of its kind. The “no subscription” framing is strategic: where OpenAI gates its most capable models behind a paywall, Google is giving its flagship AI search away for free to defend the funnel that feeds its ad business.
Gemini 3.5 Flash: The New Default Model in Search
Powering this scale is Gemini 3.5 Flash, which Google confirmed on May 19, 2026 became the default model in AI Mode globally – the latest step in a model upgrade cycle that has moved fast even by Google’s standards. AI Mode’s custom version of Gemini 2.5 gained a model switcher offering Gemini 2.5 Pro to Google AI Pro and Ultra subscribers by July 2025, per 9to5Google, before Google introduced Gemini 3 into AI Mode across roughly 120 countries in December 2025, according to the Google Blog. That cadence hasn’t slowed since: an AnswerHalo report found that a Google update on September 2, 2026 pushed Gemini 3.8 Flash into AI Mode’s stack, with Google confirming the new model is now available worldwide to AI Pro and Ultra subscribers, effectively superseding 3.5 Flash as the model doing the heavy lifting behind Search just over three months after it became the global default. Google described the 3.5 Flash generation as its “newest Flash model delivering sustained frontier performance for agents and coding,” and called it the first model in its “latest series of models combining frontier intelligence with action” – framing that Gemini 3.8 Flash now carries forward. The emphasis on “action” is deliberate. Flash models are Google’s latency-optimized tier – built to respond fast and cheap at billion-user scale – and putting a frontier-class Flash model at the center of Search tells you where Google’s priorities sit: speed and agentic capability over maximum raw reasoning.
Crucially, the same model that answers a consumer’s search query is also the model developers can build on. Google said Gemini 3.5 Flash is generally available through the Gemini API, Google AI Studio, Android Studio, and Google’s agent-first development platform, Google Antigravity – and Google has priced its successor, Gemini 3.8 Flash, at $0.75 per million input tokens and $3.75 per million output tokens through December 2026, preserving the low-cost, high-volume economics developers built around on the 3.5 generation. That unification – one model spanning consumer Search and the full developer stack – is a structural advantage. It lets Google amortize a single model’s training and inference investment across its entire product surface, from a free search box in Jakarta to a paid enterprise coding agent in San Francisco.
Why a “Flash” model at the core matters
Choosing a Flash-class model as the global default is an economic decision as much as a technical one. At a billion users running queries that double quarter over quarter, every millisecond of latency and every fraction of a cent of inference cost multiplies into enormous numbers. Google itself has drawn that line for users: alongside the December 2025 rollout of Gemini 3 in AI Mode, the Google Blog confirmed a “Thinking with 3 Pro” option, explicitly labeled as a slower mode for people willing to trade speed for deeper reasoning on hard questions. A heavier, slower “Pro” or “Ultra” model would deliver marginally better answers on the hardest questions but would be ruinously expensive to serve at that volume as the default. By positioning Gemini 3.5 Flash – and now Gemini 3.8 Flash – as “sustained frontier performance” rather than a cut-down budget model, Google is signaling that the gap between its cheap-and-fast tier and its premium tier has narrowed enough to run the world’s largest search engine on the former, while still reserving a Pro-class “Thinking” mode for the minority of queries that need it. Google’s own September 2026 benchmarks back up that claim: the company reported Gemini 3.8 Flash scoring 54.9% on HLE-Verified, lifting DeepSWE v1.1 from 65.3% to 73.7%, and raising Terminal-Bench 2.1 from 85.8% to 89.4% – coding and agentic gains that show the Flash tier closing in on results that used to require a Pro-class model.
Search Agents: The Real Story Behind I/O 2026
The user numbers grab headlines, but the architectural shift is the arrival of Search agents. Google said Search is entering an era where users can create, customize, and manage multiple AI agents directly inside Search. The company is starting with what it calls information agents – agents that operate in the background, 24/7, reasoning across information to surface what a user needs at the right moment, rather than waiting passively for a query. One of the more mature examples, Deep Search, can already issue hundreds of individual searches to assemble a fully-cited, expert-level report in minutes – work that would otherwise take a person hours of manual research to produce.
