At I/O 2026, Google rebuilt Search around AI: Gemini 3.5 Flash became the default in AI Mode, the Search box got its biggest upgrade in 25 years, and Preferred Sources, the favorite sites users select, now appear inside AI Overviews. The Google I/O 2026 AI Search announcements point to one shift for SEO and GEO teams: visibility now depends on being a cited, chosen source, not just a ranked one.

• Gemini 3.5 Flash is now the default model in AI Mode, which has passed 1 billion monthly users.
• Preferred Sources now appear in AI Overviews and AI Mode, where Google says they earn roughly 2x more clicks.
• A new Highly Cited badge flags original reporting that other articles reference.

Search agents and generative UI are coming to Search this summer, free for everyone.

Gemini 3.5 Flash and the New Search Box Make Queries Longer and Smarter

The headline change is a faster model behind a smarter query box, and it rewires how people search. Google made Gemini 3.5 Flash the default in AI Mode worldwide and called the redesigned Search box its biggest upgrade in over 25 years. The box now expands as you type, suggests intent beyond autocomplete, and accepts text, images, files, videos, and Chrome tabs as inputs. AI Mode itself has crossed 1 billion monthly users, with queries more than doubling every quarter since launch.

For SEO, the practical effect is that head terms keep losing ground to long, conversational, multi-part questions. Content tuned to a single keyword will lose to content that resolves the full intent a user describes in a sentence. The teams that adjust fastest will rewrite briefs around questions and entities, not around isolated query strings, because that is what the new box and the underlying model are built to match.

Preferred Sources Just Became the Biggest GEO Opportunity of 2026

The most actionable announcement hands part of your visibility to your own audience. A week after I/O, Google brought Preferred Sources into AI Overviews and AI Mode: users mark favorite sites in their settings, and those sites get a visible label whenever they appear in an AI answer. Google reports a 2x higher click-through rate for these sources, and more than 345,000 unique sources have already been selected.

This is the clearest generative engine optimization lever Google has shipped, because the choice belongs to the reader rather than the auction. Any site publishing fresh content is eligible, and Google offers documentation on prompting readers to opt in. The move for enterprise brands is to treat Preferred Source selection like a newsletter signup: a recurring on-site and in-email ask, tracked over time. Build the habit now and you compound an AI visibility advantage competitors cannot simply outspend.

The Highly Cited Badge Rewards Original Reporting Over Aggregation

Originality just became a visible ranking signal. Google is adding a Highly Cited badge to article links that many other stories reference, surfacing the primary reporting that downstream coverage is built on, and it will also flag when an article cites a Highly Cited source. The message to publishers is direct: be the source others quote, not the one restating them.

For AI Overviews strategy, this rewards proprietary data, firsthand research, and named expert points of view over commodity explainers. A page can sit below the top organic positions and still surface in an AI answer if it holds the original claim. Enterprise content teams should reallocate budget from broad coverage toward fewer, stronger assets designed to be cited, since those are the ones AI surfaces now amplify.

Search Agents and Generative UI Change How Brands Get Discovered

Discovery is moving from a list of links toward agents that act on your behalf, and that reshapes the funnel. Google announced information agents that work in the background 24/7 to track a topic and surface changes, rolling out this summer to AI Pro and Ultra subscribers in the US first. For tasks like booking, Search now pulls together live pricing and availability and hands over direct links to finish the reservation with the provider. And through Antigravity and Gemini 3.5 Flash, generative UI assembles custom layouts, interactive tables, graphs, and simulations on the fly, free for everyone in Search this summer.

For SEO and GEO, this raises the bar on structure and machine readability. If an agent assembles the answer, your content and your commerce data need to be parseable, current, and unambiguous, or you are left out of the action. Clean structured data, accurate availability and pricing signals, and answer-shaped content stop being technical hygiene and become the entry ticket to agent-driven discovery and conversion.

What the Google I/O 2026 AI Search Changes Mean for SEO and GEO Teams

Taken together, the announcements reward scale advantages large organizations already hold, if they actually use them, and this is where an enterprise plays a different game than a small site. The first advantage is proprietary data. Where a small publisher struggles to be cited, an enterprise sits on first-party data, transaction patterns, and category benchmarks it can turn into original research that others have to reference, which is exactly what the Highly Cited badge now surfaces. Productizing that data into recurring studies beats publishing more commodity explainers that the AI summarizes away.

The second is distribution. A large brand can campaign for Preferred Source selection across newsletters, apps, and millions of monthly visitors in a way a smaller competitor cannot, and labeled sources earn roughly double the clicks. Multi-brand portfolios face a sharper version of the question: decide which entity readers should select, rather than splitting that loyalty across sub-brands that compete with each other for the same slot.

The third is governance, and it is usually where enterprises lose. Agentic discovery and generative answers depend on clean, consistent structured data across thousands of pages, several markets and languages, and more than one CMS or PIM. That is no longer a per-page tweak but a data-integrity program owned jointly by SEO, engineering, and merchandising. Measurement has to cross the same silos: citations and AI-surface mentions sit between SEO, PR, and brand, so the teams that win build one shared view instead of three competing dashboards. None of this retires technical SEO or topical depth. It reprioritizes them around a surface where Google and its users now pick sources directly.

Where SEO and GEO Teams Should Focus This Quarter

The I/O 2026 changes converge on a single takeaway: AI Search rewards original, cited, user-selected sources, and ranking position is now a means to that end rather than the goal. The fastest way to act is to find out how often AI surfaces already cite you and where the gaps sit, then fix the content and data signals that decide selection. Primelis runs an enterprise SEO audit built to map exactly that and to turn the Google I/O 2026 AI Search shift into a plan you can ship before the next core update settles.