Google Rebuilds Search Around Gemini 3 in Its Biggest Overhaul in 25 Years

The dek: Putting a reasoning model at the center of the world's front door to the web is not just a feature update. It quietly rewrites the bargain that has funded online publishing since the late 1990s, and SEO with it.

For most of its existence, Google Search worked like a switchboard. You asked a question, it handed you ten blue links, and the click that followed sent both attention and ad revenue to the site on the other end. That loop paid for newsrooms, niche blogs, recipe sites, and most of the open web. Google's move to put Gemini 3 at the heart of Search puts that loop under pressure in a way no past redesign has.

The shift is structural. With Gemini 3 powering AI Mode in Search, the page increasingly answers the question itself. Google says AI Mode now supports more complex reasoning, generated visual layouts, and interactive tools and simulations built on the fly. In other words, the result is no longer a list of destinations. It is a synthesized answer assembled from many destinations, with the originals reduced to citations a user may or may not follow.

For publishers, that is the headline. When the answer lives on the search page, the incentive to click through drops. The web's content still feeds the system, but the traffic and the ad impressions that historically flowed back to creators no longer arrive at the same rate. Sites that depend on informational search queries are the most exposed, because those are exactly the queries an AI answer handles best.

The model rollout has moved fast. Gemini 3 Pro arrived in November 2025, and Google made a point that it was shipping in Search on launch day for the first time, rather than weeks later. In December 2025, Gemini 3 Flash followed, a cheaper and faster member of the family that Google priced at $0.50 per million input tokens and $3 per million output tokens. Flash became the default model in the Gemini app and is rolling out in AI Mode in Search globally. Cheaper inference matters here: it makes AI answers economical to serve at Google's scale, which is what turns an experiment into a default.

For the SEO industry, the ground is moving. Ranking for a keyword means less when the click that ranking was supposed to win never happens. Practitioners are already shifting language toward optimizing to be cited inside AI answers rather than to top a results list. Whether being a footnote in a generated response carries the same value as owning the destination page is the open question, and the honest answer today is unknown.

None of this means the open web disappears overnight. AI answers still need source material, and brand, authority, and original reporting may matter more, not less, when a model is choosing whom to trust and cite. But the terms have changed. Google built the most successful advertising business in history on sending people away to other sites. Gemini 3 is the clearest sign yet that it would rather keep them.

Fontes

• https://blog.google/products/gemini/gemini-3/
• https://blog.google/products/gemini/gemini-3-flash/