For the past few years, the question “which AI are you using” would likely have produced a common answer: ChatGPT, by default, for almost everything. That is no longer quite true. Market intelligence firm Sensor Tower’s State of AI 2026 report shows a market where Google Gemini, Anthropic’s Claude and a handful of others have carved out distinct audiences with distinct habits.
ChatGPT is still, by a wide margin, the AI assistant most people in the world use. It commands 1.11 billion monthly users worldwide, more than Google Gemini and Claude combined. However, the report’s data shows its market share has been on the decline, dropping below half for the first time since the generative AI boom began. At the same time, other platforms, including Claude and Google Gemini, have been gaining significant ground.
While this owes something to improving capabilities among the competition, it also has to do with what device a user is interacting with AI on, and what they are using it for. At least that is what insights from Seema Shah, VP of Insights at Sensor Tower, suggest.
The Android effect is real
Start with the most structural force driving the trend: the operating system itself. Google Gemini’s growth, from 285 million worldwide monthly users in May 2025 to 662 million by May 2026, has coincided with its deepening presence across Search, Gmail, Docs and Android itself.
“Android integration is clearly an advantage for Google Gemini and has likely contributed meaningfully to its growth,” Shah said. Roughly 81 per cent of Gemini’s users are on Android devices, compared with 70 per cent for ChatGPT, 61 per cent for Claude and 56 per cent for Grok, she noted, a gap that points to Google’s ecosystem and device integrations introducing Gemini to a large audience by default.
But that does not explain the whole story, according to Shah. “Gemini’s growth is not limited to Android,” she said. Looking only at iOS, where Google enjoys none of its Android advantages, Gemini remains one of the leading AI assistants. Between January and May 2026, it ranked second globally by downloads among AI assistant apps on iOS, behind only ChatGPT, and third by average monthly active users, behind ChatGPT and China’s Doubao. Its iOS MAU climbed more than 800 per cent year-on-year over that period.
Gemini’s popularity on iOS could be attributed to a number of reasons. For one, Google has been rolling out new Gemini features, such as Gemini Live’s camera and screen-sharing capabilities, to iOS almost as consistently as to Android. For another, Apple’s own portfolio of native AI features has remained limited, with the company only recently introducing its Siri AI assistant for iOS, leaving more room for a third-party assistant like Gemini to fill the gap.
Claude’s audience looks like a different market altogether
If Gemini’s story is about distribution, Claude’s is about who is actually opening the app. Claude’s worldwide monthly users went from 60.2 million in December 2025 to 245 million by May 2026, a roughly fourfold jump in five months that took its global market share from 3 per cent to 10.3 per cent. In India, the climb was just as sharp: 13.3 million users in December to 72.3 million in May, with market share rising from 2.2 per cent to 10 per cent.
Shah traced the growth to audience composition more than to any single event. “Claude has benefited from growing demand for AI assistants focused on professional, research, and developer use cases,” she said, adding that Sensor Tower’s traffic and audience data show Claude users have stronger ties to platforms such as GitHub and LinkedIn than users of many competing AI assistants, a pattern that points to particularly strong adoption among technical and professional audiences.
The report’s own traffic analysis backs this up, with GitHub and LinkedIn accounting for more than 2 per cent and nearly 2 per cent of Claude’s inbound traffic respectively, a meaningfully higher share than what feeds rival assistants.
However, Claude’s significant growth in recent months cannot be attributed entirely to user personas and specific use cases. As per the report, OpenAI’s agreement to work with the US Department of War triggered a wave of ChatGPT uninstalls in the United States, peaking at roughly 200 per cent above the app’s average in the week of March 9 to 15.
Shah pointed to this directly as a factor in Claude’s gains. “Claude gained traction in the US following backlash to OpenAI’s partnership with the Pentagon in late February,” she said. “Uninstalls for ChatGPT spiked in the following days and Claude recorded more US downloads than ChatGPT each day from March 1 to 5.”
ChatGPT has led on daily downloads every day since, but the gap has stayed considerably narrower than it was before March, she added. By May 2026, Claude’s US market share had climbed to nearly 14 per cent, up from 5 per cent in December.
Anthropic’s own spending has played a part too, Shah said. “Anthropic has significantly increased its marketing investment. The company ramped digital advertising in late 2025 and into 2026 and was among the AI brands advertising during the Super Bowl, helping raise awareness among mainstream audiences.” That matters because it complicates a tidy narrative where Claude’s growth is purely organic, driven by word of mouth among engineers. Some of it was bought, the same way ChatGPT and Gemini buy attention.
Popular AI’s based on personas
The Sensor Tower report also showed that user profiles vary significantly across leading AI assistants. As per data on Android mobile app users in the US in Q1 CY2026, Grok users were more than four times as likely to be crypto traders, while Claude users were approximately three times more likely to use the AI for crypto trading. By contrast, ChatGPT and Google Gemini attracted a broader mix of users, suggesting a more mainstream audience base.
