7 hours ago
3
Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
- Google Chrome is being upgraded with Gemini 3 and a revamped sidepanel UX.
- The new Auto Browse feature can navigate webpages and take actions on your behalf.
- Auto Browse launches January 28th for AI Pro and AI Ultra subscribers.
Google just announced several updates to its Chrome browser, centered on -- you guessed it -- AI. With Gemini 3, its latest model, the updates aim to "put Gemini to work in Chrome" with personal assistant-style intelligence, tab organization, and the new "Auto Browse" feature, which takes actions on your behalf in Chrome.
Available now for AI Pro and AI Ultra subscribers, Auto Browse is a bit like autopilot for the web, designed to handle digital chores like filling out forms or making routine purchases. "We live our lives on the web," Parisa Tabriz, Google's VP of Chrome, said in a recent press conference. Auto Browse's focus, then, is on mundane, repetitive tasks; the "digital laundry", so to speak.
Also: Google's Gemini 3 is finally here and it's smarter, faster, and free to access
During the press conference, Tabriz showed off new features in a few common use cases like shopping and summarizing information, but the interesting stuff was when Gemini was asked to perform more complex tasks, like managing favorited apartments on Redfin, or shopping on Etsy based on an image -- all the way to inputting credit card info and finalizing the purchase.
Chrome's latest version upgrades Gemini from the small textbox in the upper right corner to a large pane in the browser window -- taking up a full quarter of the screen. You add the browser tabs you want help with, giving Gemini permission to see their content and subsequently take actions in the browser.
One key example showed the user sharing multiple tabs with Gemini and telling the AI to source information across them to fill out a registration form.
Also: I tried Gmail's new Gemini AI features, and now I want to unsubscribe
It goes beyond text inputs, though. A more complex example showed the user telling Gemini to go through all their favorited apartments on Redfin and remove any that aren't designated as pet-friendly.
In another example, the user showed Gemini a photo of a party, and instructed the AI to go to Etsy and buy all the decorations found in the image. Sure enough, the AI navigates to the website and starts searching for specific items it identifies in the image, adding its finds to the cart.
To enable these kinds of shopping tasks, Chrome will now support Google's Universal Commerce Protocol, an open standard for commerce and agentic AI, co-developed alongside major retailers such as Etsy, Target, Shopify, and Wayfair.
Watching it in real time is a little bit like watching someone sharing their screen on Zoom. Gemini navigates to different fields, enters text, and narrates what it's doing at each step. Through it all, there's a "Pause" button at the top of the screen that allows you to take over the task at any time.
Also: The two fastest growing AI chatbots now (neither is ChatGPT)
If watching Gemini add item after item to your cart and then check out with your credit card isn't scary enough, Google says there are built-in safety measures that require the user to confirm before the transaction proceeds. Making purchases is one of them. If it reaches this point, Gemini will prompt you in the browser to approve continuing.
To aid in these hyper-personalized capabilities, Google's new Personal Intelligence feature, introduced in January but rolling out in Chrome in the coming months, aims to securely connect information across your Google apps, including Gmail, Drive, Google Photos, and YouTube, for more specific inquiries.
Essentially, Personal Intelligence adds all of your Google data to the Gemini pot, allowing the AI a more comprehensive picture of your digital identity. This feature is disabled by default, so if you do want to give Gemini access to these things, you'll have to toggle it on. Still, Google says Personal Intelligence is programmed to avoid making assumptions about sensitive topics, like health or finances.
Siccing Gemini on your ever-growing mound of digital chores sounds great, but the problem with Gemini, and AI in general, is that it still isn't completely reliable. It's very good at a select group of actions, but other, seemingly simple actions elude the AI's capabilities.
Also: Gemini is gaining fast on ChatGPT in one particular way, according to new data
For example, Gemini still struggles to generate a document in Sheets from natural-language parameters. It can certainly provide a starting point that can become a finished product with manual tinkering, but specific natural-language commands like "change the date format in column E to read mm/dd/yy" still elude the AI, often with bizarre, unpredictable results.
Ultimately, agentic AI in the browser definitely has a place in the future of how we use the internet, and while this idea of handling mundane tasks has a lot of potential, privacy and accuracy remain top of mind.
For now, its launch is limited to AI Pro and AI Ultra subscribers, with no set date for general availability. Although this isn't explicitly a "beta", Google is right to ease into these Gemini features, since the everyday user will need some time with Gemini to trust it with their credit card.
English (US) ·