Google Unveils Gemini Spark and Agentic AI Tools at Google I/O

Google has unveiled a significant expansion of its artificial intelligence capabilities, headlined by the introduction of Gemini Spark, a proactive AI assistant designed to manage routine tasks on behalf of users. Announced at the company’s annual Google I/O developers conference, these advancements signal a strategic shift toward "agentic" AI, which focuses on systems capable of performing actions rather than simply responding to queries.

The Rise of Agentic AI

Gemini Spark, powered by the new Gemini 3.5 model, is designed to operate in the cloud, allowing it to function in the background even when a user’s device is locked or inactive. The assistant can sort through emails, meeting notes, and chats to generate documents summarizing key takeaways and action items. To address security concerns, Google noted that the agent is programmed to request explicit user permission before executing high-stakes actions, such as making purchases or sending emails. A beta version of Gemini Spark is rolling out to select testers and U.S.-based Google AI Ultra subscribers, with plans to integrate the agent directly into the Chrome browser later this summer.

Gemini 3.5 and Enhanced Media Creation

Google is also rolling out its Gemini 3.5 model family, starting with Gemini 3.5 Flash, which now serves as the default model for the Gemini app and AI mode in Google Search. The company describes this as its strongest coding and agentic model to date, boasting speeds four times faster than some competitors. Additionally, the new Gemini Omni model enables users to generate high-quality, realistic video from text, image, video, or audio inputs. These videos can be edited through natural conversation with the model. To ensure transparency, all content created by Omni will include Google’s SynthID digital watermark, and the company is introducing a content credentials verification tool to help users identify AI-generated media.

Intelligent Search and Shopping

Google is implementing the most significant update to its search box in 25 years, transitioning to an intelligent interface that adapts to longer queries and provides AI-powered suggestions. This new search box supports multi-modal inputs, including text, images, video, files, and Chrome tabs. The update is currently launching in all countries and languages where AI mode is available.
Furthermore, the company announced the Universal Cart, an intelligent shopping tool that functions across various services, including Google Search, YouTube, Gmail, and Gemini. Once an item is added to the cart, the system uses Gemini models to track price drops, provide historical pricing data, and notify users when out-of-stock items become available. The Universal Cart is scheduled to arrive on Search and the Gemini app this summer, with integrations for YouTube and Gmail to follow.