Google launches a new AI agent-style "Auto Browse" tool for task automation in Chrome
Google Chrome has announced it is rolling out a new "Auto Browse" feature powered by Gemini 3, which lets the AI assistant carry out tasks directly inside the browser, pretty much like other AI-based browsers we have been seeing lately, such as OpenAI’s Atlas, Perplexity’s Comet, or the Dia Browser. For now, it is limited to desktop users with active Google AI Pro or Ultra subscriptions, keeping the feature tied to Google’s paid AI tiers.
The update turns the Gemini sidebar into an always available panel on the right side of Chrome, replacing the earlier floating window. From there, Gemini can interact with websites by clicking, scrolling, and filling in fields to complete user requests. The flow is agent style: users describe the task, get a “Task started” confirmation, and Gemini executes the steps inside Chrome.
Google says Gemini can handle workflows like shopping on sites such as Etsy by searching items, adding them to a cart, applying discount codes, and preparing checkout based on constraints like budget or theme. With permission, it can also log in using Google Password Manager credentials (if you're brave enough), and a “Take over task” button lets users interrupt and switch back to manual control. Google also plans a Personal Intelligence feature so Chrome can remember past conversations and offer more contextual help, but broader availability still depends on server side rollout rather than a specific Chrome version.
- Web Browser
- Free
- Proprietary
Related news
External links
- Chrome gets new Gemini 3 features, including auto browseGoogle The Keyword • Official source
- Google Chrome Gets Gemini Side Panel and Agentic Browsing FeaturesMacRumors
- Google Chrome unveils Gemini-powered auto-browsing featureMashable
- Google Chrome no longer needs you, as Gemini takes the driving seatNeowin
Comments
Oh great, so now people will be trained to be too lazy to even use a web browser...
"This includes researching hotels and making hospitality and flight reservations, scheduling appointments, filing tax (!) reports, and assisting in renewing driver licenses." surely nothing can go horribly wrong with that!
So now you don't have to read the AI-generated websites yourself, since AI can read them for you.
This goes beyond Perplexity Comet, which I've been asking it about.. for filling forms,.. or any other AI that I know of. Comet cannot directly move your mouse, click buttons, scroll, or type into fields on websites the way a human user does.
As I read on, apparently Atlas has some similar capabilities, but it is only available for Mac at this time.
Successfully ran for 57 minutes. 1025 tokens ($82.26) consumed.
I understand using this AI for highly repetitive tasks like filling the same web form dozens of times a day or summary of extremely big documents in sidebar. But it looks like main purpose of AI integration is "auto shopping".
It's like, this way they don't have to convince you to give them money, they can just tell your computer to do it on your behalf.
Obviously. In sales there is one truth. If you tell them that your budget is $50, they will sell you stuff on $50. And robots can do it fast.