Twelve months ago, ChatGPT owned the AI market. Not in a “market leader” way. In a “what else would you even use?” way. Similarweb had them at 86.7% of all AI chatbot web traffic. Everyone else was rounding error.
Today? That number is around 65%.
A 20+ percentage point decline in a single year. In mobile apps, it’s even steeper — from 69.1% down to 45.3%, according to Apptopia data reported by Fortune.
ChatGPT still leads. But “dominant” and “leading” are different words, and we’ve crossed from one to the other.
The Numbers Tell the Story
Here’s the market shift in actual data, pulled from Similarweb, Apptopia/Fortune, and Statcounter via Reuters:
| Platform | Jan 2025 | Jan 2026 | Change |
|---|---|---|---|
| ChatGPT (web) | 86.7% | ~65% | -22 pts |
| Gemini (web) | 5.7% | ~21.5% | +16 pts |
| ChatGPT (mobile, U.S.) | 69.1% | 45.3% | -24 pts |
| Gemini (mobile) | 14.7% | 25.2% | +11 pts |
| Grok (U.S.) | 1.9% | 17.8% | +16 pts |
The decline wasn’t sudden. ChatGPT shed roughly 2.2 percentage points per month during Q4 2025. It was a slow leak, not a burst pipe.
Tom Grant, VP of Research at Apptopia, told Fortune: “ChatGPT built the category, but as viable alternatives have scaled, users are naturally diversifying their toolkit. People aren’t just experimenting anymore, they’re building workflows around specific products for specific use cases.”
Where Everyone’s Going
Gemini tripled. From 5.7% to over 21% in web traffic, with 2 billion monthly visits by January 2026 — a 315% increase. Google’s Workspace integration is the engine. With 2.3 billion document interactions in the first half of 2025 and 73% of Gemini enterprise accounts using Workspace, Google didn’t need to win the model benchmarks. They won distribution.
Grok came from nowhere. In the U.S., it hit 17.8% market share — up from 1.9% a year ago. Nearly 10x growth. 35 million active users globally. Analysts credit X (formerly Twitter) integration more than model quality: “Distribution leverage rather than model superiority is likely the primary growth driver.” xAI’s valuation hit $250 billion.
Claude went deep, not wide. Anthropic’s web traffic share stays around 2%, but that’s misleading. Look at engagement instead — 34.7 minutes average time per daily user. Claude users spend dramatically more time per session than any other platform’s users.
And then there’s the enterprise picture.
The Enterprise Flip
Consumer market share tells one story. Enterprise spending tells a completely different one.
Menlo Ventures reported in December 2025 that Anthropic captures 40% of enterprise LLM spend — up from 24%. OpenAI fell to 27%, down from 50%.
Read that again. In enterprise, Anthropic now outspends OpenAI by a significant margin.
Ramp’s AI Index confirms it from a different angle: 1 in 5 businesses on their platform now pays for Anthropic. A year ago, it was 1 in 25. That’s 5x growth in twelve months.
Customers spending $100K+ annually on Anthropic grew 7x. Customers at $1M+ grew from 12 to over 500 in two years. And 79% of Anthropic’s customers also pay for OpenAI — they’re not replacing ChatGPT, they’re adding Claude alongside it.
Anthropic’s annualized revenue? $14 billion, growing 10x year-over-year for three consecutive years. They just closed a $30 billion Series G at a $380 billion valuation.
The Coding Gap
Claude’s enterprise story is really a coding story.
Claude Opus 4.5 became the first model to surpass 80% on SWE-bench Verified — solving 405 out of 500 real-world software engineering problems. That 80.9% score broke a barrier no other model had crossed.
Claude Code hit $1 billion annualized revenue just six months after launch. It’s now at $2.5 billion, doubling since the start of 2026. Business subscriptions quadrupled. And 4% of all GitHub public commits are now authored by Claude Code — projected to reach 20%+ by year-end.
The 2025 Stack Overflow Developer Survey confirmed the developer shift: Claude Sonnet is the most admired LLM among developers with a 67.5% admired rating, used by 43% of professional developers.
For coding tasks specifically, the migration from ChatGPT to Claude isn’t a trend. It happened.
OpenAI’s Financial Pressure
Behind the market share decline sits a financial problem that doesn’t get enough attention.
OpenAI has 800+ million weekly active users. That’s an extraordinary number. But only about 5% maintain paid subscriptions. ChatGPT subscriptions account for roughly 75% of OpenAI’s revenue.
Deutsche Bank found that consumer spending on ChatGPT in the UK, Germany, France, Italy, and Spain has been essentially flat since May 2025. Subscription growth surged 20%+ year-over-year in 2023, but that momentum evaporated.
One analyst captured the math problem perfectly: “Sam Altman announced 300 million new weekly users since March. This sounds great until you realize that about 299 million of the extra users won’t be paying, so every single one of the millions of queries they run will be costing OpenAI money.”
OpenAI projects $14 billion in losses for 2026 — nearly triple early estimates. Cumulative losses from 2023-2028: $44 billion. They’re not expected to turn a profit until 2029.
Which might explain why ChatGPT started showing ads in early February. And why an OpenAI researcher quit the next day, publishing a NYT guest essay titled “OpenAI Is Making the Mistakes Facebook Made.”
The Multi-Model Era
The deeper trend isn’t about any single company winning. It’s that the single-platform era is ending.
A16z’s enterprise research found that 81% of enterprises now use three or more model families in testing or production — up from 68% less than a year ago. And 1 in 5 AI users regularly uses multiple apps.
Alex Kantrowitz at Big Technology put it plainly: “Convergence in model quality is making distribution and ecosystem the key differentiators rather than raw capability.”
The models are getting closer in quality. So the question becomes: where do you use each one?
For most people, that’s shaping up like this:
- ChatGPT for casual questions, browsing, image generation
- Claude for coding, analysis, long documents, and writing
- Gemini for anything touching Google Workspace
- Grok for real-time social media context
- Perplexity for research with citations
Nobody’s deleting ChatGPT. But fewer people are using it for everything.
What This Means Going Forward
Analysts project ChatGPT will stabilize around 50-55% as it retains power users while losing casual traffic to Gemini and free alternatives. Gemini is projected to reach 25-30% through ecosystem integration. Specialized players collectively capture 15-20%.
But here’s the honest take: predictions in this market are worthless beyond a quarter or two. Nobody predicted Grok would go from 0 to 18% in a year. Nobody predicted DeepSeek would briefly surpass ChatGPT as the #1 free app on the U.S. App Store while training their model for $294,000 — a fraction of GPT-class costs.
The market is moving too fast for confident forecasts. What we can say: ChatGPT’s monopoly is over. The question isn’t if you’ll use multiple AI tools. It’s which combination works best for your specific workflow.
That’s a better world for users. More competition. More choice. More pressure on every platform to actually improve.
And if ChatGPT built the category — which it absolutely did — then the fact that five credible alternatives now exist might be the biggest compliment OpenAI ever received.