Sorry, OpenAI. Google is winning the chatbot rat race

A year ago, the AI race looked like a popularity contest. The scoreboard lived in screenshots: Who had the slickest demo, the snappiest answers, the most viral prompt. Then, the grown-up version of AI arrived, and the prize turned out to be the same thing it has always been on the internet: default behavior. The place people already start. The box they already type into. And on that battlefield, Google $GOOGL doesn’t need a miracle. Google just needs to keep acting like Google.

ChatGPT has a brand problem most companies would kill for: It’s famous. Fame is great for growth. But fame also paints a target on your forehead and is a lightning rod for every mistake. OpenAI made AI feel like a place you can go, and its chatbot, ChatGPT, made a new behavior mainstream fast. People just needed curiosity, a blank box, a blinking cursor, and one decent answer that felt like magic.

Google, though, doesn’t need magic. Google needs repetition. Google can ship AI at the speed of its own update cycle — Search, Android, Chrome, Gmail, Maps, Workspace, Calendar, YouTube — and make “using Gemini” feel like “using the internet.”

Google has spent two decades turning repetition into a business model, and it’s trying to do it again with AI — by turning Gemini from a destination into a default, and making “ask the machine” feel like something you do without thinking about where you’re doing it — the way it did with search, ads, and the browser tab you keep open like it’s providing emotional support. Google has been building its AI as a place you end up because the rest of the internet keeps funneling you right there. A challenger can build a better destination and still spend years fighting the fact that most people already live inside Google’s primary-colored world — and that most people are busy and most people are lazy. “Good enough” can spread faster than “best” in the right circumstances.

Alphabet reports fourth-quarter and full-year 2025 results on Wednesday. The numbers will matter, the spending will matter, the guidance will matter. But beneath all that, the earnings call is also a checkpoint in a broader fight that has shifted from “who has the best chatbot” to “which chatbot actually gets installed.” OpenAI still has the most famous destination in the category. But Google is building the default layer that the entire category runs on.

ChatGPT won the destination. Google is trying to win the default — and defaults have a way of becoming history.

Muscle memory at scale

Google still owns a lot of “before.” Before you decide where to search, there’s a search box. Before you decide which chatbot you trust, there’s a browser. Before you decide which app to open, there’s a phone, already loaded with a set of defaults.

The internet already routes through Google-shaped doorways. According to StatCounter, in January, Google held 89.8% of the global search engine market share. Chrome accounts for 71.4% of global browser use. Android powered 70.4% of the global mobile operating system market. Those numbers show who gets to turn a new behavior into a reflex without asking anyone to change their routine.

A destination wins when people make a deliberate trip. A default wins when people don’t. OpenAI did the hard part first: It made people want the trip. Google wants to make the trip unnecessary. If the “ask the machine” moment happens inside the place you already start, the chatbot stops being something you seek out and becomes a behavior you perform.

The “AI war” isn’t a single battle. It’s a stack of habits. People don’t wake up thinking, Today I’ll adopt an AI assistant. They wake up and do what they already do: search, scroll, browse, tap a home screen, open a document, refresh a tab. Google owns more of those motions than any company alive, and it’s been quietly threading its models into the motions themselves. Last fall, Google began integrating Gemini directly into Chrome, including features meant to help synthesize content and answer questions within browsing, and Google has been pushing toward more “agentic” browsing tools. If you’re OpenAI, that’s the nightmare scenario: The most common starting point on the internet quietly grows a second brain.

Google’s entire business history is a case study in what happens when the default is everywhere and good enough — good enough to keep people inside the flow, good enough to keep advertisers paying for intent.

The AI layer lets Google do the same trick again, this time with cognition. Instead of sending you out onto the vast worldwide web and hoping you come back, the AI assistant can live alongside your browsing session, your search session, your work session, your life session. In CEO Sundar Pichai’s prepared remarks on Alphabet’s Q3 2025 earnings call, he framed the company’s “full-stack” approach as infrastructure, models, and products that “bring AI to people everywhere,” then pointed straight at Chrome, saying the company was “reimagin[ing] Chrome as a browser powered by AI through deep integrations with Gemini and AI Mode in Search.”

The operative word there? “Deep.” People don’t choose a deep integration. They encounter it.

Apple $AAPL recently announced a multiyear partnership to integrate Google’s Gemini models into a revamped Siri later in 2026 — a perfect example because Siri is a distribution channel masquerading as a personality. If a company with Apple’s aesthetic and Apple’s control decides to borrow Google’s brain for the interface it uses to sit in your pocket, that sure looks like the market voting for Google as a foundation layer. If Apple is a kingmaker, it just handed out a crown. On Tuesday, Google Cloud and Liberty Global agreed to a five-year partnership to integrate Gemini across Liberty Global’s European footprint, including 80 million fixed and mobile connections. And Samsung’s co-CEO said the company expects to double the number of mobile devices with “Galaxy AI” features to 800 million units in 2026, with those features largely powered by Gemini. That’s a factory output number.

OpenAI created a destination. Google has the leverage to make the category ambient — a layer that follows people everywhere they go. If this fight ends with AI becoming a standard interface for information and action, the company that controls the interface surfaces gets to set the tolls.

