Google’s new Android assistant copies your grocery list into a cart by itself

Google’s new Android assistant can pick up a grocery list from your notes app, open the shopping app of your choice, and drop the items into the cart on its own. That demo, shown on stage at The Android Show: I/O Edition, is the clearest example of what Google is calling Gemini Intelligence — a layer of proactive, agentic features rolling into Android first. The phone is no longer waiting to be asked. It is offering to finish the chore.

Multi-step task automation across apps is the centerpiece. Gemini Intelligence reads content from one app — a list of names, a set of dates, a draft email — and uses it as input in another, all without the user manually switching contexts. The version Google demonstrated needed user confirmation before placing the order, with what the company described as built-in transparency and control over which apps the assistant is allowed to touch.

A second pillar is Gemini in Chrome for Android, with an experimental auto-browse mode that can navigate websites and complete bookings on the user’s behalf. Google framed this as a way to spare the reader the click-through tedium of a flight or event booking — the assistant performs the steps in the background and surfaces the result. The summarization features that previously lived only in Chrome on desktop are also coming to mobile, available from a button in the address bar.

Smaller features round out the package. Rambler is a voice tool that polishes spoken messages — the user dictates a rough message and Gemini Intelligence smooths the phrasing without changing the meaning. Create My Widget generates custom home-screen widgets from a natural-language prompt, with the example Google showed assembling trip information from Gmail and Calendar into a single travel card. These are the parts of the announcement most visible day to day.

The caveats are not small. The launch is staggered: select Samsung and Google phones first, broader Android device support later this year. Crucially, the proactive layer leans on the phone being able to read content from other apps and act inside them — a permissions model with real privacy weight, especially as the assistant gains the ability to fill forms, place orders and book tickets. Google said control over which apps Gemini can act in stays with the user, but the trust model has not been independently audited, and most independent privacy advocates will want to see the full permissions disclosure before recommending it to non-technical users.

The other constraint is reach. Google has confirmed select Samsung and Google phones for the first wave; specific list and timing per region were not made public on stage. Markets outside the US and Europe will get features at Google’s pace, which historically lags by months for Asia and Latin America. Local-language voice features have not been confirmed for every market — Rambler’s demo was in English.

Gemini Intelligence is the first time Google has packaged its agentic capabilities under a single brand explicitly framed for the average phone user, not for developers. Other features announced alongside it at The Android Show include Googlebook — Google’s new laptop platform — and developer tools for vibe-coded widgets, but the consumer headline is Gemini Intelligence itself.

The rollout begins this summer on select Samsung and Google phones, with broader availability on other Android devices later in 2026. Google’s full I/O developer conference takes place on May 19 and 20 in Mountain View, where Google is expected to detail the developer side of the same platform.