Google Gemini 3.5 Flash Now Powers AI Mode Globally With Major Agentic Search Features

Google I/O 2026 was packed, and if you blinked you probably missed half of it. But the headline that matters most for anyone who cares about search is this: Google Gemini 3.5 Flash is now the default model powering AI Mode globally. That’s not a small update. It’s a signal that Google is done experimenting and is now moving fast on making AI search the standard — everywhere, for everyone.

Alongside the model upgrade, Google announced a stack of new agentic features either live now or landing this summer. Information agents, agentic coding, Universal Cart, Universal Commerce Protocol expansion, and a new Agent Payments Protocol — it’s a lot to take in. Let’s go through it properly.

Google Gemini 3.5 Flash Is Now the Default AI Mode Model Worldwide

Google describes Google Gemini 3.5 Flash as delivering “sustained frontier performance for agents and coding.” Google Gemini 3.5 Flash is the newest Flash model in the Gemini family, and Google has made it the default engine behind AI Mode for every user globally — no subscription, no waitlist. The upgrade matters because Flash models are built for speed and reliability at scale, which is exactly what you need when you’re running agents that are supposed to be working in the background around the clock.

This is a meaningful step up from where AI Mode was just a few months ago. The broader global rollout of AI Mode, which we covered when Google AI Mode went global earlier this year, was running on earlier infrastructure. Google Gemini 3.5 Flash brings that experience up to a genuinely new level of performance.

Information Agents: Search That Works While You Sleep

The most genuinely interesting feature announced at I/O wasn’t a flashy demo — it was information agents. These are background agents that run 24/7, continuously scanning the web on your behalf for changes relevant to your specific queries or tasks. Blogs, news sites, social posts, real-time finance data, sports scores — the agent watches all of it and notifies you when something relevant happens.

Google gave two examples that land well: if you’re apartment hunting, you dump all your requirements into Search and the agent continuously scans listings and pings you when something matches. If you want to know the instant a favourite athlete announces a sneaker collab, the agent catches it the moment it drops. It’s the kind of thing people have been cobbling together with RSS feeds and alerts for years — but built natively into Search and powered by Google Gemini 3.5 Flash. Information agents will roll out first for Google AI Pro and Ultra subscribers this summer.

Agentic Booking Now Covers Home Repair, Beauty and Pet Care

Google launched agentic booking a while back for restaurants and a handful of service categories. I/O 2026 expanded the scope significantly. Now, if you need to book home repair, beauty appointments, or pet care services, Google can place calls to businesses on your behalf to check availability and set things up. No more bouncing between tabs and waiting on hold. Agentic booking is rolling out to all U.S. users this summer.

Agentic Coding in Search: Build Mini Apps Directly in Google

This one is harder to wrap your head around until you see it. Google Gemini 3.5 Flash is powering a new agentic coding capability inside Search, built on Google Antigravity. There are two versions of it.

The first is generative UI, where Search automatically builds custom layouts for you on the fly — interactive visuals, tables, graphs, simulations — assembled in real time based on what you’re searching for. No setup required. The second is custom coding experiences, where Search actually builds you a mini app for a specific ongoing task. Google’s example was a custom fitness tracker that taps into real-time sources like reviews, live maps, local weather, and health data. You ask for it, Search codes it, and you can keep coming back to it.

Generative UI will be available to everyone in Search this summer, free of charge. The custom coding mini apps are coming in the following months, starting with Google AI Pro and Ultra subscribers in the U.S.

Personal Intelligence Now Available in 200 Countries

Google launched Personal Intelligence in Gemini back in January, expanded it in March, and is now pushing it out to nearly 200 countries and territories across 98 languages — with no subscription required. That’s a significant jump in accessibility. Personal Intelligence uses your history, preferences, and context to make AI Mode responses more relevant to you specifically, rather than giving generic answers to everyone.

Universal Cart: One Checkout Across Multiple Retailers

Universal Cart is exactly what it sounds like, and it’s genuinely useful. You can now add products from multiple retailers — Nike, Sephora, Target, Walmart, Wayfair, Shopify merchants, and more — into a single cart and either check out with one payment through Google Pay or transfer to the individual retailer if you prefer.

But it’s smarter than just a unified basket. The cart actively works in the background, finding deals and price drops, flagging product incompatibilities (say, if two PC parts you’ve added won’t work together), applying loyalty perks and promotions from your Google Wallet, and suggesting alternatives. It runs on Gemini, which means it gets smarter as the model improves. Universal Cart is rolling out in the U.S. this summer across Search and the Gemini app, with YouTube and Gmail to follow.

Google is also expanding its Universal Commerce Protocol — the backbone that makes agentic shopping possible — to Canada, Australia, and eventually the U.K., with YouTube integration and new verticals including hotel booking and local food delivery coming soon. We covered the earlier UCP rollout in detail when Google first expanded it, including what it means for retailers — read more in our piece on Google’s Universal Commerce Protocol expansion and what it means for merchants.

Agent Payments Protocol: Guardrails for Agentic Transactions

The final piece of the puzzle is AP2 — Google’s new Agent Payments Protocol. As agents start making real purchases on your behalf, the question of trust and security becomes critical. AP2 creates a verifiable, tamper-proof link between you, the merchant, and the payment processor, using privacy-preserving technology to keep your data safe. You can set strict guardrails on what agents are allowed to spend and where. Every transaction leaves a permanent digital paper trail.

Google will start rolling AP2 out to its own products in the coming months, beginning with Gemini Spark. It’s a foundational piece of infrastructure for the agentic commerce ecosystem Google is clearly building toward.

Taken together, what Google announced at I/O 2026 isn’t a collection of isolated features — it’s a coherent architecture. Google Gemini 3.5 Flash is the engine. Information agents, agentic booking, agentic coding, Universal Cart, and AP2 are the layers built on top of it. For anyone working in SEO, content, or ecommerce, the direction is now unmistakably clear. You can read Google’s full announcement directly in Google’s official I/O 2026 Search blog post. Search is becoming an agent layer, and Google Gemini 3.5 Flash is what’s running it.