Google's 'Auto Browse' AI Agent Takes Over the Browser: Results Are Mixed

AI Takes Over Browser Control

Google has launched its new AI agent named 'Auto Browse,' integrated into the Chrome browser, for AI Pro and AI Ultra subscribers in the US. The feature allows users to automatically control the browser, open and close tabs, and perform clicks based on instructions given to the Gemini chatbot. This move is considered part of Google's strategy to fundamentally change the web browsing experience.

Mixed Results in Tests

Initial practical tests of the feature revealed that the AI agent could perform some basic tasks but fell short of fully meeting expectations. During the testing process, tasks such as ticket booking, second-hand clothing shopping, and campsite research were assigned.

In the ticket booking test, the agent followed the user's instructions verbatim and selected seats outside the aisle and orchestra sections. However, it made a fundamental logic error, such as placing two tickets in different rows instead of side-by-side. This situation demonstrated that the AI lacks human-like contextual understanding and common sense.

In the task of searching for a leather jacket on a second-hand platform, the agent added the first three options from the search results to the cart. This behavior was recorded as a mechanical following of the ranking rather than a qualitative assessment or curation.

Security and Control Concerns

The use of Auto Browse inherently brings some security concerns due to the nature of automation. Generative AI tools can be vulnerable to 'prompt injection' attacks on malicious websites. Google states that it has implemented security measures requiring user approval for sensitive transactions like purchases or social media sharing. However, the risk profile of the feature, which has not yet undergone comprehensive security testing by external researchers, is not fully clear.

Google displays a persistent warning text in the Gemini chatbot interface: "Use Gemini carefully and take control if necessary. You are responsible for Gemini's actions during tasks."

The Future of Web Browsing Experience

The release of Auto Browse is an extension of Google's broader strategy to move users away from the 'clicking' action and summarize web interactions through AI. The emergence of social networks like Moltbook, designed exclusively for AI agents, also points to developments in this direction. While the AI Overviews feature summarizes search results, Gemini integration in Gmail generates email responses. Auto Browse aims to perform transactions without step-by-step guidance.

Experts suggest that the ultimate goal of this technology is to radically change how people navigate the web and create an internet experience shaped around the needs of automated browser bots. However, current test results show that this transition is far from flawless and that human oversight remains critically important.

Technological Potential and Human Limits

Auto Browse has shown progress compared to previous similar tools in performing multi-step tasks without deviation. However, the logic errors and superficial choices that emerged during testing once again prove that AI cannot replace human intuition and contextual understanding.

As technology advances, the role AI tools will play in daily life will become clearer. These tools will either remain as assistants used only for trivial tasks or fundamentally redefine the meaning of being online. Current evidence shows that human oversight and critical thinking are indispensable, no matter how advanced automation becomes.