Gemini in Chrome with Auto Browse: the browser becomes an agent that shops, books and fills forms for you

At I/O 2026 Google brings Gemini 3 inside Chrome with a new side panel and an agentic feature called Auto Browse. It compares products across two tabs in parallel, fills forms reading data from a PDF, runs multi-step research, and always pauses before checkout or sign-in. Available now to US AI Pro and Ultra subscribers. Chrome for Android joins in late June, with Gemini Spark integration coming in the fall.

Fifteen updates for Chrome, but one stands out: at I/O 2026 Google turned the browser into the primary access channel to Gemini for consumer users. The new Gemini in Chrome side panel is always present, built on Gemini 3, and aware of open tabs. The bigger news, however, is what the browser can do without the user clicking.

Auto Browse: what it actually does

Auto Browse is the agentic feature. It takes a multi-step task and executes it: research a flight, compare prices across sites, fill out a sign-up form reading data from a PDF open in another tab, manage subscriptions, draft an email with attachments gathered during browsing. The difference from classic AI extensions is the ability to hold cross-tab state and carry the flow until a sensitive step appears.

The moment Auto Browse needs to sign in, complete a purchase, or post on a social network, the system pauses and asks for explicit confirmation. It is a design choice Google reiterated on its official blog and to TechCrunch: final responsibility stays with the user, the agent does not sign anything in their name. For payments, integration with Google Password Manager handles credentials, but on-screen confirmation is always required.

Multimodal, even with the pointer

The second piece is the multimodal pointer: select two products on the same screen and ask Gemini to compare them. The image transformation pipeline runs Nano Banana directly inside the browser window. Connected Apps expose Gmail, Calendar, YouTube, Maps, Shopping and Flights to the side panel: a question about the inbox or a booked flight no longer requires switching context.

When, where, for whom

The feature is live on desktop Chrome for AI Pro and AI Ultra subscribers in the United States. Chrome for Android joins in late June, on devices with Android 12 or later and at least 4GB of RAM. In the following months, Google confirmed, Auto Browse will integrate with Gemini Spark: the 24/7 personal agent will be able to take over the browser to complete tasks in the background, effectively erasing the last boundary between agent and UI.

What is worth watching with caution

The real issue is not technical feasibility, already demonstrated. It is the consent model: every new Connected App opens access to a data perimeter, and most users tend to grant everything at the first prompt. The second point is concentration: with Auto Browse, Chrome becomes a default agentic platform for the first time, shifting the balance against competitors like Arc, Comet and Dia, who bet on this category before Google. For enterprises, Google announced in parallel a dedicated enterprise plan (Chrome Enterprise with built-in DLP) at around $6 per user per month, which embeds the agentic features inside a corporate control perimeter.

← Back to all announcements