A few key findings from the report stand out:
-
Grok users are 4.33x more likely to fall into the Crypto Traders persona than the general population, while Claude users are 3.06x more likely to fall into the same category -
Grok users are 2.42x more likely to use the AI for fitness-related suggestions than the general population -
Grok users are also 2.42x more likely to use the chatbot for gig-related work -
Copilot users are 3.44x more likely to fall into the Peer to Peer payers persona than the general population -
Gemini users are 1.73x more likely to use the chatbot for shopping for sneakers than the general population -
ChatGPT users are 1.35x more likely to be students than the general population
Generative AI platforms are taking this into account
Analysis of the top 50 ad creatives in the US this year, included in the report, suggests that companies are well aware their platforms are being used for specific use cases, and are tailoring marketing accordingly.
As per the report, ChatGPT expanded beyond education, with everyday-use creatives rising from 8 per cent in H1 CY2025 to 28 per cent in YTD (year-to-date) 2026. Meanwhile, student and learning creatives, which accounted for 54 per cent of top ads in H1 CY2025, no longer appeared among its leading campaigns.
Google Gemini and Microsoft Copilot, meanwhile, are focusing on distinct user needs, as per the report. Gemini doubled down on image generation and education, with the two categories accounting for 76 per cent of its top creatives in YTD 2026. Copilot, in contrast, emphasised practical utility, with everyday-use creatives increasing from 30 per cent in H1 2025 to 48 per cent in YTD 2026, and productivity-focused ads rising from 10 per cent to 24 per cent over the same period.
What you ask an AI may also depend on what device you use
Sensor Tower’s report is built on app and web traffic, not on what is actually said inside conversations. Microsoft’s analysis of 37.5 million Copilot conversations between January and September 2025, the largest study of its kind at the time, fills in some of that gap, and found that what people ask an AI for depends heavily on the device they are using. Health and fitness questions skewed toward mobile. Productivity and career conversations skewed toward desktop during the workday. Philosophical questions became more common late at night, and entertainment chatter dropped off sharply during work hours.
That distinction, between a desktop tool reached for during work and a mobile companion reached for everywhere else, maps reasonably well onto what Sensor Tower is seeing across assistants. Claude’s heavier tilt toward GitHub and LinkedIn traffic looks like a desktop, work-context pattern. Gemini’s dominance on Android, where it rides along on a phone someone is using for everything from maps to messaging, looks like the mobile, ambient-use pattern.
Neither study proves the other’s findings, but together they sketch a market where the assistant a person ends up using is shaped as much by when and where they are reaching for it as by which model they think is “best.”
Maturing market and revenue
AI has evolved well beyond an experimental tool for consumers. As per the Sensor Tower report, with business and professional use cases expanding, users are increasingly willing to pay for premium features and subscriptions.
Claude, with its focus on research and professional workflows, has seen average monthly revenue per user (ARPU) in the US rise from less than $0.50 in September 2025 to $2.76 in May 2026.
This is also evident in Claude’s user distribution. As of May 2026, 13.3 per cent of Claude’s iOS users in the US were subscribers, the highest share among AI platforms. By comparison, 8 per cent of ChatGPT users and 9.7 per cent of Perplexity users in the same demographic were subscribers.
Overall, global in-app purchase revenue from AI apps is projected to surpass $4.25 billion in the first half of CY2026, up 36 per cent half-over-half.
Against that backdrop, India’s revenue position is an uncomfortable data point in the report. The country is projected to log 5.82 billion hours of generative AI app usage in H1 CY2026, more than three times the 1.66 billion hours logged a year earlier. Yet India’s projected in-app purchase revenue from AI apps for the same half-year is just $38 million, against the worldwide total of $4.25 billion, under 1 per cent of global AI app revenue from one of the world’s largest and most engaged user bases.
Shah’s explanation is structural rather than AI-specific. “The gap largely reflects broader mobile market dynamics rather than anything unique to AI apps,” she said. “India is one of the world’s largest mobile markets by users, but it remains relatively under-monetised compared with markets such as the US, Japan, South Korea, and Western Europe. In 2025, India accounted for roughly 0.6 per cent of global mobile IAP revenue across iOS and Google Play, despite representing a much larger share of users.”
That said, the trend line is improving. Shah said that India’s mobile IAP revenue grew 130 per cent between 2020 and 2025, compared with 58 per cent globally, and the country climbed to 21st globally by IAP revenue in 2025, up from 26th in 2020.
AI apps specifically are doing better than the broader market might suggest. In May 2026, ChatGPT ranked third, Claude seventh, and Grok 25th among all of India’s top apps by IAP revenue, regardless of category, which Shah read as evidence that “while revenue per user remains lower than in more mature subscription markets, there is meaningful willingness to pay for premium AI services.”
Closing the rest of the gap, in her view, will depend on continued growth in digital payments, subscription adoption and higher-value use cases, with AI providers likely to lean on premium features, more capable models and localised offerings to convert more of their enormous user bases into paying subscribers.
