Google has begun testing its new “Auto Browse” feature within Chrome, allowing an AI agent to automate tasks like booking tickets, shopping, and planning trips. While the idea is to streamline the online experience, early tests reveal a clunky, often illogical system that highlights both the promise and the limitations of AI-driven browsing. This isn’t just about convenience; it’s a preview of how Google intends to reshape user interaction with the web – whether users are ready or not.
Early Access and Initial Hurdles
Auto Browse is currently available to US users with AI Pro or AI Ultra subscriptions. Activation requires navigating Chrome settings and explicitly enabling the “Let Chrome browse for you” toggle, a step many users might overlook. Even after enabling the feature, the bot initially failed to execute commands, requiring account logouts and browser refreshes before functioning. This technical friction underscores the current instability of the system.
The Reality of Automated Tasks
The tool’s performance in real-world scenarios is mixed. When asked to book tickets for the San Francisco Symphony, Auto Browse selected aisle seats in the correct price range but placed them in different rows, demonstrating a lack of common sense despite following instructions precisely. Similarly, when tasked with finding a leather jacket on Depop, it simply chose the first three search results without any curation or qualitative judgment.
The system also struggled with more complex requests, like finding camping spots near San Francisco. After 15 minutes, it only confirmed availability for one campground and directed the user to another website for the rest, defeating the purpose of automation. These results suggest that while Auto Browse can handle basic navigation, it lacks the nuance and contextual awareness needed for truly useful task completion.
Security and Control Concerns
Google includes disclaimers reminding users that the AI makes mistakes and that they are responsible for its actions. This acknowledgement highlights the risks involved in handing over control to an automated agent. Generative AI tools are vulnerable to prompt injection attacks, and Auto Browse is no exception. Moreover, using the tool for purchases raises concerns about potential financial mishaps, despite Google’s safeguards for sensitive actions.
The Future of Browsing?
Google’s push toward AI-driven browsing reflects a broader trend of removing users from the direct experience of the web. Features like AI Overviews in Search and Gemini integrations in Gmail already summarize information and generate responses without requiring active engagement. Auto Browse takes this further by automating entire tasks, but its current limitations raise doubts about its viability.
The tool’s biggest flaw isn’t its technical instability, but its lack of spirit. A human browser would scroll through options, discover hidden gems, and make choices based on intuition. Auto Browse simply executes instructions mechanically, missing the joy and serendipity of online exploration.
Ultimately, Google’s Auto Browse is a glimpse into a future where AI mediates our interaction with the web, but it’s a future that feels far from seamless. For now, the human click remains superior.
