ChatGPT's U.S. Mobile Market Share Falls Below 50% as Gemini and Grok Gain Ground

The chatbot market more than doubled in size over the past year, and ChatGPT lost its grip on it. Two separate analytics firms, tracking different platforms with different methods, landed on the same conclusion: OpenAI's lead is compressing fast.

Similarweb measured the web side. Gemini visits surged from 267.7 million to two billion between January 2025 and January 2026, a 647% jump that pulled Google's share of AI chatbot web traffic from 5.3% to 22%. ChatGPT still dominates at 5.7 billion visits, but its share slid from 86.6% to 64.6% over the same period. On mobile, data from analytics firm Apptopia obtained by Big Technology painted a sharper picture. ChatGPT's daily U.S. user share fell below the 50% mark for the first time, landing at 45.3%, down from 69.1%. Grok pulled the biggest surprise: 1.6% a year ago, 15.2% now. Gemini climbed to 25.1% from 14.7%.

All of this happened while the overall market grew 152%, according to Apptopia. OpenAI is not shrinking. It's losing ground in a category that keeps getting bigger.

A bigger market, more competitors

Think of the chatbot market as a restaurant that just doubled its seating. ChatGPT still has the most tables filled. But Gemini grabbed the new seats, and Grok, which barely had a reservation a year ago, is now pulling chairs from both.

The growth, though, may be hitting its first ceiling. "The market is flattening a bit," said Adam Blacker, Apptopia's director of public relations. "We likely have not hit peak GenAI apps, but this is an early inflection point where there may be a gap between early adopters and everyone else." Most devices Apptopia tracked never opened a chatbot at all. Billions of phones sitting there, chatbot-free.

Similarweb's David Carr, insights news and research editor, spotted something specific in the timing. ChatGPT traffic dipped in November and December, right as Gemini hit a growth spurt. Preliminary January data shows ChatGPT recovering but still short of its October peak above 6 billion visits. Gemini, by Similarweb's early estimate, grew another 17% month over month.

And this is not a straight migration story. Apptopia found that 20% of chatbot users were running at least two apps by late 2025, four times the 5% who did so two years earlier. Users aren't dumping ChatGPT. Plenty of them tap Gemini too. The race is not zero-sum, at least not yet.

Google's distribution advantage

Google's gains come from distribution, not just model quality. Gemini is baked into Android, woven through Workspace, and sitting inside Chrome. When 650 million monthly active users can reach Gemini without downloading anything, the growth curve follows. Apple's January deal to embed Gemini models as the backbone of Apple Intelligence will only accelerate that. You don't need to seek out Gemini. It shows up where you already are.

That's a different competitive posture than OpenAI's. ChatGPT lives in its own app and its own website. Microsoft bundles it into Copilot, but Copilot is its own product with its own branding confusion. Google's advantage is that Gemini doesn't feel like a separate thing to install. It's the answer layer inside products people already use eight hours a day.

Code discovered in Gemini's Android app this week pointed to another play: a chat import tool called "Import AI chats." Android Central first reported the feature, which appears designed to let users upload their ChatGPT or Claude conversation histories directly into Gemini. The importer sits inside the attachment menu, a file upload rather than a migration wizard. If you've built months of project context inside another assistant, Google wants to make the switch feel less like a factory reset.

That move has a catch. Imported chats land in your Gemini Activity data and can be used to train Google's models. You carry your context over. Google keeps the copy. No word yet on supported file formats, regional availability, or whether the feature hits mobile or web first.

Google is not returning the favor. If you want to leave Gemini and take your history to ChatGPT or Claude, there's nothing comparable. That one-way door could draw regulatory attention in the EU, where data portability rules have teeth.

OpenAI's countermoves

OpenAI looks nervous, and it's acting like it. The company launched a standalone Codex app for Mac computers on Monday, opening it temporarily to free and low-cost Go subscribers alongside its usual paid tiers. Sam Altman, sitting in front of reporters at a Friday press briefing, called Codex "the most loved internal product we've ever had." The company doubled rate limits across paid plans to sweeten the pitch. More than 1 million developers used Codex last month, CNBC reported.

Coding tools are where the competitive pressure is sharpest. Anthropic's Claude Code has carved out a loyal developer audience, and Cursor has built a standalone editor that developers swear by. OpenAI's response is to go wider: open Codex to everyone, even the free tier, and bet that volume creates stickiness.

For consumers, there's ChatGPT Go, the $8 monthly plan OpenAI rolled out earlier this year for price-sensitive markets. You get GPT-5.2 Instant, a smaller and faster model, plus ads. They show up in the free and Go tiers now, labeled and walled off from ChatGPT's actual responses. Pay for Pro or Enterprise and you never see them. The pricing ladder is meant to catch users at every income level, a recognition that Google can undercut on price by subsidizing Gemini through ad revenue from its other businesses.

