When I was finally able to experiment with Auto Browse (in real time) I took Google’s suggestions on digital tasks as my starting point and chose online tasks that would help my own life.
Whenever interacting with generative AI tools, a healthy sense of skepticism—and caution—is essential. Google even includes a disclaimer embedded in it Gemini chatbot reminds users that it can produce errors. The Auto Browse tool goes one step further. “Use Gemini carefully and take control as needed,” reads the persistent text that appears on the chatbot’s sidebar every time Auto Browse runs. “You are responsible for Gemini’s actions during assignments.”
Before you try it, you should also consider the security risks associated with this type of automation. Generative AI tools are easily compromised quick injection attack of malicious websites. These attacks try to divert the bot from its task. Potential vulnerabilities in Google’s Auto Browse have not been fully explored by outside researchers, but the risks may be similar to other AI tools that control your computer.
Also, be very careful when you use Auto Browse to make a purchase. Google has safeguards in place that flag certain actions, such as buying things or posting on social media, as sensitive and requiring the user’s permission to proceed. However, I’m not sure how the bot will behave and am worried about the damage it might cause to mine credit cardnot to mention providing financial information to it in the first place.
Here’s the first prompt I sent, card in hand:
I want to book two tickets to the SF symphony tonight. I don’t want to pay for orchestra seats, but the tickets aren’t necessarily the cheapest available. Please select two seats along an aisle.
It’s a bit strange to see Google’s AI agent clicking the tab. First, I saw that it used Gemini 3, Google’s latest model, to strategize and define goals, such as getting two aisle seats at the symphony, in the sidebar text box within seconds. This process is similar to a chatbot using a “reasoning” modeldiscuss the steps it can take before proceeding. Then, the clicking starts. Every step the bot takes as part of a task is logged for users.
Auto Browse’s ability to perform multiple tasks without being sidetracked is better than similar agent tools I tested last year. It navigates to the right website, selects the right show, and clicks through multiple seat sections to check availability. Everything listed in the log shows what is actually being executed.
After a few minutes of working on symphony tickets, the bot stopped clicking. I received a notification to replace and press the “order now” button. At a glance, the AI tool seemed to provide what I asked for, and very quickly.
But if I had unquestioningly ordered the two seats selected by Auto Browse for a date at the symphony, the night would probably end with my boyfriend making me sleep on the couch.