This reframes Search from a reactive tool into a proactive assistant. The classic model is: you have a need, you type a query, you get ten blue links. The agentic model is: you tell Search what you care about, and an agent monitors, reasons, and acts on your behalf continuously. For a product whose core interface has barely changed in years, that is a fundamental redefinition of what “search” means.
Google also pushed deeper into agentic booking, building on commerce infrastructure it had already scaled well before I/O 2026: Think with Google disclosed that AI Mode’s shopping features draw on the company’s Shopping Graph, which held more than 50 billion product listings as of May 2025. It is expanding Search’s booking capabilities to cover local experiences and services, using the concrete example of finding a private karaoke room for six on a Friday night that serves food late – a multi-constraint task that a traditional search would force you to solve across half a dozen tabs. Google said these booking capabilities will roll out to everyone in the US this summer, and TechCrunch reported on July 16, 2026 that Google had already begun shipping the first piece of that vision: app linking inside AI Mode, launching with three initial partners – Instacart, Canva, and YouTube – so agentic sessions can hand off tasks directly into a partner app rather than stopping at a summary. Pair that with Search gaining the agentic coding capabilities of Gemini 3.5 Flash and Google Antigravity to build “the ideal response format on the fly” – a feature Google calls generative UI, which builds on earlier data-visualization work such as the automatic chart generation for sports and financial questions that Google began rolling out in summer 2025 – and you have Search not just retrieving information but constructing custom interfaces and completing transactions. Google said generative UI will be available to everyone in Search this summer, free of charge.
Google Antigravity 2.0 and the Agentic Developer Stack
Sitting underneath the consumer features is Google Antigravity, the company’s agent-first development platform, which Google Research described as relaunching at I/O as Antigravity 2.0, an improved agentic development platform. Antigravity is where Google’s “agentic era” framing becomes concrete for developers: it is the environment in which Gemini 3.5 Flash can plan, write, and execute code as an autonomous agent rather than as a passive autocomplete.
The strategic logic is that Search agents and developer agents are the same technology pointed at different audiences. The booking agent that reserves your karaoke room and the coding agent that ships a pull request both rely on the same underlying capability: a model that can reason over tools, take multi-step actions, and recover from errors. By unifying these under Antigravity and Gemini 3.5 Flash, Google avoids the fragmentation that plagued earlier efforts – Search Generative Experience, Bard, Duet AI, and assorted Workspace features that each had their own model and roadmap. The consolidation into a single agentic stack is arguably the most important under-the-radar change of I/O 2026.
From AI Overviews to AI Mode: A Two-Year Timeline
To understand why 2026 matters, you have to trace the arc. Google’s generative-search journey has moved fast and survived a rocky start.
| Stage | Year | What it was | Status in 2026 |
|---|---|---|---|
| Search Generative Experience (SGE) | 2023 | Opt-in Labs experiment with AI-written answers | Rebranded |
| AI Overviews | 2024 | AI summaries above results, rolled out broadly in the US | Live worldwide |
| AI Mode | 2025 | Dedicated conversational Search tab | 1B+ monthly users |
| Search Agents + Generative UI | 2026 | Background agents, booking, on-the-fly interfaces | US rollout this summer |
| Gemini 3.5 Flash as default | 2026 | Frontier Flash model powering AI Mode globally | Live worldwide |
The history is instructive because the launch was not smooth. When AI Overviews rolled out broadly in 2024, the feature became a punchline for hallucinated answers – most infamously suggesting users put glue on pizza. Two years later, Google told I/O 2026 attendees that AI Overviews now reaches 2.5 billion monthly users, dwarfing AI Mode’s own billion-user tally and underscoring just how far the underlying generative-search idea has scaled since that shaky 2024 debut. Google also confirmed that the unified experience connecting AI Overviews and AI Mode is now live across desktop and mobile, worldwide, letting users flow from a question to a results page with an AI Overview and then into AI Mode for follow-ups with links to learn more. The journey from a glue-on-pizza meme to a combined multi-billion-user agent platform in 24 months is the real measure of how fast this category is moving.
What Google Executives Said at I/O 2026
Google framed the announcements around a single thesis: AI is becoming the interface to everything. In the official Search announcement, the company positioned the update as a move into “the era of Search agents,” emphasizing that users will be able to “create, customize, and manage multiple AI agents” inside the product they already use billions of times a day.
Liz Reid, VP and Head of Google Search, has consistently argued that AI features grow rather than shrink search engagement – a position the I/O 2026 data was clearly chosen to support, given the disclosure that queries hit an all-time high. People are asking more questions, and more complex questions, than ever before, is the through-line of Google’s public messaging this year, and the all-time-high query figure is the evidence the company puts behind it.
On the model side, Google’s description of Gemini 3.5 Flash as delivering “sustained frontier performance for agents and coding” reflects the DeepMind philosophy that CEO Demis Hassabis has pushed publicly throughout 2025 and 2026: that the next leap is not bigger models but models that can act – plan, use tools, and complete tasks. Independent analysts have been more cautious. The recurring concern in the analyst community is that by answering more queries directly, Google risks eroding the click-through traffic that funds the open web – the very ecosystem its index depends on. That tension between a better user experience and a sustainable publisher economy is the defining debate of this announcement.
Market Impact: Alphabet Stock and the Ad-Revenue Question
For investors, AI Mode is a double-edged sword. Search advertising remains the majority of Alphabet’s revenue, and anything that changes how users interact with the search box is, by definition, a change to the company’s core economics. The bull case is that more queries and richer agentic experiences create more high-intent commercial moments – booking, shopping, comparing – where ads and transaction fees can be inserted. The bear case is that AI answers reduce clicks, compress the number of ad slots, and invite the kind of antitrust and publisher backlash that has dogged Google for years.
The market has been ambivalent. As of June 8, 2026, Alphabet (GOOGL) traded at $363.31, down slightly from a previous close of $368.53, according to market data from Investing.com. Analyst sentiment remains constructive: a consensus price target of roughly $396.25 was reported as of June 9, 2026, implying meaningful upside from the current price. In other words, Wall Street is treating the agentic-search pivot as a net positive for Alphabet’s long-term position, even amid a broader big-tech selloff in early June 2026.
Google did use I/O 2026 to sharpen one piece of the monetization picture: on May 19, 2026, it revised its Gemini subscription tiers, setting AI Plus at $7.99/month (50% off for the first two months), AI Pro at $19.99/month (50% off the first year), and splitting AI Ultra into two tiers – a new $100/month developer tier and a flagship tier cut from $250 to $200/month. But that is subscription revenue, not advertising, and Google has still not detailed how ads will appear inside agentic flows – whether a booking agent surfaces sponsored venues, whether generative UI includes ad units, or how revenue splits work when an agent completes a transaction. Until that is clear, the ad-revenue question stays open, and it is the single biggest swing factor for the stock over the next several quarters.
The Publisher Problem: AI Overviews and Click-Through Rates
No analysis of AI Mode is complete without the publisher angle, because the web’s content economy is collateral in this transition. A widely cited Pew Research Center study published in July 2025 found that when users encountered a Google search page with an AI Overview, they clicked a traditional search result link in only about 8% of visits, compared with roughly 15% of visits on pages without an AI summary – and only about 1% of users clicked a link inside the AI Overview itself. The same research found users were more likely to end their browsing session entirely after seeing an AI summary.
Those numbers crystallize the “zero-click search” fear. If AI Mode answers a question completely, the publisher whose article supplied the answer may never receive the visit – and therefore none of the ad revenue or subscription opportunity that visit represents. Google’s counterargument is that the unified AI Mode experience surfaces “links to learn more” and that complex agentic queries actually drive users deeper. But the Pew data sets the burden of proof high, and many publishers have reported meaningful referral-traffic declines from Google since AI Overviews scaled.
The stakes are existential for some media businesses. As AI Mode expands to nearly 200 countries, the click-through dynamics that Pew measured in the US will play out globally. Publishers in 98 languages now face the same question their English-language peers have wrestled with since 2024: how to remain visible and monetizable when an AI agent stands between the reader and the source.
