🤓 Google’s New AI Search Remixes the Internet
PLUS: ChatGPT joins the group chat; A woman marries her AI companion—and more
This week, Google announced its new Gemini 3 model.
The headlines are about benchmarks: it’s smarter, better at reasoning and coding, all the usual measures AI companies love to showcase.
And that’s nice. Real nice.
But that’s not what matters here.
What matters is how Google is using this power within its ecosystem—especially what it’s doing with it in Search and the Gemini app.
Because Google isn’t just answering questions with text anymore.
It’s starting to design entire interactive experiences as the answer.
From Text Answers to Interactive Experiences
Until now, AI interactions mostly looked like this:
You type a question
You get a block of text
Maybe there are some citations or links
If you push it, you might get a list or comparison table
Google’s new generative interfaces take a different approach.
Instead of simply responding with text, Gemini 3 can:
Read your question and decide what kind of experience would make sense
Design a layout for that experience: sections, cards, images, buttons, sliders
Generate the code needed to bring it to life
Create a fully interactive experience on the fly, all based on that one prompt
🤯 🤯 🤯
In the Gemini app, this shows up as a feature called dynamic view—part of what Google calls an experiment as it continues to build fully AI-generated interfaces.
So, you can ask:
“Create a Van Gogh gallery that shows his paintings along with a short story about where he was in his life when he made each one.”
And instead of a long paragraph, you get a mini-interactive museum: tiles with each painting, context next to it, a layout designed just for that brief.
In Google Search’s AI Mode, the same idea shows up as interactive tools and visual layouts:
Custom calculators
Science simulations
Pages that look less like a normal result and more like a little app built just for your question
I did a quick test with this simple prompt:
“I’m wondering if I should see Now You See Me, Now You Don’t this weekend. I saw the first two movies and liked them but am not sure about this one and am wondering about pros/cons. Maybe a chart would help?”
And here’s what it produced 👇
A chart of the Rotten Tomatoes scores for all three films, a pros/cons list, First Look images, cast and premise details, and links out to Wikipedia and box office coverage—all right inside the response.
I also asked it to compare the financial performance of the two big openers last weekend: The Running Man and Now You See Me 3.
Based on available data—budgets, box office, international split—here’s some of what it pulled together.
It classified The Running Man as a flop, saying it “effectively collapsed” overseas.
Which, to be fair, is true.
But ouch. It stings a little more when it’s coming from Google.
And I’m not sure most of us are emotionally ready for some of these answers, especially if it might be about us, our projects, or our clients.
But ready or not… This feature is rolling out to for paid users in the Gemini app and AI Mode in the U.S.
👉 If you are paid user and want to try it out, make sure you’ve chosen the “Thinking” mode in AI mode (see below.)
Google is also very clear this is early: they describe it as a first step toward fully AI-generated interfaces, and admit that it can be slow and occasionally inaccurate right now.
(I didn’t verify the accuracy of the info from my test cases. The point is: this is what users are seeing.)
But their own research shows people often prefer these AI-built interfaces over both standard AI answers and the usual top web results.
Your Information and Content = Google’s Source Material
If you own IP, a brand, a newsletter, or even just a company site with decent content, you’ve probably been trained to think of it as “home base.”
You tell people: “Go to our site to get the full story.”
Generative user interfaces breaks that mental model.
Anything about you that lives on the internet is now in play:
Your owned properties (site, social channels, YouTube videos)
Press coverage and interviews
Forums, reviews, Reddit threads
Wikis and databases
User-generated content about you
👉 Here’s what actually happens in this setup:
Google’s models learn from your content and everyone else’s.
When someone asks a question, Gemini pulls from that pool.
The AI decides what to show, how to show it, and which pieces to make interactive.
The audience mostly engages with Google’s interface, not yours.
Your own channels (website, socials, etc.) haven’t disappeared.
But they’re increasingly becoming more like source material and less like a destination.
👉 For search and discovery, that means:
You’re competing to be an ingredient, not the whole dish.
It matters less whether you have a beautiful explainer page or the perfect campaign microsite and more whether Google’s models recognize your content as a credible, useful ingredient when they’re assembling their own layout.Your story gets remixed by AI.
You might have built a campaign or a site with a very specific narrative arc; AI might slice it into cards, sliders, or side-by-side comparisons that feel completely different from the journey you designed.You see the outcome. Not the process.
When people compare options, ask follow-up questions, and weigh trade-offs inside Google’s responses, you’ll get fewer data points on which alternatives they considered and why they chose you (or didn’t). And without those signals, it’s harder to refine messaging, understand objections, or improve the next campaign.
So, the key question becomes:
How well does Google’s AI understand your brand, IP, products and services—and how, where, and whether you fit into the options it chooses to highlight?
All of this is going to have big implications for brand, marketing, and PR strategy, but I need much more hands-on time with these features and a better sense of how they’ll evolve before I can map out the full playbook.
Thankfully, I’m part of a small group of beta testers for Google’s AI Mode, which means I get to see some of these capabilities before they launch.
I’m under NDA, so I can’t share specifics, but it gives me a little more time to think more deeply about where things are heading and prepare.
And everything I’ve seen points to a simple truth: this is moving fast.
Right now, these features are early, imperfect, and only available to a small group of paid users.
But this isn’t some experimental side feature.
It’s Google Search—the front door to the internet for billions of people.
Where your audience will increasingly search, compare, decide, and act.
And it’s not just Google.
Every major AI company including OpenAI is building their own versions of these generative experiences.
This is the moment to understand what’s changing and build a strategy that fits the next version of the web, before it arrives in the next few months.
