Three years ago, the world of AI was straightforward. ChatGPT on one side, the rest of the world on the other. Today? The boundaries have blurred so much that it is almost impossible to tell where AI ends and conventional software begins. The investment firm Andreessen Horowitz (a16z) published the sixth edition of its Top 100 Generative AI Apps ranking, and the results are... surprising.
ChatGPT: Still king, but...
The numbers are clear. ChatGPT is 2.7 times larger on the web than second-place Gemini, and outperforms it by 2.5 times on mobile. It is used by 900 million people every week—more than 10% of the entire global population. In a single year, it gained half a billion weekly active users. You do not see numbers like that very often.
But take note. The competition has not stood still. Claude from Anthropic grew its number of paid subscribers by more than 200% year over year, while Gemini grew by as much as 258%. Roughly 20% of regular ChatGPT users also use Gemini every week. People are becoming less loyal to a single platform and are starting to combine them.
What did Google do right? It bet on creativity. The Nano Banana model generated 200 million images and brought 10 million new users to Gemini during its first week. Veo 3 then delivered a breakthrough in video generation. Anthropic, meanwhile, is betting on professionals—developers, analysts, and scientists. Claude Code reached one billion dollars in annual revenue in just six months.
Who will emerge as the winner? Maybe no one. Or both. The report’s authors compare the situation more to the war between mobile operating systems than to the battle between search engines. Two different approaches, two different groups of users, both with a chance of success.
The AI world is fragmenting: The West and China
The geography of the AI market increasingly resembles a Cold War map. Western tools—ChatGPT, Claude, Gemini, Perplexity—share virtually the same user base: the US, India, Brazil, the UK, and Indonesia. They cannot reach China or Russia.
DeepSeek is the only tool that crosses this boundary. Its traffic is divided among China (33.5%), Russia (7.1%), and the US (6.6%). Russia, which practically did not exist as a separate market in previous editions of the ranking, has quickly transformed into a third pole. Sanctions created a vacuum, and local products filled it within two years. Yandex Browser with the Alice assistant reached 71 million monthly active users and broke into the global top 10 mobile AI apps.
And where are AI users the most active in the world? Not in the US. In first place is Singapore, followed by the United Arab Emirates, Hong Kong, and South Korea. America came in only 20th. A little ironic, considering that most of these tools were created by Americans.
Video takes the baton from images
Remember when Midjourney and Stable Diffusion dominated the first wave of generative AI? In September 2023, seven of the nine creative tools in the ranking were image generators. Today, there are only three. They have been replaced by video, music, and voice tools.
Why? Because ChatGPT and Gemini improved their own image generators so much that standalone apps stopped making sense for the average user. Midjourney, which was once in the top 10, now sits in 46th place. The only tools that survive are those serving specific communities with particular needs—such as Leonardo, Ideogram, or CivitAI.
Video is a different story. Kling AI, Hailuo, and Pixverse have built genuine user bases. Chinese models long led the way in output quality. Sora from OpenAI? An interesting case. After the launch of version 2.0, it spent 20 days at the top of the US App Store and reached one million downloads faster than ChatGPT itself. But interest then declined—the viral wave ended, and it did not make the mobile ranking. Even so, more than 3 million people use it every day.
Music and voice remain on firmer ground. Suno held on to 15th place, and ElevenLabs has been in the ranking from the very beginning. Voice cloning, dubbing, audio production—these are capabilities that the major players have not yet replicated as simple checkbox features.
AI agents enter the scene
AI is no longer just a chatbot that answers questions. It is starting to act.
The trend began with vibe coding tools such as Lovable, Cursor, or Bolt, which write code for you. But in January 2026, something different arrived. OpenClaw, a project by Austrian developer Peter Steinberger, went from zero to 68,000 stars on GitHub within a few weeks. An agent that connects to your messages on WhatsApp, Telegram, or Signal and carries out tasks in the background, as if you were texting a friend who takes care of everything for you.
In March 2026, OpenClaw became the most-starred project on all of GitHub—surpassing both React and Linux. OpenAI acquired it in February 2026. We will see what comes of it.
The ranking also featured Manus and Genspark—platforms where you enter an open-ended task (research, data analysis, a presentation) and the AI completes the entire thing on its own. Meanwhile, Manus was acquired by Meta for an estimated $2 billion. Genspark announced annual revenue of more than $100 million.
AI is leaving the browser and heading everywhere
The ranking has one fundamental problem: more and more AI users do not appear in it at all. A developer who spends eight hours a day in Claude Code, or an employee who dictates every email using a voice tool—none of them are visible in web traffic data.
AI is shifting from a destination to a feature. Google added Gemini to Chrome and launched Personal Intelligence, which connects the assistant to Gmail, Google Photos, YouTube, and search. Anthropic released Claude for Excel and PowerPoint. OpenAI launched ChatGPT for Excel and is testing ads because Sam Altman wants to bring AI to the billions of people who will not pay for a subscription.
Note-taking tools such as Fireflies, Fathom, Otter, or Granola have collectively reached more than 20 million users. Notion, which made the ranking for the first time, reports that AI features account for more than 50% of its revenue—up from 20% just a year ago.
The world of AI apps is changing faster than we can keep track of. And the most interesting things are surely still to come.
