On January 28, Google introduced key updates to Chrome that integrate its Gemini AI model more deeply into the browser. It aims to help users multitask, automate workflows and complete tasks across the web more efficiently.
The company said Gemini in Chrome is now built on Gemini 3 and will roll out on macOS, Windows and Chromebook Plus. The update introduces a persistent side panel that allows users to interact with Gemini without switching tabs.
Parisa Tabriz, vice president of Chrome, described the new Gemini in Chrome as akin to having an assistant that makes it easier to discover information and complete tasks on the web than ever before.
The side panel enables users to compare information across tabs, summarise content such as product reviews and manage scheduling tasks while keeping their primary webpage open. The company said testers have used the feature to reduce tab switching and handle parallel tasks.
Chrome is also adding image transformation features through Nano Banana. Users can modify images directly from the browser by typing prompts into the Gemini side panel, without downloading or re-uploading files. Google said this can be used for design exploration or creating visuals such as infographics.
Gemini in Chrome now supports Connected Apps, allowing it to work across services including Gmail, Calendar, YouTube, Maps, Google Shopping and Google Flights. According to Google, this enables Gemini to pull context from emails, calendars and travel data to assist with tasks such as booking flights or drafting emails.
In the coming months, Google plans to bring ‘personal intelligence’ to Chrome. The feature, which requires user opt-in, allows Gemini to retain context from past interactions to provide more tailored responses. “With Personal Intelligence, you’re always in control,” Tabriz said, adding that users can connect or disconnect apps at any time.
For paid AI Pro and Ultra subscribers in the US, Google is also introducing Auto Browse, an agentic browsing feature that can handle multi-step tasks.
Auto browse can research travel options, fill out forms, collect documents and manage subscriptions. With user permission, it can also sign in using Google Password Manager.
Google noted that auto-browse will pause and prompt confirmation for sensitive actions, such as purchases or social media posts.
The company also announced that Chrome will support Google’s Universal Commerce Protocol, an open standard for agent-driven online commerce developed with partners including Shopify, Etsy, Wayfair and Target. Google said the protocol is intended to allow AI agents to complete transactions on behalf of users within the browser.
Google emphasised that the new features are designed with added security controls and user oversight. “Auto browse is designed to pause and explicitly ask for your confirmation,” the company said.
The updates position Chrome as a more active platform for AI-assisted browsing, as Google expands Gemini’s role from information retrieval to task execution within the browser.
Google has also introduced Agentic Vision, a new visual reasoning capability in Gemini 3 Flash that allows the model to actively inspect images using code execution to reduce visual errors and improve accuracy across vision tasks.
According to Google DeepMind, Agentic Vision shifts image understanding from a single-pass process to an iterative one, enabling the model to zoom into details, manipulate images, and ground responses in visual evidence. The company said enabling code execution delivers a “consistent 5–10% quality boost across most vision benchmarks.”
Agentic Vision operates through a Think, Act, Observe loop. The model first analyses the image and query, then generates and executes Python code to crop, rotate, annotate, or compute visual data. The transformed images are reintroduced into the model’s context to inform the final response.
The feature is available through the Gemini API in Google AI Studio and Vertex AI, and is beginning to roll out in the Gemini app under the ‘Thinking’ model option.
ALSO READ: GitHub Introduces Copilot SDK to Embed AI Agents in Applications
