Google just flipped the switch on one of Chrome‘s most ambitious features yet — a version of Gemini that doesn’t just answer questions about what’s on your screen; it can actually go out and do things on the web for you.

Starting this week, Gemini in Chrome is rolling out to every Android user in the U.S., and if you’re paying for AI Pro or AI Ultra, you also get access to something Google calls “auto browse” — an agentic tool that can handle the kind of tedious, multi-step errands most of us put off because they’re just annoying enough to be a hassle.

What auto browse actually does

Think about the last time you had to renew a parking pass, tweak a recurring Amazon order, or piece together flights and hotels for a trip across three different tabs. That’s exactly the category of task Google is targeting here. According to Chrome’s product management director Charmaine Dsilva, auto browse can “book parking for an event, update a recurring online order, or organize travel for your next vacation” — essentially taking the wheel on the digital busywork that eats up a chunk of everyone’s day.

This isn’t a brand-new idea for Google. The company first previewed auto browse back in late January, folding it into a broader wave of Gemini 3-powered updates for Chrome that also included the Nano Banana image tool and a “Personal Intelligence” feature. What’s new now is that the assistant itself — the core Gemini in Chrome experience, minus the full agentic powers — is finally available to everyone on Android in the U.S., not just subscribers. The auto browse piece, though, is still gated behind a paid AI Pro or AI Ultra subscription.

Why Google is being cautious here

Letting an AI click buttons and fill out forms on your behalf sounds great until it goes wrong — imagine an assistant accidentally checking out with the wrong shipping address or approving a purchase you never meant to make. Google clearly built the feature with that risk in mind. The company says its models are specifically trained to catch prompt injection attacks, a well-documented vulnerability where malicious instructions get hidden inside a webpage to hijack an AI agent’s behavior. On top of that, auto browse is designed to pause and ask for your confirmation before it completes anything sensitive, rather than barreling ahead on autopilot.

That caution makes sense given how new this whole category of software still is. Rick Osterloh, Google’s SVP of Platforms and Devices, described auto browse as a way for Gemini to handle “the tedious, mundane digital laundry tasks of the web on your behalf” — a phrase that captures both the appeal and the slightly unsettling idea of handing over the keys to your browser.

You’re not getting in free just yet

If you’re on Android and just want the basic Gemini in Chrome experience — summarizing articles, answering questions about a page, generating images, pulling info from Calendar or Keep without switching tabs — that’s available to everyone now, no subscription required. But auto browse itself is locked to AI Pro and AI Ultra tiers, and it’s U.S.-only for now. Google hasn’t said when, or if, it’ll widen that access, though its pattern with past Gemini features suggests a broader rollout is likely eventually.

To try it, you just tap the Gemini icon or hit the three-dot menu at the top of Chrome on your Android phone.