I let Chrome’s AI agent shop, research, and email for me – here’s how it went
Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
- Chrome Auto Browse lets Gemini complete multi-step tasks across the web.
- You stay in control, with confirmations required for purchases and more.
- The feature is desktop-only and limited to eligible US subscribers.
In the past couple of years, web browsers have made major leaps in what they can do. They've long been able to remember your passwords, auto-fill forms, and help you organize and group countless open tabs. With the rise of consumer AI chatbots, however, that is no longer enough. Browsers now need to be far more ambitious and, some would argue, autonomous.
Or at least that is what Google and its competitors, including OpenAI and Microsoft, believe. They've been rolling out AI agents in their browsers with the goal of clicking, searching, comparing, and navigating the web for you. Auto Browse, for example, is Google's biggest move yet toward agentic browsing. It can take actions on your behalf and is built directly into the browser you use.
What can Chrome Auto Browse do?
Auto Browse is technically called "Auto Browse in Gemini in Chrome." Yeah, cumbersome name aside, it's an AI agentic feature in Chrome that lets Google's Gemini AI perform multi-step tasks for you across the web.
Instead of having to manually jump from tab to tab, Chrome can now handle much of the work for you, whether that is comparing products, researching flights, gathering and emailing documents, filling out forms, or checking out. Just tap the Gemini icon in your browser, describe what you want in the side panel that opens, and Auto Browse will take over once you give it permission.
Auto Browse works with other Google apps, including Docs and Gmail. You will need to grant permission in the side panel for it to access those apps if it is your first time using Auto Browse. For example, if you ask Gemini to send an email, you will get a prompt to connect with Gmail. Auto Browse will also ask for confirmation before completing sensitive actions, such as sending messages.
How to use Chrome Auto Browse
What you'll need: The latest desktop version of Chrome for Mac, Windows, or Chromebook, a personal Google account with Gemini in Chrome enabled, and an active Google AI Pro or Ultra subscription. You also need to be 18 or older and located in the US.
1. Enable Auto Browse in Chrome
First, check whether Auto Browse is available to you. Open Chrome, click the three-dot menu, go to Settings, select AI innovations, and then open Gemini in Chrome. Look for "Let Gemini browse for you" under Permissions. If you do not see it, Auto Browse is rolling out and may not be available for your Google account yet.
That said, I did not see the option either, but I was still able to trigger Auto Browse through the Gemini side panel.
Also: I've used nearly every browser out there, and these are my top 4 (spoiler: Chrome is out)
Auto Browse lives inside Gemini, which appears as a persistent side panel in Chrome. You can access Gemini on any tab, but not in Incognito mode. To open it, click the Gemini icon in the Chrome toolbar. If this is your first time using it, follow the opt-in prompts. Once enabled, Gemini opens as a side panel on your current tab.
To activate Auto Browse, I went to Walmart.com, where I already had items in my cart. I asked Gemini in the side panel to check out for me. Gemini prompted me to allow Auto Browse, and once I approved it, the feature took control of the page and began completing the checkout process on my behalf. Nice workaround, right?
How to trigger Auto Browse:
- Click the Gemini icon in the Chrome toolbar to open the side panel.
- Complete the opt-in prompts if needed.
- Ask Gemini to perform an agentic task, such as checking out on a site.
- Approve Auto Browse when prompted.
2. Ask Auto Browse to complete a task
Now that you have Auto Browse enabled, you can start describing tasks for it to complete. This is where Auto Browse feels meaningfully different from the standard Gemini experience in Chrome. Instead of asking a question for it to answer, you give it an instruction, and it works through the task step by step.
You can ask Auto Browse to assemble a cart and check out, compare flights across several dates, find and book a hotel under a certain budget, research products, trips, or ideas, collect receipts or documents from multiple sites, schedule appointments or reservations, fill out forms, connect to other Google apps, and more.
Also: I used one simple script to remove AI from popular browsers (including Chrome and Firefox)
As part of my test, I asked it to email Corey, my husband, a reminder about ordering groceries for our Friday party. In the side panel, Gemini prompted for permission to connect to Gmail. Once approved, Auto Browse took over. It correctly inferred which Corey I meant, drafted the message, and waited for me to review and send it.
