Summary: At Google I/O 2026 on May 19–20, Sundar Pichai declared the agentic Gemini era and called this year’s Search revamp the largest in more than 25 years, with agents placed directly into Search. The event also introduced the Gemini Spark assistant and Gemini 3.5 Flash. The article argues the change shifts sites from clicks toward uncertain citation or licensing mechanisms.
Google’s I/O 2026 developer conference ran May 19–20. In his keynote, CEO Sundar Pichai announced the “agentic Gemini era”: this year’s Search revamp — agents placed directly into Search — is billed as the largest in more than 25 years, and the same stage brought the Gemini Spark assistant and Gemini 3.5 Flash.
Search’s Biggest Wager in 25 Years
Search anchors Google’s business, and it is the asset the company is most reluctant to touch. That Google itself bills this as the biggest overhaul in over 25 years tells you how urgent things feel inside: generative AI has changed how people ask questions, and a Search that stands still will watch queries drift to newer entry points.
Agents inside the search box rewrite more than an interface. The old deal — type keywords, get a list of links — is becoming: hand over a task, get a finished result. The move defends and attacks at once for Google, guarding query volume against newer entry points while converting decades of accumulated index and understanding into moves an agent can execute.
When Agents Run the Search Box
For the content ecosystem, no ranking update ever cut this deep. With an agent reading, researching, and acting in the user’s place, websites stop collecting clicks and start collecting a different kind of settlement — a citation, a referral, a license, nobody yet knows which. One thing is sure: the traffic economy’s distribution logic is headed for a rewrite. Users get convenience they can feel; publishers who live on search traffic face risk just as real.
Spark and 3.5 Flash, Each to Its Own
The two same-stage launches split the work cleanly. Gemini Spark is a new assistant that wraps agentic capability in an interface everyday users can just talk to; Gemini 3.5 Flash covers the model layer, in the Flash line’s usual role of driving speed and cost down until deployment at scale makes sense. Front-of-house experience, back-of-house economics — each has its place.
The Developer and Publisher Watchlist
- The citation and referral habits of Search agents will shape content sites’ traffic structure
- More Flash-tier models are joining the lineup, and quick, low-cost inference is well on its way to default status
- Whether Spark ever opens up to third-party integration will tell us if it is a closed assistant or an agent platform at population scale
Pichai has named an era and pushed the bet onto the table. Whether Search’s biggest revamp in 25 years lands without a stumble, the coming months of real use will show.