At I/O 2026 in May, Google unveiled what it calls the largest overhaul of its Search in the service’s 25-year history. The revamped search box will process text, images, documents, videos and open Chrome tabs, and return synthesized answers in place of ranked link lists. Liz Reid, Vice President of Google Search and head of the Search division, said query volumes hit a record last quarter and that AI Mode, one year after launch, now reaches one billion monthly users, with queries doubling each quarter.
Robby Stein, Google Search product lead, says users will reliably receive AI summaries for natural-language queries, and the broader aim is to remove friction between traditional Search and AI Mode to streamline the user experience. Users can move into AI Mode by asking follow-up questions in an AI overview or by attaching documents, photos, videos and Chrome tabs directly to the search box.
Information agents: continuous background monitoring
Google also announced so-called information agents: AI services that continuously monitor user-specified topics — stock prices, housing markets or sports results — and alert users when something significant happens. TechCrunch describes these as the next-generation Google Alerts, enhanced with an argumentative layer capable of synthesizing opposing viewpoints.
The feature is slated to launch this summer in the United States, initially for Google AI Pro and Ultra subscribers.
Gemini Spark: an always-on agent
Google introduced Gemini Spark, an always-on AI agent built on Gemini 3.5 and the Google Antigravity framework. Spark can run from the cloud even when a user’s device is locked or offline. It deeply integrates with Workspace apps (Gmail, Docs, Slides), automates repetitive workflows, summarizes meeting material and can draft emails on a user’s behalf.
Mashable notes that Spark can connect via the MCP (Model Context Protocol) to more than 30 third-party services, including Adobe, Asana, Dropbox, Uber, Zillow and Zocdoc. MCP is an open standard that provides a single connector for AI models and external services; Anthropic developed the protocol in 2024 and it has since seen broad industry adoption.
Josh Woodward, vice president of Google Labs, characterized Spark’s autonomy as similar to giving a teenager their first bank card: usage is governed by constraints and limits that Google will refine over the year. In enterprise deployments Spark runs in a fully managed, secure environment on Google Cloud — tasks execute on fresh, isolated virtual machines and traffic passes through a gateway that enforces data-protection rules.
Rewriting commerce: Universal Cart and Agent Payments Protocol
Google unveiled Universal Cart, a unified shopping interface built on Google Wallet that works across Search, the Gemini app, YouTube and Gmail. The system automatically tracks price drops and inventory, notifies users when items return to stock, and can surface hidden savings using Google Wallet loyalty points and payment offers. Universal Cart is powered by Google’s Shopping Graph, which contains 60 billion products.
Behind the commerce flow is the Agent Payments Protocol (AP2), a standard that creates verifiable connections among the user, the merchant and the payment processor and ensures an AI agent acts only within user-defined limits. Mashable reports that for now every transaction requires user approval and AP2 provides a digital audit trail for disputes and refunds.
Google previously launched the Universal Commerce Protocol with Amazon, Shopify and Walmart to give AI assistants a common language without bespoke integrations. Universal Cart will roll out this summer in the United States and will arrive soon in Canada, Australia and the United Kingdom.
Models and financial context
Google says the Gemini 3.5 Flash model is four times faster than competing frontier models and is currently the company’s strongest agentic and coding model; it has become the default in the Gemini app and Search AI Mode globally. A higher-performance Gemini 3.5 Pro is expected next month. DeepMind and Google’s in-house AI infrastructure helped accelerate training for deeper reasoning capabilities.
These product moves come against a strong revenue backdrop: Google Search and advertising revenue reached $60.4 billion in the first quarter, up 19 percent year-over-year, which the company says undercuts concerns that AI-generated answers would undermine its search ad business.
Other announcements and timing
I/O 2026 also featured additional launches: Gemini Omni, a multimodal video-generation model; Google Pics, an AI-based image editor; Daily Brief, a personalized morning AI summary; AR smart glasses built on Android XR with Warby Parker and Gentle Monster; and Docs Live, voice-driven document editing. The smart glasses will offer real-time translation, navigation and camera-based AI queries and will be compatible with Android and iOS.
Most features will roll out gradually, typically over the summer and fall, initially in the United States and first to premium subscribers.