Google AI Mode vs ChatGPT vs Perplexity: The Competitive Landscape
Google is not operating in a vacuum. The agentic-search race is now a three-way contest between Google’s distribution advantage, OpenAI’s consumer momentum, and Perplexity’s answer-engine focus.
| Platform | Scale metric | Default model | Pricing | Core positioning |
|---|---|---|---|---|
| Google AI Mode | 1B+ monthly users | Gemini 3.5 Flash | Free, no subscription | Agentic search inside the default search box |
| OpenAI ChatGPT | 900M weekly active users (Feb 2026) | GPT-5 series | Free tier + paid Plus/Pro | Conversational AI with web search |
| Perplexity | 780M+ queries (May 2025) | Multi-model (incl. own Sonar) | Free tier + Pro | Citation-first answer engine |
| Microsoft Copilot | Bundled across Bing, Edge, Windows | OpenAI + in-house | Free + Microsoft 365 add-on | AI assistant embedded in Microsoft surfaces |
OpenAI’s ChatGPT is the closest competitor on raw engagement. OpenAI announced on February 27, 2026, that ChatGPT had reached 900 million weekly active users, up from 800 million reported in October 2025 and 700 million in September 2025 – a trajectory that has it on pace to cross 1 billion before year-end, per reporting from TechCrunch. ChatGPT’s reported annualized revenue run-rate has been cited at roughly $25 billion. The key distinction is metric type: Google reports monthly AI Mode users while OpenAI reports weekly active users, so the figures are not directly comparable. But both crossing the billion-user gravity well in the same year underscores how fast generative search has gone mainstream.
Perplexity competes on answer quality and citations rather than scale. CEO Aravind Srinivas said the company handled more than 780 million queries in May 2025, and Perplexity reached a roughly $18 billion valuation in its 2025 funding activity, with higher figures reported in subsequent talks. Perplexity’s bet is that a citation-first answer engine can win users who distrust both Google’s ad-laden results and ChatGPT’s opaque sourcing. Microsoft Copilot, meanwhile, leans on distribution through Bing, Edge, and Windows rather than a single headline user number.
Google’s structural edge is distribution. Statcounter data has consistently placed Google’s global search market share at just under 90% through 2025, meaning AI Mode reaches users who never had to download an app or change a habit. That default-position advantage is exactly why OpenAI and Perplexity are racing to build browsers and operating-system integrations of their own – they cannot win on a level field where Google already owns the search box.
Five Predictions for Agentic Search Through 2027
Based on the trajectory Google laid out at I/O 2026, here is where the agentic-search shift is likely heading.
- Prediction 1 – AI Mode passes 1.5 billion monthly users by early 2027. With queries doubling quarter over quarter and a free, no-subscription global rollout, the user curve has structural momentum. The constraint will be inference capacity, not demand.
- Prediction 2 – Ads land inside agentic flows by Q4 2026. Google cannot leave booking and generative UI unmonetized for long. Expect sponsored placements in agent results and transaction-fee economics on completed bookings before the holiday season.
- Prediction 3 – Publisher referral traffic from Google keeps falling, and licensing deals accelerate. As zero-click behavior spreads to 98 languages, more large publishers will sign content-licensing agreements with Google and OpenAI rather than rely on declining organic clicks.
- Prediction 4 – A regulatory response to agentic search emerges. Antitrust and publisher-protection scrutiny, already intense in the US and EU, will expand to specifically address AI answers and zero-click search, likely producing new transparency or attribution requirements.
- Prediction 5 – The “Flash-class default” becomes the industry norm. Competitors will follow Google in shipping fast, cheap, frontier-adjacent models as their default, reserving heavyweight reasoning models for premium tiers – because nobody can serve billion-user agentic search on a flagship model economically.