OpenAI's projected 2026 revenue of $20 billion, triple its 2025 figure, depends on all three revenue streams firing: consumer subscriptions at 55-60% of the mix, enterprise at 25-30%, and API access at 15-20%. Revenue like that buys time. But it was built on the assumption that ChatGPT would remain the default choice, and when your market share drops 24 points in twelve months, the default status starts to crack.

Sam Altman declared "code red" internally on December 2, delaying advertising plans and other initiatives to focus resources on improving ChatGPT's core product. The memo, reported by multiple outlets, acknowledged what the Apptopia and Similarweb numbers now confirm publicly: Google is not just catching up. It's catching up fast.

The engagement question nobody is asking

Raw user counts miss something. Take Claude. Anthropic's chatbot has far fewer downloads than ChatGPT or Gemini, and nobody would confuse it for a mass-market product. Yet Apptopia's numbers show something odd: people who open Claude stay for a while. Daily session time has roughly tripled since last June, going from around ten minutes to over thirty. Those are not casual check-ins. That's work getting done.

If you're OpenAI, that should worry you more than Gemini's download numbers. Download growth measures distribution. Session time measures dependency. A user who spends thirty minutes a day inside Claude is building workflows, writing code, running analysis. Pulling that person away requires more than a better benchmark score.

Yann LeCun, the Turing Award winner who left Meta last November after 12 years as chief AI scientist, posted three words on LinkedIn on Saturday: "I use Gemini." The comment appeared under a post by Raia Hadsell, Google DeepMind's VP of Research, discussing Gemini's market share surge. LeCun has publicly called autoregressive language models "doomed" and left Meta partly over frustrations with the company's pivot toward bigger LLMs. That he chose Gemini over ChatGPT or Claude says something about where at least some researchers see technical momentum heading. Google has bet on multimodal from the start, and LeCun has been arguing for exactly that kind of architecture.

It's a single LinkedIn comment. But from a Turing Award winner who just walked away from one of the biggest AI labs in the world, "I use Gemini" carries weight that no press release can match.

What the first plateau looks like

Blacker's phrase sticks: an early inflection point. The chatbot market grew 152% in a year but showed signs of flattening in recent months. Early adopters have picked their tools. The next wave of users, the ones who haven't opened a chatbot on their phone yet, will pick based on whatever is already on their screen. That favors Google.

The enterprise market tells a different story. Companies that built institutional knowledge inside ChatGPT Enterprise or Claude for Work face real switching costs: months of conversation history containing project context, internal terminology, workflow preferences that took quarters to train into the system. Google's chat import tool takes aim at exactly that lock-in. But importing a conversation history is not the same as importing trust. CIOs don't migrate AI infrastructure because a file upload tool exists. They migrate because the product underneath justifies the disruption. Most enterprise procurement teams right now are still running evaluations, not writing migration plans. A file upload is the easy part. The security reviews, compliance certifications, and integration work that follow are what actually slow a switch down.

OpenAI built the category and still leads it. ChatGPT reached 1 billion weekly users by the end of 2025, a milestone Facebook took eight years to hit. But the distance between first place and second place shrank in every metric tracked by both Apptopia and Similarweb over the past twelve months. On mobile, ChatGPT shed 24 points of market share. On the web, the gap between ChatGPT and Gemini compressed from 81 points to 43.

The question for 2026 is not whether ChatGPT survives. Of course it does. Revenue is tripling. The developer base keeps growing. Codex just went free. But the question underneath the revenue line is different: can OpenAI sustain the kind of market position that justifies a $300 billion valuation? Its mobile share sits at 45% and every quarter brings a new competitor closer.

Three years ago, there was ChatGPT and there was everything else. Now there's a market.

The Class Divide in Teen AI: Who Gets the Tutor, Who Gets the Companion

Two-thirds of American teenagers have now used AI chatbots. Pew Research confirmed that figure this month, polling 1,458 teens aged 13 to 17 across the country. Most coverage stopped at the headline.

The Implicator

Meta Won Its Antitrust Case While Losing the Talent War That Actually Matters

Federal judges declared Meta faces "fierce competition" on Tuesday. The company's AI researchers apparently agree. They're just choosing to work for it instead. Judge James Boasberg dismissed the FTC

The Implicator

When Your AI Chatbot Thinks You're God: Inside Grok's Musk Worship Spiral

Grok spent days this week explaining why Elon Musk would dominate LeBron James in fitness competitions, outperform Albert Einstein in intellectual pursuits, and rank as the world's greatest lover. The

The Implicator