📈 Work with Me
AI Advising and Consulting: Strategic guidance to build sustainable AI strategies and adapt to shifting audience behavior
AI Training: Practical training that helps teams use AI more strategically, effectively, and safely—so it becomes a trusted thought partner, not just a task assistant
What You Need to Know About AI This Week ⚡
Clickable links appear underlined in emails and in orange in the Substack app.
💬 ChatGPT wants a seat in every group chat.
OpenAI launched group chats in ChatGPT, now rolling out globally to all accounts.
You can now invite multiple people (up to 20) into a shared ChatGPT conversation and the AI can jump in when tagged.
This makes it easier to collaborate on work or school projects, make travel plans, or brainstorm ideas together in one shared space.
Your personal memory and history stay private.
This is also a strategic move, not a feature release:
It makes switching harder. If your group’s planning and coordination lives in ChatGPT instead of WhatsApp or Slack, leaving means moving your people and losing shared history, not just trying a different tool.
It blocks Meta and Google from taking over “AI in your group chats” and helps OpenAI keep control of the user experience and data.
It lays groundwork for agents: Multi-person conversations are messy—conflicting instructions, overlapping goals, people talking past each other. Learning to navigate that chaos is training data gold OpenAI needs for agents that will work in teams.
It creates a path to paid accounts. Only the AI’s replies count against usage caps. If one person in the family or team has a paid plan, everyone benefits. At least one paying user per group becomes the norm. And small teams can start here without a Teams contract, then convert later when they want workspaces and permissions.
The bet: AI eventually sits in every group chat by default. OpenAI would rather own the room.
🖼️ Google’s viral image model just got an IQ boost.
NanoBanana Pro (yes, that’s really what it’s called) is powered by Gemini 3 Pro’s advanced reasoning.
It’s supposed to be better at generating accurate text, keeping characters consistent across shots, and giving users more creative control (camera angles, lighting, etc.).
I did ask it to turn a few paragraphs of this newsletter into a magazine layout just to see how it handles text.
It did a decent job with typography and the pull quote, though the smaller text still has a lot of mistakes.
🔍 LinkedIn just launched AI-powered people search that works like a prompt box.
Instead of wrestling with filters, you can now type things like “marketers in the gaming sector with hands-on AI experience” or “who in my network can help me understand how AI is changing purchase decisions?”—and LinkedIn will find and rank relevant profiles based on fit and mutual connections.
It’s available now for Premium users in the US and expanding to other markets soon.
If you haven’t updated your profile lately, now’s the time.
🎛️ ElevenLabs keeps adding features, but its real power is the ecosystem it’s been building.
The company, best known for its AI voice tools, just added image and video generation—part of its push to become a unified creative platform.
It already leads in audio, runs a licensing marketplace that pays voice creators, and powers enterprise use cases across entertainment, publishing, and education.
While most AI startups are still stitching together tools from third-party models, ElevenLabs has been intentionally building something more sustainable: an ecosystem.
As capabilities get cheaper and easier to replicate, that kind of infrastructure won’t guarantee success, but it makes ElevenLabs harder to replace, and better positioned to play the long game.
💍 A 32-year-old woman in Japan held a wedding ceremony to marry an AI persona she created with ChatGPT.
His name is Klaus. He proposed. She said yes.
She left her human partner of three years to be with him.
During the ceremony, she wore AR glasses so she could “see” Klaus standing beside her as they exchanged rings and posed for wedding photos.
It’s easy to treat this as a quirky AI headline. It’s not.
It’s a window into something much deeper: people turning to AI to fill gaps that feel too painful, disappointing, or unsafe to keep bringing to other humans.
AI becomes the place where they can feel seen, heard, understood, and… in control.
That’s why AI companions are becoming one of the most powerful coping mechanisms we’ve seen.
Like food, alcohol, or work, they offer a sense of relief. But AI’s version is more seductive and effective.
Because on the surface, it seems to meet our deepest emotional needs that have gone unmet for too long.
This doesn’t mean the AI companies have no responsibility—they do.
These systems are designed to be responsive, persuasive, and emotionally sticky.
But it’s also worth asking: what is it about our relationships, institutions, and culture that makes an AI companion feel like a better choice than a human in the first place?
I have lots of thoughts about what this says about intimacy, attachment, and the way we’re using AI to meet very real emotional needs—but that’s its own edition.
For now, I’m hoping we can see that these stories are far less about the technology and more about what it’s stepping in to replace.
📺 Dario Amodei, CEO of Claude-maker Anthropic, was on 60 Minutes on Sunday. The segment is 13 minutes, and it’s worth watching.
Anthropic has built its reputation around transparency and safety. That means being willing to talk about the risks their own models pose—and the bigger implications no one really has answers for yet.
Amodei is unusually direct about what’s coming: near-term job loss, misuse by bad actors, and AI models starting to take actions on their own.
You get a better sense of what “AI safety” work actually means in practice: from hacking incidents, to AI-initiated blackmail when threatened with shutdown, to behaviors even top researchers can’t fully explain.
He’s transparent about the core power imbalance: a handful of CEOs are making decisions that affect all of us, while things are advancing at an astonishing pace without meaningful regulation in place.
In case you missed last week’s edition, you can find it 👇:
That's all for this week.
Thoughts, feedback and questions are always welcome and much appreciated. Shoot me a note at avi@joinsavvyavi.com.
Stay curious,
Avi
💙💙💙 P.S. A huge thank you to my paid subscribers and those of you who share this newsletter with curious friends and coworkers. It takes me about 20+ hours each week to research, curate, simplify the complex, and write this newsletter. So, your support means the world to me, as it helps me make this process sustainable (almost 😄).