Google just handed its AI the keys to your browser. The company's new Auto Browse feature, launched Wednesday, lets Gemini take control of Chrome to autonomously complete tasks like booking flights, shopping online, and filing expenses. Available initially to Google AI Pro and Ultra subscribers in the US, the feature represents a major shift in how we interact with the web - putting AI in the driver's seat while users watch from the sidelines.
Google just made its biggest bet yet on AI agents controlling your digital life. The company's new Auto Browse feature for Chrome launched Wednesday, letting its Gemini AI model take over your browser to handle everything from online shopping to expense filing. It's a watershed moment for consumer AI - the technology that's been hyped in enterprise settings is now coming for your everyday browsing.
The feature arrives as Google doubles down on weaving AI into every corner of Chrome. Last year brought Gemini in Chrome mode for answering questions about web pages and synthesizing details across tabs. Auto Browse pushes that vision much further, transforming the browser from a window you look through into an agent that acts on your behalf.
Access comes with a price tag. Auto Browse launches today exclusively for US subscribers to Google's AI Pro and Ultra plans, accessed through the Gemini sidebar in Chrome. The company hasn't announced when it'll expand to free users or additional countries, suggesting this is a test run before a broader rollout.
Charmaine D'Silva, director of product management for Chrome, demonstrated the feature's promise in a pre-launch demo. "Instead of having to remember where I bought something and try to reorder something," she told Wired, "I can now delegate to Auto Browse within Gemini to be able to go ahead and buy jackets for me." She typed a request into the Gemini sidebar asking it to reorder a jacket from last year and hunt down a discount code before purchasing.
When activated, Auto Browse creates a ghostly digital presence in its own tab, making autonomous clicks and navigating sites while users watch. It's both fascinating and slightly unnerving - your browser moving on its own, following instructions from an AI model that's interpreting your commands.
But Google isn't handing over complete control. A disclaimer on the demo version reads: "Use Gemini carefully and take control if needed. You are responsible for Gemini's actions during tasks." That last line is crucial - even though the AI is doing the work, Google considers users accountable for whatever happens.
The automation has built-in guardrails for sensitive actions. Posting to social media or entering credit card information triggers human oversight requirements. The Chrome bot pauses, shows the steps it's taken, and asks permission to proceed. It's a sensible safety measure, though it undercuts the promise of true hands-off automation.
Security researchers are already raising red flags about prompt injection attacks. Despite Google's safeguards, Auto Browse remains vulnerable to malicious websites that could trick the AI into unintended actions. These attacks work by embedding hidden instructions in web content that override the user's original commands - a fundamental weakness in current AI agent architectures.
The launch puts Google in direct competition with OpenAI, which is developing its own Atlas browser built from the ground up around AI agents. Nearly every major browser now features some level of AI integration, with the Vivaldi browser standing out as a holdout for users who want to avoid AI-powered browsing altogether.
This represents Silicon Valley's broader vision for the future of web interaction - more AI, less human clicking. Whether users actually want an AI agent handling their digital chores remains an open question. The technology tends to be overhyped and unreliable in practice, struggling with tasks that require nuanced judgment or dealing with unexpected website layouts.
Google's phased rollout strategy suggests the company knows Auto Browse needs real-world testing before going mainstream. Early adopters paying for AI Pro or Ultra plans become de facto beta testers, helping Google iron out issues before potentially pushing the feature to Chrome's billions of users.
The timing is strategic. As AI agents move from buzzword to product category, Google needs Chrome to remain the dominant gateway to the web. Letting that position slip to AI-first browsers from competitors would be disastrous for a company that built its empire on search and advertising.
What's unclear is whether autonomous browsing solves a problem people actually have, or creates new ones. Shopping online isn't that hard. Neither is booking flights or filing expenses. These tasks require attention to detail - comparing options, reading reviews, checking terms. Handing them to an AI that might misinterpret instructions or miss important details could cost users money and time.
The feature also raises questions about web economics. If AI agents start doing the browsing, who sees the ads? How do e-commerce sites optimize for bots instead of humans? The entire digital advertising ecosystem that funds free content online assumes human eyeballs. Auto Browse and similar tools could upend those assumptions.
Google's Auto Browse represents a pivotal moment in the evolution of web browsing - the point where AI agents move from research projects to consumer products with real consequences. The feature's success will depend on whether it proves genuinely useful rather than just impressive, and whether users are comfortable letting algorithms handle tasks that require judgment and carry financial risk. With security vulnerabilities still unresolved and reliability questions looming, Auto Browse feels like a capability in search of a use case. But given Google's track record of gradually expanding experimental features, this limited launch to paying subscribers is likely just the opening move in a much larger transformation of how we interact with the web.
