Google launches Gemini Intelligence, bringing agentic AI automation to Android phones

Google has announced Gemini Intelligence, a new set of AI features designed to bring full agentic automation to Android phones. The system can handle complex, multistep tasks across apps, including browsing the web, filling out forms, dictating speech, and interacting with commonly used mobile apps, kinda similar to Claude Cowork or Perplexity Personal Computer but for your phone. It can also use what is visible on the phone screen or in an image to better understand the user’s request.

Tasks are started by long pressing the power button and giving Gemini verbal instructions, but Google says the assistant only begins working after the user asks it to. Purchases still require final confirmation, data access can be managed through Android’s permissions menu, and users can follow Gemini’s activity through a progress bar or stop it at any time. Google’s demos include booking exercise classes, copying grocery lists between apps, and finding similar travel options from a photo.

The update also brings Gemini to Chrome for Android, where it can summarize pages and answer questions about web content. In Gboard, a new Rambler feature will support more natural voice dictation by transcribing speech, preserving the user’s tone, and removing filler words. Google is also adding a natural language widget builder, letting users describe the Android widget they want to create, which in theory would be fully usable on their home screen even if it did not originally come from a specific app. All these new agentic features will arrive first on recent Pixel and Samsung Galaxy devices, with broader Android availability planned later this year.