What AI Mode Means for SEO and Content Strategy
For marketers, publishers, and anyone who depends on Google traffic, AI Mode forces a strategic rethink. Optimizing for ten blue links is no longer sufficient when an agent may answer the query, build a custom interface, or complete a transaction without ever rendering a traditional result page. The emerging discipline – variously called generative engine optimization or AI search optimization – prioritizes being cited within AI answers, structuring content so models can extract and attribute it, and capturing high-intent commercial moments that agents still route to external sites. Publishers who want out entirely will get a formal lever: Google Search Console’s toggle to block AI Mode from generating responses based on a site’s content takes effect June 17, 2026, and Google has confirmed it carries zero ranking penalty for sites that opt out.
The practical implications are concrete. Content that simply restates facts an AI can synthesize will lose value; original reporting, proprietary data, expert analysis, and transactional depth will gain it. Schema markup, clear sourcing, and brand authority become more important, not less, because models preferentially cite trustworthy, well-structured sources. And because agentic booking and shopping route users to specific providers, having a directly bookable or purchasable offering – rather than just an informational page – becomes a survival strategy in categories Google’s agents now touch. Publishers now have more visibility into that shift, too: SEO practitioner Tushar Agrawal flagged in a June 3, 2026 post that Search Console rolled out a dedicated AI Mode report breaking out impressions, pages, and hourly performance data for how sites appear inside AI answers, giving site owners their first real measurement tools for this traffic.
The Bottom Line on Google AI Mode in 2026
Google AI Mode at I/O 2026 is the clearest signal yet that the search box is becoming an agent. A billion monthly users, queries doubling every quarter, a frontier Flash model serving 98 languages for free, and the arrival of background agents, agentic booking, and generative UI together amount to the biggest redefinition of Search since the mobile transition. The growth numbers vindicate Google’s bet that AI expands rather than cannibalizes querying – at least so far.
But the hard questions are unresolved. How Google monetizes agentic flows without breaking trust, how it sustains the publisher ecosystem its index relies on amid falling click-through rates, and how it fends off a fast-scaling ChatGPT and a citation-focused Perplexity will determine whether the billion-user milestone is a peak or a base camp. What is no longer in doubt is the direction: search is becoming agentic, it is becoming global, and in June 2026, Google is leading the race it started.
Frequently Asked Questions
How many users does Google AI Mode have in 2026?
Google announced at I/O 2026 that AI Mode has surpassed 1 billion monthly users, with queries more than doubling every quarter since launch. The feature is available in nearly 200 countries and territories across 98 languages at no cost.
What model powers Google AI Mode?
As of I/O 2026, Gemini 3.5 Flash is the default model in AI Mode globally. Google describes it as its newest Flash model delivering “sustained frontier performance for agents and coding,” and it is also available to developers via the Gemini API, Google AI Studio, Android Studio, and Google Antigravity.
Is Google AI Mode free?
Yes. Google has repeatedly emphasized that AI Mode and the expanded Personal Intelligence features require no subscription. Agentic booking and generative UI are also rolling out free of charge to US users this summer.
How does Google AI Mode compare to ChatGPT?
Google reports more than 1 billion monthly AI Mode users, while OpenAI reported 900 million weekly active ChatGPT users as of February 2026. Google’s advantage is distribution inside the default search box; ChatGPT’s is conversational engagement and a paid subscriber base. The metrics aren’t directly comparable, but both are nearing or past the billion-user mark.
Does AI Mode hurt website traffic?
Evidence suggests it can. A Pew Research Center study from July 2025 found users clicked a traditional result link on only about 8% of visits to pages with an AI Overview, versus roughly 15% without one. Many publishers have reported declining referral traffic as AI answers reduce the need to click through.
What are Search agents in Google AI Mode?
Search agents let users create, customize, and manage multiple AI agents inside Search. Google is starting with “information agents” that run in the background 24/7 to surface relevant information, plus agentic booking for local experiences and services, both expanding in the US this summer.
What is Google Antigravity?
Google Antigravity is Google’s agent-first development platform, relaunched at I/O 2026 as Antigravity 2.0. It is where developers build agentic applications on Gemini 3.5 Flash, and it underpins the agentic coding and generative-UI capabilities now appearing in consumer Search.
External references: Google Search I/O 2026 announcement, 100 things Google announced at I/O 2026, TechCrunch on ChatGPT’s 900M weekly users, Pew Research Center, Statcounter search engine market share.