And Google has spent 25 years perfecting tolls.

Clearing the “good enough” threshold

Of course, distribution alone doesn’t end a platform war. The closing mechanism is quality — not perfection, not universal supremacy, simply the point at which the default stops feeling like a compromise.

For a while, that’s where the story tilted toward OpenAI. ChatGPT felt like the place you went when you wanted the best answer. Google felt like the place you went because your finger already knew where the box was. That gap made it possible to imagine a future where the destination won, even if it had to fight for every click. But the quality narrative around Gemini has stopped sounding like internal morale and started sounding like external consensus, including on public leaderboards that aren’t run by Google.

On LMArena’s Text Arena leaderboard — a large, continual set of head-to-head user preference matchups — gemini-3-pro is ranked No. 1 as of Jan. 29, 2026, with more than 5.1 million total votes across models. That’s a real-time signal that the “Google can’t ship a great model” stereotype is becoming outdated, at least for many users in many everyday interactions.

The behavioral economics of AI are brutally simple. When the default output is mediocre, people override it. Overriding is an act of migration: another tab, another app, another ecosystem. When the default output holds up, the override impulse weakens. That’s the moment distribution becomes sticky. It’s also the moment the destination starts paying an invisible tax: the tax of being the option you choose deliberately, while the other option shows up automatically.

Alphabet’s own numbers frame this as scale. Pichai, in his Q3 2025 remarks, said the Gemini app had “over 650 million monthly active users,” and that queries “increased by 3x from Q2.” He also said first-party models like Gemini were processing 7 billion tokens per minute via direct API use by customers. Take those figures for what they are: corporate statements delivered on an earnings call, meant to persuade. Even with that caveat, the direction is telling. Google is treating Gemini as a consumer product, an enterprise product, and an infrastructure product all at the same time — because it can. It has enough surfaces to deploy across, and enough money to keep the lights on while use patterns harden into habit.

Google’s distribution only becomes destiny once its model quality clears the “good enough” threshold — and right now, a lot of signals say it has. And that’s important. A default that frustrates users is brittle. A default that performs competently becomes the path of least resistance. And the path of least resistance is where most people live.

Who can afford ubiquity?

AI is expensive in a way consumer software hasn’t been for years. The costs don’t stop after launch; they ramp with use. Serving intelligence at scale is a power bill, a chip bill, a data center bill, and a talent bill — and then it’s a patience bill.

Google can pay patience bills.

Google may be better positioned to make AI cheap at scale because it can optimize hardware, software, and distribution — together. OpenAI can absolutely compete, but it has to keep locking down supply in a world where compute is strategy.

Alphabet has been raising its capex forecast as demand for AI infrastructure accelerates, and 2026 is likely to carry an even larger bill. The AI-race winner won’t just be the company with the smartest model. The winner will likely be the company that can subsidize the most intelligence, for the longest, without wrecking the core business that pays for the subsidy. Google has a cash engine built on intent (and ads). It knows how to monetize the moment someone wants something.

OpenAI said it plans to begin testing advertising in the U.S. for ChatGPT’s free and Go tiers, with the initial tests limited to logged-in adults. OpenAI’s brand advantage is also its fragility — the destination is where people form feelings, and feelings are easy to sour when money shows up near the answer box. Sure, ads might be a rational move for a company serving an enormous user base with escalating compute costs. But that move is also an admission that the economics of mass-scale AI are pushing everyone toward the same destination: ads, commerce, and whatever else can turn attention into margin. Google already lives there. Google already has a payment system for the entire internet. It’s called Search.

Heading into Wednesday’s earnings call, the market is listening for whether Alphabet sounds like a company extending an empire or paying insurance premiums. Capex will be read as either disciplined investment tied to demand or a defensive spiral. Cloud commentary will be weighed for signs that AI infrastructure is becoming a durable growth engine rather than a margin-eating arms race. Search will be interpreted as either resilient in the face of interface change or quietly vulnerable to it.

Maybe nobody “wins” the AI race. Maybe the market splits by context: Google dominates “built-in” everyday tasks; OpenAI dominates high-intent creative work; Anthropic’s Claude dominates technical work; enterprises run mixed stacks; and the leader changes depending on whether you’re coding, shopping, searching, writing, or automating.

Maybe OpenAI figures out a way to leapfrog Alphabet — Sam Altman’s long-teased hardware push with Jony Ive turns into a device that ships at real scale and makes ChatGPT something you carry; Apple reverses course and lets someone else (or itself) run its AI; Microsoft $MSFT tightens, rather than loosens its grip on OpenAI inside Office and Windows. Maybe OpenAI could lock in the enterprise “before” the way Google owns the consumer one. Maybe, maybe, maybe. Or maybe not.

OpenAI built the breakout product of the era. It proved the market. It trained users. It made “chat with a robot” feel normal. It also created an interface category that the incumbent with the biggest default footprint on Earth gets to absorb.

In platform wars, the destination can be beloved and still lose. The default doesn’t need to be beloved. It needs to be present, competent, and paid for. Google has all three advantages sitting on top of one another. Wednesday’s call won’t end the war, but it will tell you something important: whether Alphabet is speaking like the company that already owns the next interface, or the company still negotiating for it.