Google brings Gemini 3 agents directly into Chrome

For the past year, AI startups have been trying to convince you to ditch Google Chrome for shiny new AI browsers like OpenAI's Atlas or Perplexity's Comet. Most users simply said no thanks.

Google just made that decision a lot easier. On January 28, the company rolled out a massive wave of Gemini 3 upgrades directly into Chrome, turning the world's most popular browser into an autonomous agent that can surf the web for you.

A persistent sidebar that actually understands your tabs

Google previously tested Gemini as an awkward floating window in Chrome, but it is now moving into a permanent side panel. The rollout covers Windows, macOS, and now Chromebook Plus machines.

One clever addition is tab context grouping. If you open five different links from a single page, Gemini automatically treats them as a shared workspace. You can have it compare specs across product tabs or synthesize reviews without flipping back and forth.

Google also baked in its Nano Banana creative tool. Instead of saving an image, opening a new tab, and re-uploading it, you can now modify images or turn web data into infographics right from the sidebar.

Deep links to Google Apps like Gmail, Calendar, and Flights are now live in settings, with a broader Personal Intelligence rollout promised in the coming months to pull context from your personal data.

Hands-free browsing with built-in guardrails

Beyond basic answering, Google is launching auto browse, a full-blown agentic browsing feature powered by Gemini 3. It is rolling out first to AI Pro and Ultra subscribers in the United States.

Instead of just autofilling your credit card, auto browse can navigate websites across multiple steps. In early testing, users leaned on it to gather tax forms, get contractor quotes, handle expense reports, and speed up driver license renewals.

In one shopping demo, the agent analyzed an inspiration photo, tracked down similar vintage party items across the web, factored in budget limits, applied discount codes, and added them to carts. It can even leverage Google Password Manager to handle logins without exposing raw credentials to the AI model.

Google is being cautious, though. The agent is hard-coded to pause and ask for explicit human confirmation before taking sensitive actions, like checking out or posting to social media.

To standardize how agents shop online, Google also partnered with Shopify, Etsy, Wayfair, and Target to launch the Universal Commerce Protocol.

Why it matters

For power users, Chrome just took away the friction of context switching. If you have access to the Pro tier, auto browse is worth testing for repetitive administrative slogs like cross-checking travel dates or filling out repetitive forms, though it is still too brittle for complex multi-step workflows that require nuanced human judgment.

For the browser market, Google just delivered a brutal blow to AI browser challengers. Startups tried to unseat Chrome with sidebar copilots and agentic demos, but Chrome’s default market share proved nearly impossible to overcome. By folding those exact features into Chrome, Google basically rendered third-party AI wrappers redundant.

We think Google played its hand smartly here by locking down ecommerce standards through the Universal Commerce Protocol rather than just showing off party tricks. The real hurdle is reliability: web agents break constantly when websites change their layouts, and if users have to babysit the browser through every checkout screen, the novelty could wear off fast.