Google I/O 2025: Major AI News - Gemini 2.5, Project Astra, and a Complete Overview of All New Features
At its I/O 2025 developer conference, Google unveiled a range of groundbreaking developments in artificial intelligence that fundamentally change how we will work with technology. This year's keynote focused almost exclusively on AI innovations across the entire Google ecosystem. Let's take a detailed look at all the major new features the company announced.
AI Mode in Google Search - A New Era of Search
Google introduced an entirely new "AI Mode" for the Google Search engine, which is now available to all users in the US. This mode uses the most advanced Gemini 2.5 model and delivers a more intelligent and agentic search experience. In AI Mode, users can ask longer and more complex questions than ever before, receive detailed AI-generated overviews, seamlessly follow up with additional questions, and take advantage of deeper research capabilities, including visualization and live multimodal responses. According to the Google Search blog, these advanced features will gradually be integrated into the core search experience for all users. "AI Mode is available through a separate tab in Google Search and provides a space where we can implement some of our most advanced AI features for complex searches," states the official Google blog. In this mode, the search engine offers much more contextual answers and can maintain conversational continuity, enabling natural dialogue when searching for information.
Google has not yet announced the exact date when AI Mode will become available in the Czech Republic. According to available information, it will gradually be rolled out to additional countries after its successful launch in the US. However, AI Overviews are already fully available in the Czech Republic. These overviews generate summarized answers to more complex queries directly in Google search results. The feature is active in more than 200 countries, including the Czech Republic, and supports over 40 languages, including Czech. When users enter a more complex query, they automatically receive a generated overview above the standard search results, with links to relevant websites.
Gemini 2.5 - The New Generation of AI Models
The Gemini 2.5 Pro model now serves as the foundation for many new products and developer tools. Google has significantly improved the Gemini API for faster prototyping through Google AI Studio, where developers can generate web applications directly from text, image, or video prompts. The new SDKs support the Model Context Protocol (MCP), making integration with open-source tools easier. Gemini 2.5 delivers significantly improved contextual understanding and multimodal processing capabilities, enabling more sophisticated interactions between users and AI.
Project Astra - A Visual Assistant for the Real World
Project Astra, a mobile visual assistant that understands the context of the real world through the device's camera, has received significant improvements. According to information from the official Google DeepMind blog, demonstrations during the keynote showed Astra providing real-time advice (for example, while repairing a bicycle) through natural dialogue while simultaneously visually analyzing its surroundings. Astra is being rolled out today on both Android and iOS, and its visual recognition capabilities are also being added to AI Mode in Google Search. Project Astra can identify objects, read text from images, and provide relevant information based on what it sees, making it a powerful tool for everyday use. "Project Astra represents a significant step in our vision of creating a visual assistant that truly understands the world around you," Google DeepMind states on its blog. The system combines advanced computer vision with natural language processing to provide useful and contextually relevant information in real time.
Generative Media: Imagen, Veo, and Flow
Google also presented its most advanced generative models - Imagen 4 for image generation and Veo 3 for video generation. These models are now much more capable of creating realistic and relevant content based on text descriptions. Another new development is the Flow application, which is aimed at creators and enables rapid video production using the generative capabilities of the Veo model. According to an article from The Verge and the official Google blog, these tools dramatically lower the barriers to creating professional-looking media content and open up new possibilities for creative expression. "Flow allows creators to make videos simply by describing what they want to see, and then edit details using simple text instructions," Google states in an official blog post.
Agent Mode and More Agentic AI Experiences
Google introduced Agent Mode within Gemini, which enables tasks to be performed in the background, such as searching for rental offers or planning trips based on personal preferences. This mode demonstrates how Gemini can act proactively rather than merely respond reactively to queries. "Agent Mode represents a shift from simply answering questions to actively solving complex tasks," Google states. The system can work with many different information sources simultaneously, retain context, and perform multi-step tasks that would otherwise require considerable manual effort.
Project Mariner - Dynamic Search Filters
Project Mariner from Google DeepMind is a new tool that helps dynamically refine search filters during tasks such as apartment hunting. According to information from the official DeepMind blog, Project Mariner analyzes user preferences, monitors their feedback, and continuously adjusts search parameters so that the results better match their actual needs. "Project Mariner represents a new approach to searching for complex information, where user preferences cannot always be clearly defined in advance," explains Google DeepMind. The system learns from interactions and continuously improves its results, leading to a much more efficient decision-making process.
Android XR and Smart Glasses Integration
Google officially unveiled Android XR with partners such as Samsung and Xreal. Gemini-powered smart glasses can provide live translation directly on the device, answer questions about what you see around you, navigate cities using Maps overlays, and even help find cafés based on dietary preferences. According to an article from The Verge, "Google emphasizes that AR glasses powered by Android XR will be able to process advanced AI tasks directly on the device, ensuring privacy and fast response times." This integration represents a significant step toward ubiquitous augmented reality in everyday life.
Audio and Multimodal Improvements
Gemini 2.5 Flash Native Audio delivers improved conversational flow for voice interactions across devices in several languages. This feature enables more natural and fluid interactions with voice assistants and significantly improves the quality of conversations with AI. "Our latest audio-processing models in Gemini 2.5 achieve unprecedented quality in understanding human speech and generating natural responses," Google states, suggesting that future voice interactions with technology will be much more natural and intuitive. Czech is among the more than 24 languages supported by the new native audio output feature.
Google Shopping with AI and Virtual Try-On
Google has significantly enhanced its shopping platform, Google Shopping, with AI features. According to the official Google Shopping blog, an AI shopping mode was introduced to help customers more easily find exactly what they are looking for, even when they do not have a clear idea of a specific product. Another major new feature is the expansion of virtual try-on capabilities, which now support many more types of products. "Thanks to our advanced computer vision and machine learning models, customers can now virtually try on clothing, glasses, makeup, and other products with a high degree of fidelity and realism," states the Google Shopping blog. The system also integrates product reviews and ratings more effectively, with AI helping extract the most relevant information from user reviews and present it in a clear format. Users can use Czech localization when setting up their account, managing product feeds, or troubleshooting issues.
Google Beam - A New Way to Share Content
Another new development is Google Beam, which introduces a new way to share content between devices. According to the official Google blog, this technology enables seamless sharing of URLs, text, images, and other data between different devices simply by pointing a phone. "Google Beam combines ultra-wideband (UWB), Bluetooth, and visual recognition to create an almost magical content-sharing experience," Google states. This technology is part of Google's broader vision of connected devices that communicate with one another naturally and intuitively.
Google's Future Lies in AI Features
Google I/O 2025 clearly demonstrated that Google's future is deeply intertwined with advanced generative artificial intelligence. The company has transformed everything from search and creative workflows to the way we interact with devices in everyday life. With the Gemini 2.5 model serving as the foundation for many new Google products and features, the company is clearly charting a course toward advanced AI as a key pillar of its strategy. Whether it is more agentic search, visual assistants such as Project Astra, or augmented reality through Android XR, Google is bringing AI into every aspect of digital life. All these innovations signal a significant shift away from traditional human-computer interactions toward much more natural, context-aware, and proactive systems that not only respond to our queries but actively help us achieve our goals.
