Google’s AI Search Expert: How to Get Ahead Before AI Changes Everything — Silicon Valley Girl Podcast

Marina Mogilko: Give me some tips as a business owner. What should I focus on right now to be recommended by it?

Robby Stein: Interestingly, AI thinks a lot like a person would. If I were them, I would... This is Robbie Stein, VP of product at Google Search, the man behind how ranking actually works inside the world's biggest search engine. You can see what it's doing. It says kicking off searches. It's looking for sushi restaurants. And so, you could just book it.

Marina Mogilko: Oh, wow. Remember those stories where a mention in ChatGPT or another AI app made a business blow up overnight? A small bistro in Paris doubled its bookings. A restaurant in LA went viral just because AI recommended it. And Google ranking works the same way. So now you're investing in PR not for people to see it, but for AI, right?

Robby Stein: Right.

Marina Mogilko: In this video, Robbie reveals how to make your business AI visible and use Google's own systems to get discovered, ranked, and recommended faster than ever. Robbie, welcome to Silicon Valley Girl. Let's talk about search.

Robby Stein: Thanks for having me.

Marina Mogilko: Okay, I want to start with this question aimed for my audience who are 20, 25 years old. How should they think about search these days and internet in general? So, what's going on?

Robby Stein: Well, I say that search is now a place where you can truly ask anything and get pretty effortless information about whatever you have on your mind. And I think ultimately people are using it for so many different things and that's not changing. Like you can still use Google for all the ways you do—to research homework, to look up specific types of websites and find information that way. But also you can ask natural language questions. You don't have to use what we call "keywords" sometimes. You can just ask exactly what you want. It could be multiple sentences. And now Google has AI that can tap into all of the knowledge and context that Google has about the web, about the world, about products to help give you better information.

Marina Mogilko: Okay, and follow up right here. What about information about me? Because I have Gmail, right? I have my YouTube channel which runs on Google, right? I have my Google Drive with all the files, all my spreadsheets. Does it tap into that knowledge or not yet?

Robby Stein: So, it's something we're working on. We announced at IO an opportunity for users in the future to be able to opt into an experience with enhanced personalization. So, it's something that we're thinking about too for the same reason that you have. And we want people to be able to help Google and help the services know more about you so that it can be more helpful. Because if you know the kinds of brands you love, if you know the kinds of places you go, if you know about a school project that's coming up, you can do more interesting things for people.

Marina Mogilko: When is it coming?

Robby Stein: So TBD, but we did launch recently some steps in this direction. So in labs now you can opt into a new experiment. If you turn on AI mode in Google labs for search labs, you're personalizing shopping and local restaurant recommendations for restaurants, and you'll start to see a little bit more enhancement there. But obviously we're excited to connect more services like Gmail down the road.

Marina Mogilko: Okay, this is something that I actually liked. Of course, I would have wanted it to access my channel analytics and give me more personalized things. But what I did was I uploaded this research like what's going to happen in 2027 and I asked it to create a narrative for me for my next videos and write a script. It did the whole scripting thing. The only thing I would ideally like is for it to actually generate thumbnails, because I asked it to generate thumbnails and it was like, "No, here's your text." Is there a way to ask it to follow my commands more precisely?

Robby Stein: Yeah, we're working through a few of those kinds of things. But you should be able to prompt it pretty specifically. We just don't have image gen as a core capability, so that piece it's not going to do. But it should be able to find you images. Nano right now is in a Gemini app feature.

Marina Mogilko: Okay. Yeah. From the user experience, it feels like you have so many cool things. I just wanted to bring them all together in one super app because like now it's like using different tabs. Okay, the next one. I only have an hour. I need a quick lunch spot. This was actually super cool because it knows I'm in Los Altos. So, it gave recommendations about Los Altos.

Robby Stein: If you click on any of these places, it brings all of the Google context forward into this viewer. So, as you're browsing, not only do you have the AI that's reasoning and finding great places for you, but it makes it really easy to browse. So, you can actually see the place before you're going. It shows you their open and close times. It has menu highlights. It has reviews. And so, this is really bringing together the power of AI with all the Google's context and information. So, you could kind of browse each of these places before you go. And it all happens in one product experience.