I could edit the message directly in the side panel or open it in Gmail if I preferred. Neat.
How to start an Auto Browse task:
- Click the Gemini icon in the toolbar.
- In the side panel, describe what you want done.
- Click Submit when done describing your task.
- You will see Auto Browse take over, with an icon visible in the active tab.
- You can pause or stop Auto Browse at any time to take over or refine your prompt further.
My experience with Auto Browse so far
Auto Browse is not a trigger-and-forget AI tool. It keeps you involved, pausing to ask permission to review a form before submission, confirm details, take over for a step it cannot complete, or approve a sign-in or purchase.
In another task I tried, I asked Auto Browse to create a Disney World itinerary for late August and compile everything into a Google Doc for review. It took over my screen, with the option to resume control at any time, and pulled information from multiple travel blogs, each offering different tips and suggestions. It still needed my permission before compiling everything in a Doc, though.
Once I approved, the result was a neatly organized itinerary tailored to me and the specific trip I was planning.
Also: I used these viral Gemini prompts to find the cheapest flight possible - here are the results
For some tasks, Auto Browse can sign into websites using Google Password Manager. This is optional and always requires permission. In the Walmart example I mentioned earlier, Auto Browse asked for permission to sign in and needed me to confirm that I wanted it to navigate through the checkout process. After I approved, it loaded my saved cart and moved through checkout, but stopped at the final review page so I could double-check everything, including my saved payment details, before placing the order.
No follow-ups
One thing to note is that you can't add follow-up tasks midstream. Instead, you have to stop and enter a new prompt or take over manually. I wanted to test Auto Browse's flexibility, so I stopped the Walmart task to see whether it could reroute and add sourdough bread to my cart. Auto Browse was able to go back to shopping and successfully add "ACE Bakery Sourdough" to my cart.
The thing is, that's not the brand I buy, so in the future I will need to be more specific or ask it to use my past purchase history.
Clunky
I have used a few different agentic browsers, and so far, I prefer OpenAI Atlas the most. While Auto Browse is powerful at launch, it feels constrained by Chrome's side panel. It comes across as a tacked-on feature rather than the core focus, though that may be intentional for people who want to keep using the Chrome they know and love with minimal AI distractions.
If that's the case, Google should make Auto Browse much easier to enable or disable. I found the setup clunky. Because the setting wasn't clearly visible, I had to use a workaround to trigger it. The feature's official name, "Auto Browse in Gemini in Chrome," is also very confusing. But this is a first-generation attempt from Google, so maybe the functionality and name will improve over time.
Usage limits
Another thing to note is that Atlas is available to ChatGPT Plus and Pro users, and I have yet to hit a usage limit as a Plus subscriber.
Auto Browse is not unlimited, even with a paid plan. Currently, Google AI Pro subscribers get up to 20 multi-step tasks per day, which I quickly hit while writing this article. Meanwhile, Google AI Ultra subscribers get up to 200 multi-step tasks per day.
When you hit a limit, Gemini will tell you.
FAQs
Will Auto Browse make purchases without asking me?
No. It will request confirmation before completing sensitive actions, such as purchases, submissions, or sending messages.
Does Auto Browse work in Incognito mode?
No. It is disabled in Incognito.
Can I use Auto Browse on my phone?
Not yet. It is currently limited to desktop Chrome.
Why can't I use Auto Browse in Chrome?
Google has a list of requirements, including being 18 years old, located in the US, subscribed to AI Ultra or Pro, and using the desktop version of Chrome. The company also says it is gradually rolling out Auto Browse in Chrome, so it may not be available to you yet.
Also: Which AI tools are actually worth paying for? I'm keeping these subscriptions in 2026 - here's why
Can I turn off Auto Browse?
Yes. Auto Browse requires explicit permission before it can browse for you, and you can review or change its permissions at any time. Open Chrome Settings, go to AI innovations, and select Gemini in Chrome. From there, open Permissions. You can then manage Gemini's access to your location, mic, and current tab, as well as view and control any connected apps.
You may also see a "Let Gemini browse for you" option under Permissions. Turning it off completely disables Auto Browse.
Show more
Post Comment