ChatGPT Slips Below 50%: The AI Assistant Market Just Fragmented

For most of the generative AI era, there was one default. You wanted an AI assistant, you opened ChatGPT, and that was the end of the conversation. As of June 2026, that default is gone. For the first time, ChatGPT has slipped below 50% of the AI assistant market, and the way the rest of the field has filled in tells a more interesting story than the headline number.

The practical takeaway is simple. There is no single AI assistant to standardize on anymore. Buyers are starting to pick assistants the way they pick any other software category: by task fit, not by reflex. Below we lay out what the data actually says (it is messier than the headlines suggest), and then what it means if you are choosing tools for yourself or a team.

The number depends on who is counting

The "below 50%" milestone is real, but the exact figure swings depending on methodology, and you should know which one you are looking at before you quote it.

Measured by app and user base, the firm Sensor Tower puts ChatGPT at 46.4% by the end of May. That is a clean drop under the majority line. Measured by web traffic, Similarweb tells a different story: ChatGPT holds about 54.7% of visits across the seven largest chatbots, still a majority on that axis. Both numbers come from reputable trackers, and both were reported widely (TechCrunch among them). They are not contradicting each other so much as measuring two different things. App installs and active users skew toward consumer mobile behavior. Web visits skew toward desktop and work sessions.

The honest framing is that ChatGPT either just lost its majority or is about to, depending on the lens. Either way, the direction is the same, and the trend has been pointing this way for a while.

Where the share is going

On the web-traffic measure, the runner-up is not close, but it is no longer a rounding error. Google's Gemini sits at roughly 27.7% of visits across the major chatbots. Anthropic's Claude holds about 10.3%. After that the field thins out fast: Grok, Perplexity, DeepSeek, and Meta AI each hold under 5%.

The monthly active user figures reinforce the shape. ChatGPT still commands around 1.1 billion monthly active users, an enormous base that is not going anywhere. Gemini reports about 662 million, helped by its distribution across Google's products. Claude sits near 245 million, smaller in raw reach but punching well above its weight in specific professional workflows.

So this is not a collapse. ChatGPT is still the largest assistant by a wide margin on most measures. What changed is that it is no longer the assumption. The market went from one product with a long tail to a genuine top three, with a credible second tier behind it.

The enterprise signal is louder than the consumer one

The most telling part of this shift is not happening on phones. It is happening in procurement.

Two recent moves make the point. Microsoft, whose own Microsoft Copilot is built on a deep OpenAI partnership, is reportedly weighing a hosted DeepSeek model for its enterprise Copilot Cowork product. That is notable because it signals a willingness to mix model providers inside a single product rather than betting everything on one. Separately, India's Wipro launched an AI hub built around Anthropic's Claude. A services giant standardizing a major practice on Claude rather than the market leader would have been surprising a year ago.

Neither move is a referendum on ChatGPT's quality. They are bets on fit. Microsoft wants a cost and performance profile that DeepSeek can offer for certain workloads. Wipro wants Claude's strengths for the kind of long, document-heavy reasoning its clients pay for. When the biggest buyers start treating assistants as interchangeable components selected per job, that is the clearest sign yet that the default is dead.

What fragmentation means for buyers

Here is the part that matters if you are the one choosing. A fragmented market is not a problem to solve by finding the new winner. It is a market that finally rewards matching the tool to the job. Based on how these products have actually diverged, this is the rough division of labor we would draw today.

General-purpose and consumer tasks. ChatGPT remains the safest default for broad, everyday work: drafting, brainstorming, summarizing, quick research, and general question answering. Its scale means the widest plugin and integration ecosystem, and most people you collaborate with already know how to use it.

Coding and long-context work. Claude has earned a reputation among developers for code generation, refactoring, and reasoning over large documents or codebases. If your work involves feeding in a lot of context and getting careful, structured output back, this is where a lot of professional users have landed. The Wipro move is the enterprise version of the same preference.

Google ecosystem and search-grounded answers. Gemini makes the most sense if you live in Gmail, Docs, and the rest of Google Workspace, and when you want answers grounded in current web results rather than a model's training cutoff. Its distribution advantage is also its product advantage here.

Research and citations. Perplexity is built around answer-with-sources rather than open-ended chat. When you need something you can verify and link, an assistant that shows its citations by default saves real time over one that you have to fact-check afterward.

Real-time and X-native context. Grok leans into live information and tight integration with X. If your work touches current events, social conversation, or anything where the last few hours matter more than the last few years, that real-time posture is the differentiator.

Cost-sensitive and high-volume workloads. DeepSeek keeps showing up in conversations about price-performance. For teams running large volumes of inference where the marginal cost per call actually moves the budget, the appeal is straightforward. Microsoft's reported interest is the same calculation at enterprise scale.

No single one of these wins every category, and that is the whole point. The right answer for a research team is the wrong answer for a backend engineering team, and both differ from what a marketing department reaching for quick drafts should pick.

How to actually decide

If you are evaluating now, resist the urge to anoint a replacement default. A more durable approach is to start from the jobs, not the brands. List the two or three things your team does most often with an assistant, then map each to the tool that does that job best. Many teams will end up with more than one, and that is fine. The cost of switching between chat assistants is low, and the cost of forcing every task through a single model is paid quietly in worse output.

If you want a structured starting point, our guide to the best AI chatbots compares the major assistants head to head, and our roundup of the best AI search engines covers the research-and-citation use case specifically, where the differences between tools are sharpest.

The bottom line

ChatGPT slipping below 50% is less a story about one product losing and more a story about the market growing up. A year ago the question was "which AI assistant." Now it is "which AI assistant for what," and that is a healthier question. The data from Similarweb and Sensor Tower, the procurement moves at Microsoft and Wipro, and the steady rise of Gemini and Claude all point in the same direction: the era of one default is over, and the era of fit-for-purpose has begun. For buyers, that is good news. You just have to do a little more thinking, and a lot less defaulting.