Google Unveils Sweeping AI Advancements at I/O, Overhauling Search and Creative Tools

MOUNTAIN VIEW, Calif. – Google, at its annual I/O developer conference on May 20th, presented a significant array of new artificial intelligence tools and model upgrades, signaling a deeper infusion of AI across its services. Key announcements centered on the evolution of its Gemini AI models, a fundamental transformation of its search engine experience, and new generative AI capabilities for content creation and software development. The breadth of these announcements underscores a strategic direction where AI is not merely a feature but a foundational element reshaping how Google's core products operate and how users will interact with them.  

Gemini 2.5: The Upgraded AI Engine

At the heart of many new capabilities are enhancements to Google's Gemini 2.5 series of AI models. The Gemini 2.5 Pro model is set to receive an experimental enhanced reasoning mode called Deep Think, designed to tackle highly complex problems, particularly in areas like mathematics and coding. This development suggests a push towards AI capable of more sophisticated analytical tasks. Simultaneously, updates to the lighter-weight Gemini 2.5 Flash model focus on improved reasoning, multimodality, coding, and handling extensive information, all while being more cost-effective. This dual approach indicates Google is tailoring AI models for diverse needs, from high-end research to broader, high-volume applications. Furthermore, Gemini 2.5 models are gaining expressive, human-like speech generation through a new Live API, aiming for more natural AI interactions. For developers, Google is providing features like thought summaries and thinking budgets within the Gemini API, encouraging wider adoption and innovation on its platform.  

Search Reimagined: AI Mode and Beyond

Google Search is undergoing a significant overhaul with the wider rollout of AI Mode in the United States, now powered by a custom Gemini 2.5 model. This feature offers a conversational, ChatGPT-like interface, allowing for follow-up questions and more intricate interactions.  

Within AI Mode, Google is introducing several new capabilities. A Deep Search option will allow for in-depth exploration of complex topics, generating detailed, cited reports. Live Search will enable users to conduct searches using their device's live camera feed, asking questions about their surroundings. Perhaps most notably, AI Mode will gain agentic features, empowering the AI to perform tasks such as purchasing event tickets or booking appointments directly from the search interface. This includes new agentic shopping tools like virtual try-on using personal photos and AI agents to track prices and automate purchases. This move signals a shift for Google Search from primarily an information provider to an action-oriented assistant, aiming to complete tasks for users within its ecosystem.  

Alongside these changes, AI Overviews, the AI-powered summaries appearing at the top of search results, will expand to over 200 countries and more than 40 languages. While these enhancements aim to improve user experience, the increased capacity for direct answering and task completion within Search has led to observations that it may reduce click-through rates to external websites, a concern previously noted with AI Overviews.  

Empowering Creativity and Development

Google also unveiled a suite of new AI-powered tools for creators and developers. Flow is a new AI filmmaking tool built with the Veo, Imagen, and Gemini models, offering camera controls and a scene builder to simplify video creation. It is complemented by Veo 3, an advanced AI video generation model capable of creating realistic video sequences with native audio generation, including environmental sounds and character dialogue—a feature that distinguishes it from some competitors. Imagen 4, an upgraded image generation model, promises more detailed images, improved tonality, and better text rendering. These tools collectively lower the barrier to entry for high-quality media production.  

For developers, Jules, an autonomous coding agent, was introduced in public beta. It is designed to read code, understand intent, and perform tasks like writing tests, building features, and fixing bugs, integrating directly with GitHub. This represents a step beyond AI as a simple "co-pilot" towards AI as an autonomous partner in software development. Additionally, Gemini Code Assist is now generally available for individuals and GitHub, powered by Gemini 2.5.  

AI Expanding to New Realities and Interactions

The company demonstrated how AI is moving into new forms of interaction and hardware. Google Beam, formerly Project Starline, is an AI-first 3D video communication platform that uses AI to transform 2D video streams into realistic, immersive 3D video calls, with HP as an initial hardware partner. Google also detailed its Android XR platform, which will bring Gemini's AI capabilities to smart glasses and headsets for hands-free conversational AI, image capture, and device control. Samsung's upcoming Project Moohan smart glasses will utilize this platform. These initiatives suggest Google envisions AI, particularly conversational and multimodal AI, as the primary interface for next-generation hardware.  

AI Embedded Across Google's Ecosystem and Subscriptions

AI enhancements are being more broadly integrated across Google's services. For desktop Chrome users with AI Pro or Ultra subscriptions, Gemini will be available to clarify complex information or summarize web pages, initially for English speakers. Google Workspace is also receiving Gemini-powered features, such as personalized smart replies in Gmail that adapt to a user's writing style and AI-assisted drafting in Google Docs.  

To access the most advanced features, Google introduced the AI Ultra plan, a new premium subscription tier priced at $249.99 per month in the US. This plan offers the highest usage limits for AI tools, early access to models like Veo 3, 30TB of storage, and access to tools such as Flow and Project Mariner. This tiered subscription model indicates a strategy to monetize Google's most advanced AI capabilities, targeting power users and enterprises.  

Addressing AI Content Authenticity

In a move to address concerns about AI-generated content, Google launched the SynthID Detector. This new portal aims to help identify AI-generated images, videos, audio, and text created with Google AI tools by detecting an imperceptible SynthID watermark. The launch of this tool alongside powerful new generative capabilities like Veo 3 and Imagen 4 suggests an effort to provide mechanisms for transparency regarding AI-generated media.  

The announcements from Google I/O on May 20th collectively underscore a company-wide commitment to embedding advanced AI into nearly every facet of its product lineup, aiming to redefine user interaction, content creation, and developer productivity. The overarching theme is an AI that is more powerful, increasingly agentic, and accessible, albeit with premium options for its most advanced capabilities.