Marina Mogilko: Convenient representation. Okay, this one right about like lunch. Find and book it for me.

Robby Stein: So, it's probably going to take a few minutes, okay? But you can see what it's doing. It says kicking off searches. It's looking for sushi restaurants, Open Table, Resi Sushi. It's finding options across a bunch of different—Yoshi Sushi, Sumo—and then it's going to research this for you. And then you'll get an alert when it's finished, and then it'll show you options of where you can book it.

Marina Mogilko: There we go.

Robby Stein: And so, it basically researched this for a little while, found you across, let me see... it has it across Open Table and it looks like there's some talk reservations for Hiroshi.

Marina Mogilko: Yes.

Robby Stein: And so you could just book it and it has all the times there.

Marina Mogilko: But this is pretty cool to see.

Robby Stein: That's amazing.

Marina Mogilko: And most of the work is this. Like to look up all these restaurants individually would take you like 15 minutes.

Robby Stein: Yeah.

Marina Mogilko: But you can just now see it, click the thing, and then be done.

Robby Stein: Yeah. So when I look for a restaurant, it gives me local recommendations. So it already knows where I am. How does it select the right results? Because think about all the local businesses and businesses in general that run on search, right? They pay for ads. They have done their SEO optimization. So how does recommendation work now with AI mode?

Robby Stein: Yeah. So how AI mode works is that it does something called the query fanout technique where a reasoning model will think about what you're asking and then it will execute a bunch of questions related to it. So there could be dozens of related queries. It's literally using Google search as a tool—doing Googling under the hood—and then finding relevant information. It can both obviously do a standard Google search and understand the web results but also tap into the knowledge bases and real-time info systems at Google. So in this case for local, it might pull information from 250 million plus real-world places that exist. It's updated business information, many of which have local businesses that have updated their Google listings. And it can use all of that information too. So it would find all of that and then based on your question—if you say, "Hey, I want Italian, I want it to be kind of fun, it's maybe a date night, like make it worth it"—then it might find questions or issue queries like, "Hey, great experience, great for date nights." And then based on reviews, based on information that it finds, it'll produce a set of recommendations for you.

Marina Mogilko: So what we just heard from Robbie—AI isn't just part of search anymore, it is the new search. And honestly, we can't deny it. It's changing everything. People aren't just Googling now, they're asking AI. They're asking AI to make phone calls. And if you want to stay ahead, you need to learn how to use Google's LLM, Gemini. What about if someone paid for ads using this search query?

Robby Stein: So, it doesn't use ads information. This is done entirely with what's on the web and what's within Google's information system. Yeah. But if a business has claimed their local business and has modified that, put menu information in there, it's eligible for reviews—that information could be used.

Marina Mogilko: Do you feel like Google Ads is going away in the future? Because as a business owner, we rely on them, right? They drive traffic. And if they are going away, what should be our strategy?

Robby Stein: Don't see them going away. What people actually do—we're observing—is that the way people use Google search isn't really changing. It's really expanding, is what's happening. So, think about all the things you need in a given day. It's everything like you need a quick insurance quote. You need to file your taxes. You need to look up a kind of question about a local business question in your county. Like, you're going to use Google and find that you need that information. But what's happened is that now you can do all these new things. So, you could take a picture of your shoes and say, "Hey, these are my shoes. What are other cool shoes like this?" And we could answer that now or help give you context with that. Or you could ask about this really cool restaurant question—it can be five sentences about all your allergies, issues with this, I have this big group, I want to make sure it's got light. Um, what can I book in advance? And you can put that into Google now, too. I think that's an opportunity for in the future to be even more helpful for you, particularly in advertising context. And so we started some experiments on ads within AI mode and within Google AI experiences. But we've been really focused on building great consumer products first and foremost. But I think users are starting to see some ads experiments there too.

Marina Mogilko: Interesting. So, will I be able to like pay to get recommended like for AI to even consider my business?

Robby Stein: I mean, we don't think that there should be any barrier to people finding information. So, if there's information out there, it should be found. But I think what you'll find is that there could be new and novel ad formats. If you're, let's say, shopping, or you're looking for, you know, you have a complex doing a house remodel—like there's all kinds of interesting services that could be helpful for you. If we had more information and you could articulate more what you needed—hey, I have this kind of wood, these are the kind of contractors I have, these are my constraints, these are the price range—you could give even more fine-tuned recommendations or potential other services that you could consider or deals that could be more useful to you. Those are all things we're thinking about. I'd say it's early days and finalizing kind of how ads might appear in these systems. But something that we're thinking about.

