AI Mode becomes central to Google Search: “AI Mode,” a conversational search feature Google introduced in 2025, is moving from an experimental option to a core part of the product.
These features are built on a system called “Antigravity,” which creates persistent dashboards and lightweight task interfaces inside Search.
Search Live introduces real-time multimodal interactions: Google also expanded Search’s multimodal capabilities with “Search Live,” a feature powered by Project Astra.
The system lets users interact with Search in real time through their phone cameras, enabling Search to interpret live video and respond conversationally.
This feature builds on Google DeepMind’s previous work on multimodal AI assistants that integrate understanding of text, images, audio, and video.
Read the Google blog post here.
Google has announced a broad redesign of Search in over 25 years, featuring an AI-powered interface, autonomous task execution, deeper integration with Google apps, and tools for building customised “mini apps” within Search. These updates were unveiled at the company’s annual developer conference, Google I/O 2026.
These changes further shift Search from a list of links to a conversational, AI-driven assistant powered by Gemini models. Google is adding what it calls “agentic” capabilities, allowing users to request multi-step tasks instead of just retrieving information.
AI Mode becomes central to Google Search: “AI Mode,” a conversational search feature Google introduced in 2025, is moving from an experimental option to a core part of the product. It runs on Google’s Gemini 3.5 Flash model and handles longer, more complex queries. Users can feed it documents, images, and browser tabs alongside their search prompts. The search box itself has been redesigned to accommodate back-and-forth conversations rather than short keyword queries.
AI Mode is now live in nearly 200 countries and supports 98 languages.
Search gains autonomous task execution features: Users will be able to ask Search to track projects, monitor updates, organize information, and interact with websites and services autonomously. Google demonstrated use cases such as wedding planning and moving house. These features are built on a system called “Antigravity,” which creates persistent dashboards and lightweight task interfaces inside Search. The feature will launch first for subscribers.
Integration with Gmail, Photos, and Calendar: Users can connect their Gmail and Google Photos accounts to Search, allowing it to pull in emails, receipts, travel bookings, photos, and other personal data to personalize responses. Calendar integration is planned but not yet available. Users can choose whether to enable these connections.
The broader use of personal data in Search will likely draw increased scrutiny over privacy practices and how the system stores and uses what it learns about users.
Search Live introduces real-time multimodal interactions: Google also expanded Search’s multimodal capabilities with “Search Live,” a feature powered by Project Astra.
The system lets users interact with Search in real time through their phone cameras, enabling Search to interpret live video and respond conversationally. This feature builds on Google DeepMind’s previous work on multimodal AI assistants that integrate understanding of text, images, audio, and video.
What does this mean for the web? Google is shifting Search from providing lists of links to generating direct AI answers. This change is already decreasing website and publisher traffic from Search. As zero-click experiences increase, advertisers and publishers that rely on search referrals face growing uncertainty.
Accuracy problems remain unresolved: AI-generated search results continue to be unreliable. Research published in 2026 found that generative search engines sometimes cite AI-generated or low-quality sources. Google’s AI Overviews have previously produced inaccurate summaries and buried original source material. Google says its systems are improving, but independent concerns about accuracy, citation quality, and transparency remain unresolved.
Read more:
