Google on Tuesday used its annual I/O developer conference to announce the deepest restructuring of its search product in decades, embedding AI agents directly into the query box and releasing a new lightweight model called Gemini 3.5 Flash, as the company moved to keep its core advertising engine from being outflanked by standalone AI chatbots.
The announcements signal how acutely Google views the rise of AI chatbots as a threat to the franchise that generates the bulk of its revenue. Search is the cash cow that funds Google's sprawling empire, according to Axios, and the company is moving proactively to upend its own core business before someone else does.
What shifted
The headline product change is the expansion of the search box itself to accept longer queries and back-and-forth exchanges — what Google billed as the biggest change to the search box since its debut. More consequentially, Google is embedding standing queries into search results: a user who wants alerts whenever a favorite band announces nearby concert dates can now set one up directly from the search interface, with an AI agent monitoring the question over time.
"Agents in search is the next step," Google DeepMind chief Demis Hassabis said. "One of the cool things we get to do here at Google is build technologies that get immediately deployed into multi-billion dollar products."
Google also announced Gemini Spark, a general-purpose personal assistant inside the Gemini app capable of reasoning across connected applications and taking action on a user's behalf. Spark launches in beta next week for Google AI Ultra subscribers and a set of trusted testers.
The model lineup
On the model side, Gemini 3.5 Flash becomes the default for the Gemini app and AI mode in search globally. CEO Sundar Pichai described it as "remarkably fast" and said it delivers cutting-edge performance at roughly half — and in some cases one-third — the cost of comparable frontier models. Google said the updated model is "less likely to generate harmful content and mistakenly refuse to answer safe queries."
A heavier version, Gemini 3.5 Pro, is in internal use and is expected to reach wider availability next month. Google also unveiled Omni, a world model that simulates physical environments based on user actions, with applications in video editing, image generation, YouTube Shorts and the company's Flow product.
On the hardware front, Google said AI smart glasses co-developed with Samsung, Warby Parker and Gentle Monster will be available this fall. Hassabis described the broader wearables push as a way to move AI beyond screens — "It's a really exciting use case for where AI can get out into the real world," he said.
The counterpoint
Today's coverage comes entirely from center-leaning outlets — CNBC and Axios — and no critical or dissenting voice appears in the source material. Missing from the dossier is any assessment from independent researchers or rival companies about whether Google's agents represent a genuine capability leap over offerings from OpenAI or Anthropic, both of which are preparing for IPOs and have attracted more market attention in recent months. Also absent is any examination of the publisher and advertiser displacement that agent-driven search entails: if a standing query answers a user's recurring question automatically, the downstream click — and the ad revenue attached to it — may never materialize. An editor seeking balance should consider adding sourcing on those two angles before publication.
Google's next major milestone is the broader release of Gemini 3.5 Pro, expected in June, which will give analysts a fuller picture of where the company's models stand against the frontier.