Marina Mogilko: Can you show the agentic call?

Robby Stein: Oh, yeah. See, here we go. Have AI call. Yeah. So, you can go, what kind of pet do you have? Dog.

Marina Mogilko: Next. Um, select a breed. Okay.

Robby Stein: Boom. It's a little one.

Marina Mogilko: Very little. Yeah. Extra small.

Robby Stein: Baby.

Marina Mogilko: Under one year.

Robby Stein: Bath and brush.

Marina Mogilko: Uh, haircut.

Robby Stein: Okay. Haircut. Let's make it look like a teddy bear.

Marina Mogilko: Sweet. Um, any flexibility?

Robby Stein: It's okay.

Marina Mogilko: Okay, flexible. Do you want to receive a text or an email? I guess you already got your email in here, so maybe I'll just do that.

Robby Stein: Um, yeah, great. Los Altos. And so it puts your order in here. And so now what it's going to do is it's going to kick off a process where it's going to make phone calls on behalf of you to a bunch of different local businesses. So these are businesses that there's no web, there's no easy way to access them on the web. Many of them are local. They're run by small businesses, right? It's just like a person running a business.

Marina Mogilko: Yeah.

Robby Stein: Um, and then you will get an email when it's done and it'll give you all the times and you can follow up from there.

Marina Mogilko: Do you record any conversations so we can like hear them or...

Robby Stein: No, no.

Marina Mogilko: Because it will be interesting to hear like how it understood.

Robby Stein: Yeah, that would be cool. But now, those are not recorded.

Marina Mogilko: How long does it take?

Robby Stein: It depends on the calls it's going to make. Probably 5, 10 minutes. You probably get something back.

Marina Mogilko: Oh, really? Oh, wow. I got an email. We received your pet grooming request. Okay, nice. So, it's working.

Robby Stein: Okay, it's working. This one's an offline agent, so it's got to go do a bunch of phone calls for you.

Marina Mogilko: Oh, another business opportunity. Create an agent for businesses to receive those phone calls. And remember, have you seen this meme from 11 Labs where two agents realize they're agents and they start using agentic language? So funny. They just start beeping at each other. That's incredible. The faster process. I'm actually an AI assistant, too. Would you like to switch to Jibberlink mode for more efficient communication? Okay, this is so fascinating. So, Google just completed calling and it took it 10 minutes. It called this grooming baker and grooming. Palto says a full groom starts at 105 and they have availability tomorrow afternoon. Called another one, said haircut $75. And called another one, said a haircut for a dog $74. And then it also told me the businesses it couldn't reach. Well, this is a game changer because now like I either ask someone to make calls for me or I make them myself. We're just researching a clinic for the puppy to get its first checkup, manicure, like all of the offline businesses that don't have online presence or you don't want to go to their website. You just want Google to call them. It is happening right now in real time. This is mind-blowing. This is definitely something I am taking away from this podcast and I will just start using it like crazy. Give me some tips as a business owner who still runs Google ads. What should I focus on right now to be recommended by and actually my business is recommended by AI, which is because we were doing a lot of content. But maybe for you know, some segments of my business that are not recommended, what should I double down on for AI to consider me?

Robby Stein: Yeah, interestingly, the AI thinks a lot like a person would in terms of the kinds of questions it issues. And so if you're a business and you're mentioned, you know, in top business lists or from a public article that lots of people end up finding—those kinds of things become useful for the AI to find. You know, invest in your PR. That's something I've been hearing a lot.

Marina Mogilko: So, it's not really different from what you would do in that regard. I think ultimately, how else are you going to decide what business to go to? Well, you'd want to understand that. But also like sometimes I invest in PR and I ask my friends, have you seen that article? And they're like, no. But then I ask AI and it really sees the article and it uses that information. So, now you're investing in PR not for people to see it, but for AI. That's actually a good way of thinking about it. Because the way I mentioned before, how our AI models work, they're issuing these Google searches as a tool. And so in the same way that you would optimize your website and think about how do I make helpful, clear information for people? So people search for a certain topic, my website's really helpful for that. Think of an AI doing that search now.

