Google I/O Unveils AI Agents That May Undermine Its Own Ad Revenue

Google makes $200 billion a year selling advertisements against human attention. Its newest AI products are designed, in part, to eliminate the browser visits that make those ads worth buying.

That contradiction is the thing nobody at Google — or among its advertisers — has publicly resolved. At Google I/O last week, the company unveiled a suite of AI agents that book travel, monitor prices, and handle cancellations without returning to a search results page. Every time an agent completes a transaction in the background, the display ad that would have appeared on that page does not fire. The cost-per-click metric that Google uses to charge advertisers for their spend does not register.

The $180 billion Google is spending this year on capital expenditures is, in one reading, a bet that it can build the infrastructure to stay indispensable. In another reading, it is a bet that the advertising model adapts before the agents eat it.

Google's search agents represent a deliberate attempt to have it both ways. The company remains the interface — but the interface is now a task execution layer rather than a results page. An agent assembled from Gemini 3.5 Flash, Antigravity, and AI Mode does the looking. Google still sits in the middle, still owns the query, still controls the APIs the agent calls. The company is moving from owning the search box to owning the reasoning layer underneath it.

That is the case for why Google survives the transition intact. Every query, even an agentic one, is still a Google query. The agents run on Google's infrastructure and surface Google's synthesized results.

The counterargument is where the story lives. When a purchasing agent already knows a person's size, budget, and preferred retailers, the brand that wants to reach that person does not need to advertise on a results page. The agent makes the decision, not the click. The trace of human attention that the entire advertising model rests on — the click, the impression, the moment a person was watching — disappears when the agent acts on the user's behalf.

Google's own product choices reflect this tension. The persistent background agents announced at I/O — 24/7 processes that monitor topics across the web and synthesize updates without prompting — are explicitly designed to reduce the number of times a user opens a browser. That is the stated product goal. It is also, potentially, an act of self-cannibalization that the advertising sales team has not fully accounted for.

The numbers Google reported at I/O tell a consistent story: Gemini crossed 900 million monthly users, AI Mode surpassed 1 billion monthly users within a year of launch, and daily API requests are growing roughly sevenfold. Google's capital spending will reach $180 billion to $190 billion this year, up from $31 billion in 2022 — a sixfold increase in four years. Those are not the numbers of a company defending a product. Those are the numbers of a company buying its way into a future it is not sure it can reach organically.

No Google executive has addressed the advertising tension publicly in these terms. Earnings calls talk about AI expanding engagement and growing the total addressable market for search. They do not address what happens to cost-per-click when the entity clicking is a machine.

The information agents roll out to Google AI Pro and Ultra subscribers this summer. The broader rollout, including agentic booking and business-calling features, follows for all US users. Whether the agents drive more human engagement or fewer is the open question that will determine whether Google's biggest bet in a decade is a moonshot or a managed decline.

The premise has shifted either way. Google is no longer arguing that search will remain what it was. It is arguing that it will remain the layer underneath whatever search becomes next. That is a different claim — and a harder one to prove.

Research for this story drew on Google's official I/O 2026 announcements, the Forbes account of the keynote, and independent technical analysis of the Gemini agent stack from Ken Huang's Substack. Capex figures are Google's own guidance as reported at the event.