Ask someone under 40 which AI they use and you'll almost always get "ChatGPT" as a reflex answer, the way "Google it" replaced "search for it." But the traffic numbers tell a different story than the reflex does, and the businesses actually paying for AI tell a different story still.
ChatGPT is still the biggest single name in the room. It holds roughly 54-58% of worldwide generative AI web-visit share as of mid-2026 and counts more than 900 million users. That sounds dominant, and it is. But a year earlier that same share was closer to 79-87%. Losing that much ground while still growing in absolute users means the category itself got bigger, and the new users didn't all land in one place.
Where the traffic actually went
Google's Gemini is the biggest beneficiary. It sits at roughly 28% of global chatbot web traffic, up close to 450% year over year, and it's the default assistant surfaced inside Search, Gmail, Docs, and every Android phone shipping today. That distribution advantage matters more than any single benchmark win — most people don't choose Gemini, they just already have it open.
Anthropic's Claude is the smaller but faster-growing story: around 9% of global consumer share, but up roughly 855% year over year. Claude's consumer footprint is still modest next to the other two. Its enterprise footprint is not. A June 2026 Pew Research Center survey found 49% of US adults now use an AI chatbot at all, up from 33% in 2024, with 24% using one daily. That's the headline number worth sitting with: half the country crossed the threshold in about eighteen months.
The enterprise story is almost the inverse
Here's where "everyone uses ChatGPT" stops being true in any way that matters to a business owner. Ramp's spending data — pulled from actual payment records across tens of thousands of US companies, not survey responses — shows Anthropic's share of business AI spend climbing from roughly 1 in 25 companies to 1 in 4 in a single year, and Anthropic crossed OpenAI in that measure in April 2026. Separate deal-tracking data puts Claude winning around 70% of new enterprise contracts it competes for head-to-head against OpenAI.
In the enterprise market, precision and trust win. Consumer market share and enterprise market share are moving in opposite directions in 2026, which makes "everyone uses ChatGPT" a poor reason to standardize a business on it.
Financial firms like NBIM and IG Group, and security shops like HackerOne and Palo Alto Networks, have specifically cited Claude's more cautious, harder-to-jailbreak output as the reason they standardized on it for internal tooling over ChatGPT. That's not brand loyalty. That's a risk calculation, made by companies whose job is managing risk for a living.
The tools nobody puts in the headline
Past the big three, there's a real second tier doing real work for real people, each solving a different problem:
- xAI's Grok is the tool of choice when you need to know what's happening on X right now — live, socially-informed answers the others can't match because they aren't sitting on that firehose of real-time posts.
- DeepSeek has become the default for developers watching an API bill, running frontier-tier reasoning at roughly a tenth of what OpenAI charges per token.
- Perplexity has quietly built an answer-engine habit: 34 million monthly active users on its core search product, more than 100 million across its full stack including the Comet browser, and 35-45 million queries a day as of May 2026.
- Meta AI rides distribution the same way Gemini does — baked into Instagram, WhatsApp, and Facebook, reaching people who never typed a URL for a standalone AI product in their life.
The overall enterprise market still runs as something close to an oligopoly: Anthropic, OpenAI, and Google account for roughly 90% of enterprise share between them, with Meta's open-weight models and DeepSeek picking up most of the cost-sensitive remainder.
What this actually means for you
If you're choosing a tool based on "what everyone uses," you're optimizing for the wrong variable. ChatGPT remains the strongest general-purpose default for someone who wants one app that does everything reasonably well. Claude is the stronger pick for long documents, careful client-facing writing, and anything where a hallucinated fact costs money or reputation. Gemini is the obvious choice if your business already lives inside Google Workspace. Grok earns its keep only when real-time social context is the actual job. DeepSeek is a developer's cost play, not a consumer product.
I'd argue the mistake isn't picking the "wrong" model — it's picking only one and assuming it covers every job. The businesses moving fastest in 2026 aren't loyal to a single logo. They're routing writing to one model, coding to another, and research to a third, because the data now backs up what should have been obvious months ago: no single AI model is actually best at everything, and the market has finally stopped pretending otherwise.