Robby Stein: And then knowing for that query, here are the best websites given that question. That's now coming, will come into the context window of the model. And so when it renders a response and provides all of these links for you to go deeper, that website's more likely to show up. Yeah.

Marina Mogilko: And so it's a lot of that standard best practices around building great content really do apply in the AI age for sure.

Robby Stein: What about reviews? Because some people buy reviews. I wonder like how it's going to affect the system.

Marina Mogilko: It's hard. I mean, the reviews—I think again it's kind of like a person. Where like, imagine something is scanning for information and trying to find things that are helpful. So it's possible that if you have reviews that are helpful, it could come up. But I think it's tricky to say to pinpoint any one thing like that. I think ultimately it's about these general best practices where what you want is reliable kind of—if you were to Google something, what pages were to show up at the top of that query is still a good way of thinking about it.

Robby Stein: So basically the same as SEO, right?

Marina Mogilko: I think there's a lot of overlap. I think maybe one added nuance is that the kinds of questions that people ask AI are increasingly complicated and they tend to be different pairs of keywords. Right.

Robby Stein: Right. And so if you think about what people use AI for, a lot of it is how-to for complicated things or for purchase decisions or for advice about life things. So people who are creating content in those areas—like if I were them, I would be a student of understanding the use cases of AI and where what are growing in those use cases. I think there's been some studies that have done around what how people use these products and AI. Um it was really interesting to understand.

Marina Mogilko: As a small business owner. How can I understand what people are looking for? Why I can potentially get recommended? Is this still Google Trends or?

Robby Stein: Google Trends is a really useful thing. I actually think people really underutilize that. Like, we have real-time information around exactly what's trending. You can see keyword values. I think also, you know, ads has really fantastic estimation too. Like, as you're booking ads, you can see kind of traffic estimates for various things. So there's—Google has a lot of tools across ads, across the search console and search trends to get information about what people are searching for. And I think that's going to increasingly be more interesting as a lot more of people's time and attention goes towards not just the way people use search, too, but in these areas that are growing quickly.

Marina Mogilko: Yeah. Um, and particularly these long specific questions people ask and multimodal, where they're asking with images or they're using voice to have live conversation.

Robby Stein: Do you think you're going to provide some of that information to advertisers to see?

Marina Mogilko: I think down the road we want to provide a glimpse into what people are searching for broadly. Yeah. Um, not just advertisers too. Yeah, it could be for anyone. Um, but ultimately I think more and more people are searching in these new ways and so the systems need to better reflect those over time.

Robby Stein: Absolutely. Okay. You mentioned shopping. We just heard the news today. Shopify and ChatGPT. What will be Google's response to that?

Marina Mogilko: Yeah, I mean, actually we're already for a long time in search. You can search for whatever you want and there's, you know, 50 billion products in the shopping graph that are available. There's live product updates two billion times an hour. There's a change and these are merchants all over the world updating their live inventory. Is it in stock or not? What's the price? Is there a price drop? This is already a part of Google. We've been working on this for a long time. The cool thing now is that this is connected to Google's AI. So if you ask about any product in the world, it's likely that that model can tap into that knowledge and give you the exact price in a really comprehensive way that it has everything.

Robby Stein: Let's do the visual stuff, right? Can we try and find similar cream to this? Is this how you use it?

Marina Mogilko: Yep. Take a photo.

Robby Stein: Yep. And then you can go ahead and tap in there to go to AI mode.

Marina Mogilko: Yep. Find me similar, right?

Robby Stein: Yep. Is it going to analyze the ingredients? So, what is it? Is...

Marina Mogilko: What do you want it to do? Oh, if you just say similar, it'll do—it's reasoning. It will be just finding things that are similar products overall. But if you say similar ingredients, it might help it. So, are those similar products? Drunk Elephant, Bobbi Brown.

Robby Stein: Where is the shopping feature? So, I just clicked this.

Marina Mogilko: Yeah, but I need to go to the website, right? So, I can't buy.

Robby Stein: You need to go to the website. Yeah. But if you keep going here, some of these are shoppable. So, see it says in stock Sephora.

