Google Replaces Static Search With 24/7 Information Agents

At Google I/O 2026, the company announced a structural shift in its core product, rebuilding Search around persistent AI agents. The platform is moving from reactive query resolution to continuous, proactive execution driven by Gemini 3.5 Flash. For developers and publishers, this signals an immediate shift in how automated systems interface with the open web.

Information Agents Replace Static Alerts

The centerpiece of the May 19 release is the introduction of information agents. These entities run continuous background sessions to monitor specified topics, synthesize developments across formats, and deliver actionable notifications.

Instead of relying on keyword-triggered alerts, these agents leverage Gemini 3.5 Flash for continuous reasoning across news, financial markets, and real-time social streams. The system evaluates conflicting perspectives and filters noise before pushing updates. If you build long-running AI agents, Google’s implementation provides a clear template for persistent task execution at scale.

Multimodal Inputs and Antigravity

Google replaced the traditional search box with a dynamically expanding multimodal input field. Users can now drop open Chrome tabs, files, video, and images directly into the prompt.

This interface acts as the frontend for Antigravity, Google’s new agentic application platform. Antigravity allows users to generate specific mini-apps, like a real-time moving tracker or automated wedding planner, through natural language. Search parses the intent, codes the logic, and deploys the tool immediately.

Alongside consumer tools, Google introduced Gemini Spark in beta. This agent integrates deeply into Google Workspace for proactive scheduling, email management, and automated booking.

The WebMCP Standard

To facilitate reliable interactions between these background agents and external data, Google proposed WebMCP. This open standard allows webmasters to expose structured server-side tools directly to visiting AI agents.

By adopting WebMCP, a real estate site can offer a JSON-based search function directly to an information agent, bypassing HTML scraping. This builds upon the underlying principles of the Model Context Protocol, formalizing agent-to-website transactions and creating a predictable pipeline for machine-driven traffic.

Pricing and Availability

Feature	Availability	Plan Requirement
AI Mode (Gemini 3.5 Flash)	May 19 (Global)	Free / All Users
Generative UI Dashboards	Summer 2026	Free / All Users
Information Agents	Summer 2026 (U.S.)	AI Pro / Ultra
Antigravity Mini-Apps	Summer 2026 (U.S.)	AI Pro / Ultra

Google simultaneously restructured its subscription tiers. The company introduced a $100 per month AI Ultra plan and lowered the ceiling on its top-tier enterprise plan to $200 per month. AI Mode query volume has already doubled quarterly, recently passing 1 billion monthly active users.

If you manage high-traffic web applications, the transition to agentic search requires immediate infrastructure planning. Implement structured tool endpoints via WebMCP to ensure persistent agents can reliably query your data without relying on brittle DOM scraping.