Agentic search is not a future concept at Google. It is already in progress. In an interview with Alex Heath on the Sources podcast, reported by Search Engine Journal on October 22, Google CEO Sundar Pichai described a near-future where Search and Gemini do not merge into a single product but are unified by an agentic layer (a persistent AI system that runs continuously, maintains context, and takes actions on behalf of users) that operates across both.
What did Pichai actually say about Search and Gemini combining?
Asked directly whether Search, AI Overviews (Google’s AI-generated answer panels in search results), and Gemini would ever fully combine, Pichai said he does not frame it as converging different products. His answer: the agentic layer handles that convergence on its own.
“I expect our agents to provide that unifying experience across these products,” Pichai said in the interview, as reported by Search Engine Journal. “In fact, that agentic layer, rather than try to converge or think about different products and how they come together, I think the agentic layer naturally ends up doing that.”
He described the mechanism: agents run in the cloud, 24 hours a day, carry user context across sessions, and pick up tasks wherever the user left off, whether they started in Search or in Gemini.
What is the “one box” vision?
Pichai’s stated goal for Gemini is to make it “personal, proactive, and powerful.” Gemini already has both chat and task capabilities, he told the interviewer, and the near-future vision is “one box,” a single interface that handles both conversational queries and the execution of actions.
He cited Google Spark, introduced at Google I/O, as the direction Google is moving. Spark is a cloud-based personal agent that integrates Search and Chrome, using information agents in Search to conduct web research on a user’s behalf.
How does the enterprise Gemini agent connect to consumer Search?
Pichai clarified that enterprise and consumer Gemini share the same underlying architecture: “the same model, same harness, same runtime.” The difference is context, not capability. He described the enterprise setup as provisioning a micro VM (a small, isolated virtual machine) for each task, running securely around the clock with real-time monitoring and full auditability.
“We’re bringing all that to consumers as well,” he said, according to Search Engine Journal. The implication: consumer-facing agentic search will eventually operate with the same execution infrastructure currently reserved for enterprise.
What is driving the shift to more agentic queries?
Pichai said the move is user-driven, not product-led. Google’s own usage data shows the share of agentic-style queries in Gemini has been “steadily increasing over time.” These capabilities have so far been available mainly to Ultra and Pro users, with broader rollout planned over the next few months, Pichai said. He noted that the newer Argon models “deliver a huge step change.”
What this means for AI-search visibility
The clearest business implication: an AI agent that runs continuously and executes tasks on your behalf does not behave like a person typing into a search box, and brands that treat both as the same visibility problem may be caught off-guard.
Consider the difference in practice. A user today searches for “best project management tools,” clicks a result, reads the page, and decides. In the agent model Pichai describes, the user might instead say to Gemini: find me the best project management tool and set up a free trial. The agent researches, evaluates, and acts. There is no guarantee it surfaces the same sources a search result would, and no guarantee there is a click to your site at all.
In Hingewise’s view, this separates brands into two groups: those whose content is structured so that AI agents can read, interpret, and act on it, and those whose content is optimized only for human readers arriving from a search link. The distinction matters because an agent selecting a source is unlikely to reward click-appeal; it is more likely to reward clarity and machine-readability.
Pichai did not say how agents will choose sources, so which pages they favor for retrieval (the process by which an AI pulls information to complete a task) remains an open question.
One implication the interview does not address: when agents run persistently and maintain user context, there is no discrete “moment of search” where a brand can appear. In Hingewise’s view, an agent’s accumulated context about trusted sources may then matter more than any single query, which could give brands already cited in these systems an advantage.
Hingewise’s position: being “found” in traditional search and being “selected by” an agent are not the same outcome and should not be treated as the same goal. The agentic layer Pichai describes is a distinct visibility problem that requires a distinct strategy.
Quick checklist: is your brand ready for agentic search?
- Does your key content answer specific questions directly, not just describe features or categories?
- Can a machine extract a clear, actionable summary from your product or service pages (name, category, key differentiator)?
- Is your brand already cited in AI-generated answers in Search or Gemini for your core topic area?
- Does your content include specific, verifiable data (numbers, dates, named sources) rather than general claims?
- Have you tested how Gemini or other AI assistants describe your brand when asked about your category?
- Is there structured markup on your pages that helps agents identify what the page is about and what action it supports?
What to watch
Pichai indicated the broader rollout beyond Ultra and Pro users is planned over the next few months. How Google handles referral traffic as agents absorb more of the decision-making process, a question the Search Engine Journal article raises but Pichai did not address directly, will be one of the more closely watched details when that wider rollout begins.
Source: Search Engine Journal, October 22, 2026
Lam Nguyen · Hingewise
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.