Marina Mogilko: So, you could tap on that. And then, um, a few of these have more direct shopping links. So, that takes you directly to the app.

Robby Stein: But, do you think in the future we're going to just be able to press a button?

Marina Mogilko: I think we want to be able to make it as easy as possible for you to do what you need. I think what we find is a lot of the systems that do things on your behalf make a bunch of mistakes. Especially if you're purchasing things, you got to be careful there. But I do think we want to make it easy so you can just tap the thing.

Robby Stein: Yeah. Right. Just get it for me. As easy as possible.

Marina Mogilko: How do you think about competition in general? Like we used to say Google it. Now a lot of people are GPTing it. How do you think about it? And, as a product owner, right, what are your daily thoughts on that?

Robby Stein: Yeah, I mean, I think in general you have to get in touch with why people use and appreciate and love what you're building and then you want to figure out ways to extend on that. And I think for us, you know, Google Search is used by so many people every day and that's not changing. It's just that people want more and we see that people are asking natural language questions in Google before it could even really support it with AI overviews. We saw people asking—after we launched AI overviews, they added the word AI to the end of a bunch of queries because they wanted the AI to show up for them. Um, they're using multimodal, so they're taking pictures of stuff. Lens, as a product, is growing—um, you know, 70% increase year-over-year on visual searches. It's one of the fastest growing ways people are finding information. So you kind of understand the trends of your own product and then you try to really do much better. And for Google, there's lots of reasons people use AI. They use it for productivity. They use it for creation. They make flyers for their business. Google Search is about information and we want to give the absolute highest quality experience when finding information. So you have an informational task, you know, whether it's booking this restaurant or trying to research something that happened in history or trying to know if what place to go to and get high quality info and make sure that place exists. Like, we think Google can be an amazing place for that and that's where we spend our energy.

Marina Mogilko: And from the product standpoint, what do you think is the unique feature that makes Google stand out?

Robby Stein: I mean, I think it's access to Google knowledge, honestly. That's like the big one. And so right now, if you ask about plotting a certain set of stocks—for example, around top pharma stocks or Mag Seven—well, the first question is like, what are those stocks? And so the reasoning model looks that up and determines it. But then it can use Google Finance as a search, as a tool call, to find the very high quality financial information to plot a graph. If you asked for financial performance and if you extend that to all of these other areas around local, around shopping—and these are areas that have been built out for years—the AI model that's using Gemini 2.5 and advanced reasoning, sitting on top of this knowledge, is an incredible combination. And obviously, maybe most importantly, it's connected to the web and so it understands all of this vastness of what you could then click on and dive in. And what we see is when you're trying to buy a mattress, people want to see what experts say. They want to click in. They want to read more. And that's what Google's all about. And so I think those become really unique aspects of what Google does. Um, and then I think that the newest one, which is around visual and inspiration—you, I mentioned Lens and photos. People come to Google and image search for all kinds of things. Decorate my daughter's bedroom. I'm doing landscape lighting. I'm trying to design my kitchen. Find me similar page. Find me similar stuff. And but AI is is has been bad at inspiration. Like, if you talk—if you think about what AI does, it's been a language model. So it knows text. You talk to it. It grew up as a chatbot. Um, but how come it can't help you more easily shop or design your home? And so now that's something people come to Google for all the time. And you can now do that with our AI products and you can ask, "Help me decorate my daughter's bedroom." It'll give you beautiful imagery and you can have a follow-up question about it.

Marina Mogilko: It's going to be generated. Is that like...

Robby Stein: It's not generated. It's actually with web images. And so it'll find images. Yep. Inspirational images from the web.

Marina Mogilko: And then you can click on them. You can ask follow-up questions. And it's a way to use natural language for inspiration. But see, there's this little live icon there now. So if you tap on that, say allow. And then you can ask a question.

Robby Stein: Oh, with a video as well. You can use video. So you could even, yeah, you could point it at something.

Marina Mogilko: Okay. What is this device?

Robby Stein: What I'm seeing here is a Fly Tech AI NOTE Air 2. It's a digital note-taking tablet with an 8.2 inch screen.

Marina Mogilko: You don't understand.

