At Google I/O on May 19, 2026, the company announced what VP of Search Elizabeth Reid called “the era of Search agents” — a pivot from retrieval-based engine to active, agentic platform. Four months into the rollout, the structural significance is becoming clearer: the same opinionated-default pattern that reshaped developer infrastructure this year has now reached the world’s most-used search interface.
Three Pillars of Agentic Search
Google’s strategy rests on three announcements. First, information agents — persistent, 24/7 background monitors that reason across the web, blogs, social posts, and real-time data feeds, then deliver synthesized updates with the ability to take action. A user hunting for an apartment can describe their requirements and the agent continuously scans listings, notifying them when matches appear. Information agents launch first for Google AI Pro and Ultra subscribers this summer.
Second, expanded agentic booking. Google Search now aggregates real-time pricing and availability for local experiences and services, with direct links to complete booking through the provider of choice. For select categories — home repair, beauty, pet care — Google can call businesses on the user’s behalf. These capabilities roll out to all US users this summer.
Third, agentic coding brought directly into Search. Powered by Gemini 3.5 Flash and Google Antigravity, Search now builds custom generative UI — visual tools, simulations, dashboards, trackers — on the fly, tailored to the user’s query. The generative UI capabilities are available to everyone in Search this summer, free of charge. Custom experiences via Antigravity — persistent mini-apps users return to over time — will follow in the coming months, starting first for AI Pro and Ultra subscribers in the US.
Flash as the New Default
Underpinning all three pillars is Gemini 3.5 Flash, which Google set as the new default model for AI Mode globally. Sundar Pichai claimed Flash is “four times faster than other frontier models” while delivering frontier-level capabilities at less than half the price. The company also announced that AI Mode has surpassed 1 billion monthly active users, with AI Overviews reaching 2.5 billion and Search queries hitting an all-time high.
Google’s infrastructure investment makes this scale legible. Pichai noted capex of approximately $180 to $190 billion in 2026, up from $31 billion in 2022 — a sixfold increase in four years, anchored by the eighth generation of custom TPU silicon.
The Pattern Reaches Consumer Scale
The reason this matters for agent infrastructure is not the features themselves but the architectural pattern they represent. This publication has documented the opinionated-default pattern across the stack this week: AWS Strands, DigitalOcean, and Aiven shipped opinionated agent harnesses at the cloud provider level. LangChain shipped the same pattern at the framework layer. Docker shipped it at the container runtime layer. Google’s own AX shipped it at the orchestration layer.
Search agents represent the pattern reaching its ultimate distribution surface. When agents are the default search experience for over a billion users, the question is no longer whether opinionated defaults win — it is what happens when the world’s information gatekeeper is also the world’s default agent platform.
The Click-Driven Web Problem
Industry analysts have framed this shift as structural rather than incremental. Visibility is moving from blue-link rankings to AI-composed surfaces — information-agent outputs, generative UI, agentic booking flows. Businesses that have optimized for traditional search engine visibility now face a platform that mediates rather than directs traffic. The economics of the web — built on the assumption that search sends users to websites — are being rewritten in real time.
The self-report caveat is important here: all adoption metrics (1 billion AI Mode users, 2.5 billion AI Overviews users, queries at all-time highs) are Google’s own numbers. The actual efficacy of information agents, the reliability of agentic booking, and the quality of generative UI in production remain to be evaluated as the rollout expands beyond its initial gated access. The availability caveats matter too — most features are rolling out “this summer” to US users, with information agents and Antigravity gated to paid subscribers.
What the Stack Needs Next
As this publication has argued, the rapid proliferation of autonomous agents performing real-world actions — booking appointments, calling businesses, managing background monitoring tasks — makes an agent governance stack not optional but essential. Google’s own Antigravity expansion from coding environment to “a platform for developing and managing cohorts of autonomous AI agents” confirms the direction. The infrastructure pattern is now standardized at every layer. The governance, accountability, and trust infrastructure to support it at consumer scale is still forming.