Google Gemini has 900 million users and barely any of them chose it

Google's Gemini app reached 900 million monthly active users and 27.9% of the AI chatbot market by May 2026, but almost all growth came from Google pushing it through Search and Android rather than user choice, according to Similarweb and fatjoe data. ChatGPT still leads in engagement with 800 million weekly active users, 11.52 average visits per period versus Gemini's 4.2, and 44% longer session lengths, highlighting the gap between distribution and habitual use.

Gemini has rocketed to 27.9% of the AI chatbot market and 900 million monthly active users, but almost all of that growth came from Google pushing it through Search and Android, not from users picking it. Whether that distinction matters is now the central question in the AI platform war. Numbers can flatter. Google's Gemini app crossed 900 million monthly active users at Google I/O in May 2026, up from 400 million a year earlier, and its share of global AI chatbot web traffic has climbed past 27%, according to Similarweb data. That is a roughly 450% year-over-year surge. On raw figures, it's the fastest market-share expansion in AI platform history. But look past the headline and the story gets complicated fast. ChatGPT still commands more than 800 million weekly active users. Gemini's 900 million is a monthly figure, which means Google is counting people who opened the app once in a 30-day window. More telling: the average ChatGPT user visited the platform 11.52 times in a given period, compared to just 4.2 times for the average Gemini user, according to data compiled by fatjoe. ChatGPT's session length runs about 44% longer than Gemini's. The gap in habitual use is not small. Where did Gemini's user base come from? Mostly from Google. AI Overviews, embedded at the top of Search results, now reach more than 2 billion users monthly across more than 200 countries. Gemini is the default assistant on Android, which runs on over 3 billion devices globally. Twice as many U.S. Android users encounter Gemini through their operating system as through the standalone app. That is an extraordinary distribution advantage, and it does not require a single person to decide they prefer Gemini over anything else. This is the crux of the skepticism. You can build a massive monthly active user number by being the thing that appears when someone presses their phone's button. That doesn't mean you've won the platform war. It means you've won placement. Whether those passive encounters convert into committed users is a genuinely open question, and right now the engagement data suggests they mostly don't. Frankly, Google knows this. The company's own playbook here mirrors what it did with Chrome: pour the product through every surface you control until it becomes the default, then gradually raise the quality until switching feels pointless. It worked. Chrome commands roughly 65% of the browser market. But search was a relatively simple utility. AI assistants, where habitual use and trust accumulate around a specific product's personality and reliability, may be a harder default to make sticky. What this means if you're building on an AI stack For founders and enterprise teams making AI procurement decisions in H2 2026, the distribution-versus-preference distinction matters in concrete ways. Gemini's API pricing is cheaper: $2.00 versus OpenAI's $2.50 per million input tokens, and $12.00 versus $15.00 per million output tokens, according to published rate cards. If your organization already runs on Google Workspace, you likely don't need a fresh legal review to deploy Gemini, while adding ChatGPT Enterprise typically means a new data processing agreement. In compliance-heavy environments like finance or healthcare, that paperwork difference is not trivial. ChatGPT still leads on one thing that matters for production applications: the depth of its developer ecosystem and the sheer volume of real-world deployment experience. OpenAI's GPT-5.4 remains the default for applications where accuracy is non-negotiable, with reported performance metrics that justify the higher per-token cost for coding assistants and AI-driven DevOps pipelines. You're not just buying a model, you're buying the documentation, the community, and the three years of people who have already hit your edge cases. Both platforms now score roughly identically on overall intelligence benchmarks as of May 2026. That parity means the capability argument has largely collapsed. What differentiates them now is context window size, multimodal depth, and ecosystem fit, not raw model quality. ChatGPT's share has slipped from above 75% in mid-2025 to around 64-68% now, as TechCrunch noted in June 2026 when it dipped below 50% on some measurement methodologies. The direction of travel is clear. But Gemini's path to durable dominance runs through a conversion problem it has not yet solved: turning 900 million monthly touchpoints into 900 million people who actually want to use it. Those are very different things. Also read: LG Electronics wins Korea's first Nvidia liquid cooling certification as AI data centers hit a heat wall https://startupfortune.com/lg-electronics-wins-koreas-first-nvidia-liquid-cooling-certification-as-ai-data-centers-hit-a-heat-wall/ • OpenAI connected 300 million weekly health queries to real medical records and the implications are enormous https://startupfortune.com/openai-connected-300-million-weekly-health-queries-to-real-medical-records-and-the-implications-are-enormous/ • Twenty-one banks just lined up behind SoftBank's $40 billion OpenAI bet https://startupfortune.com/twenty-one-banks-just-lined-up-behind-softbanks-40-billion-openai-bet/