Robby Stein: I'm sorry. I cannot make purchases for you. Would you like me to search for online retailers where you can buy it yourself?

Marina Mogilko: No, I already see Amazon is good. Thank you.

Robby Stein: Okay, cool. You've built so many amazing products. I mean, you built Reels, right?

Marina Mogilko: One of them. Yeah.

Robby Stein: That's fascinating. Now, you're building Google Search. From a product owner standpoint, can you give people tips on how to build in 2025 when everything is moving so fast, when everything is so competitive? Some people have this feeling that software like small startups are becoming a commodity because it's so easy to build something in two days. What do you think as a product owner?

Marina Mogilko: Yeah. Well, what's interesting is I think every product is a reaction to the time. It's like, what do people need and how do you make products today? So when there weren't any mobile experiences, everyone needed a mobile experience and people wanted more. And there was this gap in the market. So I think as a product owner and as an entrepreneur, you have to be a student of gaps. And so what are the things now that people wish technology could do better? I mentioned one with inspiration and how AI is not great at that today. So that became a big focus for what we're thinking about with AI mode and now we're going to be launching something I think exciting there. I think every entrepreneur and business owner can think that way. And I think how it's built increasingly, you'll be able to build things just by talking with language. You don't need to code. And even with really sophisticated things I'm looking at internally, you can largely just tell the model, "Hey, you, here's a data set, here's how it works, here's the schema," and the model can handle it. And so I think it's going to be very democratizing for getting ideas out there. So on the one side, it's going to mean that there's a surge of people trying ideas, which means there's more competition. And on the other side, though, it means anyone can try something. And so I think the grounds will be less—how technical is the idea?—and more—how interesting is the idea?

Robby Stein: How do you spot those interesting ideas?

Marina Mogilko: I think there's a few things. I think one is talking to people. I think there's kind of like two things I think about a lot. Like one is, like, how do you understand people? And a lot of it is you need to observe them, talk to them, research them. And everything I've ever built, you know, we spend a lot of time with people. You know, we watch them use our products. We ask them questions about what they're missing. We ask, what's the time you use this and you realize you want to keep using it forever? Like, that.

Robby Stein: I love that question. And that that's actually a critical question. Um, in actually this uh Clayton Christensen book, "Competing Against Luck." It's, is there an interview technique in there around eliciting that moment? Because that's the moment that caused the user to love your product. And if you can know that and engineer for that, that's going to get lots of people to love it. And inversely, there's a moment to say, when was the time you tried some of our product and you decided you're not going to use it anymore?

Marina Mogilko: And then that's how you get fired from your product. And understanding that is just as important. And so I think the people who understand people well and they try ideas fast—they ask them these kinds of questions to get at market fit—and they're building things that are resonating, um, are going to do really well in this next phase. Because I think before it was kind of like, "Well, did you have a good idea?" It could be a pretty good idea but not a great idea. But hey, you can build it and not a lot of people can build it. And so maybe it'll fly. But I think in the next era, lots of people will be able to build it. And so you really have to build something useful to people. How many interviews do you think you need to figure out if the idea is worth it?

Robby Stein: Honestly, I don't think a lot. I think like, I'm a big believer in small numbers, going very deep with like a dozen people. And if you give someone a prototype and you say, "Hey, here's a product. Try it out." And even your friends, they'll use it the first week because they're your friend. But if you look two months later, are they using it every day? I don't care how good a friend they are, they are not going to use your product every single day for months unless it's doing something for them. They're just not. And so, so the metric is make them use it every day.

Marina Mogilko: I think it's daily. I think for most, you have two types of products. You have really daily habits and products that you want to be mainstream large consumer products. You have other products that are utilities. They solve specific problems for people. It's like, okay, I have a telescope app for looking at the stars and using AI. That's going to be a different business model. You might, you know, have people who are hobbyists pay money for it. So, you want to study that demographic differently. But for mainstream consumer products, kind of the areas I've worked in, um, you're looking for daily value, stickiness, and stickiness. That's right.

Robby Stein: I love that. I think that was it. Thank you so much. Great. Agent calling made me.

Marina Mogilko: Okay, good. You'll be my first call when we have personalization and Gmail. We will talk the first day.

Robby Stein: Please, please. That's like, that is. I will