Google Rolls Out AI Agent Within Chrome

Google has begun rolling out Auto Browse, an artificial intelligence (AI)-powered browsing agent inside Chrome that can navigate websites and assist users with multi-step tasks. The feature is currently available in preview to Google AI Pro and Google AI Ultra subscribers in the United States. While access remains limited, Auto Browse marks Google’s clearest move yet to push Chrome beyond information display and towards task execution on the web.

For context, the rollout comes amid growing competition among AI companies to control how users interact with the web. While OpenAI and Perplexity have launched AI-native browsers, Google has chosen a different path.

Because it already controls Chrome, Google does not need to launch a separate AI browser. Instead, it is embedding agentic AI capabilities directly into its existing product, allowing it to scale AI-driven workflows without fragmenting users or forcing a switch. Notably, this keeps Chrome central to how people search, browse, and behave online.

What is Auto Browse?

Auto Browse allows users to offload complex browsing workflows to Google’s Gemini AI assistant. Users can ask Chrome to assist with tasks such as researching travel options across multiple dates, filling out long online forms, collecting documents, managing subscriptions, shortlisting apartments or services, and comparing products across websites.

The Auto Browse AI agent operates within Chrome’s browsing environment, opening and navigating tabs as needed and running in the background while users continue with other tasks. Importantly, Auto Browse pauses for user confirmation before sensitive actions such as making purchases or posting social media content. With permission, it can also use Google Password Manager to handle tasks requiring sign-ins.

This marks a shift from earlier Gemini features in Chrome, which focused primarily on summaries and contextual assistance. With Auto Browse, Chrome begins to help prepare and carry out tasks inside the browser, while keeping users involved in final decisions.

How does Auto Browse work inside Chrome?

Technically speaking, Google has built Auto Browse on Gemini 3, its latest AI model, incorporating research from Project Mariner: the company’s experimental initiative focused on autonomous AI agents.

Rather than relying on Application Programming Interfaces (APIs) or specialised integrations, Auto Browse simulates keyboard-and-mouse interactions within Chrome. In practice, this means scrolling pages, clicking links, opening tabs, and inputting information in a way that mirrors human browsing behaviour.

Chrome visually marks AI-controlled tabs so users can see where the agent is active. Users can also run multiple Auto Browse tasks simultaneously while continuing to browse in other tabs.

However, Auto Browse does not run locally on a user’s device. Google processes content from AI-operated tabs in the cloud, where Gemini performs the reasoning and task assistance.

What does Google say about data, privacy, and control?

Because Auto Browse operates in the cloud, data handling becomes a central issue. Google says it has built Gemini in Chrome with “security and control by design”, and that agentic features such as Auto Browse are meant to keep users “in the loop”.

According to the company, Auto Browse pauses and asks for confirmation before carrying out sensitive actions, including making purchases or posting on social media. It has also framed deeper integrations as optional, stating that users can choose whether to connect apps and disconnect them at any time.

Google has also emphasised that Personal Intelligence, which will allow Chrome to remember context from past conversations to provide more tailored responses, will roll out gradually and only with user consent. “You’re always in control,” the company said, adding that users can decide what context Gemini can access.

At the same time, the company has not clarified whether it uses webpage content processed by Auto Browse to train AI models or how long it retains such data. It has only said that Gemini in Chrome operates under existing Gemini policies.

Notably, some risks have already surfaced in other AI agentic browsers. In December 2025, OpenAI said internal testing found that its browser-based AI agent could be manipulated through malicious instructions hidden inside webpages or emails, causing unintended actions.

Reflecting these concerns, analyst firm Gartner has advised organisations to block AI-powered browsers for now, warning that agentic AI features can prioritise user experience over security, and increase the risk of automated errors across logged-in accounts.

How does Auto Browse fit into Gemini’s expansion in Chrome?

Auto Browse forms part of a wider expansion of Gemini inside Chrome. Google has introduced a persistent Gemini side panel that allows users to interact with the AI agent alongside their active tabs, enabling multitasking without switching windows.

Chrome also supports Connected Apps, which allow Gemini to pull context from Gmail, Calendar, Maps, YouTube, Google Flights, and Google Shopping. As a result, Gemini can reference emails, schedules, and bookings while assisting with tasks such as travel planning or drafting messages.

Google has also added image editing directly inside Chrome through Nano Banana, an AI-powered image tool that lets users transform images from webpages using text prompts without downloading and re-uploading them.

Why is Google pushing agentic browsing now?

As AI systems mature, competition has shifted from standalone chatbots to controlling where user intent forms and how tasks are executed. Browsers play a central role in this shift because they remain the primary gateway to the open web.

OpenAI and Perplexity have responded by launching AI-native browsers such as Atlas and Comet last year. However, Google’s strategy differs: by embedding agentic AI capabilities directly into Chrome, it avoids the friction of user migration while positioning the browser as a place where discovery and task execution converge.

This approach also lets Google absorb the transition from traditional search to AI-generated responses within Chrome, keeping control over how users discover information and take action on the web.

What does this mean for trust and the future of browsers?

Agentic AI browsers demand a higher level of trust than traditional ones. With Auto Browse, users are asked to let an AI assistant help with actions across websites, accounts, and workflows. Over time, this could change how people relate to browsers, shifting them from tools users directly operate to systems they increasingly supervise.

And this, in turn, raises unresolved questions:

• How should responsibility be assigned when an AI-assisted action leads to an error or harm?
• How transparent are prioritisation and default choices when a browser takes actions rather than presenting options?
• How much autonomy are users willing to grant AI assistants that sit close to their identity, payments, and personal data?

Overall, these questions point to a broader transition in how web browsers function. Chrome is no longer positioning itself only as a window to the internet, but increasingly as a layer that understands intent and assists with execution, while stopping short of acting without users’ approval.

Ultimately, the AI browser wars are no longer about search quality alone. They hinge on who controls intent, action, and trust on the web, and Google’s decision to embed agentic AI capabilities directly into Chrome suggests it intends to shape that future from inside the browser itself.

Also Read: