Gemini Spark integrates with Chrome to automate tasks

Google is connecting Gemini Spark with Chrome so it can complete tasks online using your signed-in accounts and saved passwords, always with your permission. The feature aims to take repetitive tasks off your hands, but it does not take full control of sensitive actions.

What Gemini Spark can do in Chrome

Spark is Google's agent for carrying out multi-step tasks. With the new integration, it can browse websites in Chrome and move through processes that would normally require you to open several tabs, fill out forms, and compare options.

For example, it could:

• Schedule visits to apartments you have saved.
• Research flight options.
• Start a booking process.
• Use signed-in sessions and saved passwords to move through the required websites.

The difference between asking a chatbot to explain how to do something and using Spark is that Spark tries to do it directly on the web. Even so, that does not mean it can complete every process without supervision.

You still control the sensitive steps

Google says Spark includes safeguards against prompt injection attacks, a technique that tries to trick an agent with hidden instructions embedded in a webpage.

The system also hands control back to you when it reaches a sensitive action, such as a payment. Spark can prepare the process, but you will have to review and confirm that step yourself.

That matters because an agent browsing on your behalf needs access to particularly sensitive information: your accounts, your passwords, and the data you enter into external services. The convenience depends on knowing exactly what you are authorizing and when you need to step in.

More countries, but with initial limits

Google is also expanding access to Spark for Google AI Pro subscribers in more than 160 additional countries. The expansion begins on July 30, 2026, according to the company's announcement.

The Chrome integration, however, is launching initially in the United States. Google plans to bring these features to other regions later.

For you, the potential change is straightforward: some online tasks could go from requiring your constant attention to needing only a final review. But the important point is not that AI can click buttons. It is how much access you are willing to give it to your accounts. The next limit to watch will be how these agents respond when a website changes its rules, requests unexpected data, or demands a decision that cannot be automated safely.