How Google is rethinking search in an AI-filled world

Google has built a massive business selling ads that appear around search results: In its 2024 10-K filing with the Securities and Exchange Commission, the company reported roughly $198 billion under “Google Search & Other,” its largest profit segment and more than half of its parent company Alphabet’s total revenue. But search is undergoing a foundational shift toward accessing the web’s information with the help of powerful AI models, and nobody has yet found a winning model for placing ads around AI search results. 

At the same time, new generative AI models can now handle much of the cognitive efforts users typically expend to arrive at their intended web content—and they’re doing it faster. This shift was evident in the search products and features that Google unveiled at this week’s I/O developer event, many of which are powered by this very reasoning capability. Fast Company spoke to Liz Reid, Google’s head of search, and Nick Fox, SVP of knowledge and information product, about how the company is navigating this seismic shift. This interview has been edited for length and clarity.  

How are you thinking about where search is heading?

Nick Fox: What [AI] means for search is probably the biggest shift in search ever. We’re talking about a shift from basic information retrieval to intelligence. These models enable a much deeper understanding of information [and] the ability to transform information. If we rewind a year, some of this was just theory.

Liz Reid: One of the exciting things with AI Mode is that while it’s our cutting-edge AI search, it also provides a glimpse of what we think can be more broadly available. Our current belief is that we will take the things that work well in AI Mode and bring them right to the core of search and AI Overviews. We’ve started doing this with technology like query fan-out [in which the AI calls for a number of sources for information related to the user’s search], so you can just ask whatever question you have right in the search box. When we think about the future of search, we consider a few different areas:

AI could be the most powerful engine for discovery because the ability for you to specify what you want means we can connect you to that really interesting niche page or artist that has something different to say that you’re interested in. We think it can end up transforming the web and people’s ability to connect.

LLM technology allows multimodality both in inputs and in outputs. Humans speak in different ways—conversationally, looking at images, seeing things before us. This is how we like to talk, describe our needs, and understand. AI allows you to say how you would really like to consume this information and makes that possible.

What does the introduction of reason agents mean for search?

LR: [Google] has been thinking about agentic work for a while [but] oftentimes that’s been confined by API integrations. The goal is to make search gather information, pull it together, and make it easy for you to take action on your information needs. This will be really exciting. We’re starting with bringing a lot of the Project Mariner (AI agent) technologies into search. This was something that was once easy to talk about, but now it feels like it’s actually going to be possible. This is related to the new reasoning aspect of LLMs and the tool calling ability to do tasks as opposed to just searching for a static piece of information.

Thinking about traditional search with the “10 blue links,” my brain was doing the work to figure out which link to click and to process all the information. It seems like with the new AI approach, AI is doing some of that mental work for me.

NF: While a lot of the narrative out there is “AI or the web,” we don’t view it as AI or the web, but rather the two of those really in concert together, building a holistic experience. A big part of that is exactly what you said, which is AI can help contextualize the web. AI can help organize, contextualize, and bring some of the simpler parts, and then you’re going to go deeper, often in a web page.

Robby Stein, VP of search products, mentioned giving AI Mode permission to look into search history, Gmail, and other Google services to personalize search results. What are the privacy concerns about accessing that kind of data?

LR: A key part of this is that it is genuinely an opt-in thing; we really want people to actively decide that they want to do it. We’re really starting with business messaging and recommendations. It’s not about personalizing something like a health response based on email content, but using information like the type of brands you like to shop for to make shopping queries easier, or the type of restaurants you order takeout from to recommend places in a new country. We’re going to start in Labs (for experimental products) to see what the feedback is and focus on recommendation spaces where there’s an overwhelming number of choices and it can be hard to express what you want, to specify your taste.

NF: My own usage has been super useful for recommendations, especially restaurant recommendations based on OpenTable confirmations in Gmail. Search understands some of this already, but the notion of which restaurant I actually liked is a particularly useful piece of information. 

LR: Our UX research shows people have very different views. Some people really don’t want it, which supports the opt-in. But a lot of younger users actually expect apps to be very personalized, and they assume we’re already doing a lot of personalization in search. For some users, we’re not meeting their expectations in that space, and we should do more. Others may just never want it, and that’s fine.

It’s like you have to get a sense of the zeitgeist around privacy expectations, and it’s a generational thing. I always think of it in a transactional way: I would be happy to expose my information if the return was obvious, and if you earn and keep my trust, and don’t start doing something with my data that I didn’t know you were going to do. 

NF: Yes, the user value of it has to be really high. We believe it can be and will be, but this is what we’ll learn with our users. It must be rooted in whether there is truly compelling user value by getting highly personalized, highly relevant recommendations for the things people are really looking for.

You built this huge business on showing search ads to people, and now we’re talking about this foundational shift in the way we’re doing search. Has your thinking around how you’re going to monetize AI search evolved as you’ve learned more about what it is and how people use it?

LR: We still see that a critical class of information is commercial information, where people are still often making [product] choices. So there’s still a large opportunity for ads. AI is expanding what’s possible. People are asking more queries, sometimes more specific queries. They’re telling more about what the intent is, which allows us to do more useful ads; you’re not just guessing, you can get more explicit. That is a real opportunity [for Google] from a business perspective, with more various opportunities.

It’s also really important for [businesses] that there are these opportunities for ads. If you’re a small merchant, often the only way you have a chance to break out is with ads. Otherwise, you just cement the brands everybody knows. That is how new brands come in and new merchants stand out, advertising to people looking for something they don’t know by name. If people get more specific about what they want, it gives more space for small merchants to show up and meet niche needs.

NF: What I’d add is the hallmark of our approach to advertising and search has been showing ads when they’re relevant and highly useful for the user, and showing very few or no ads when they aren’t relevant or the query wouldn’t benefit. I’m looking for a Mother’s Day present or gift ideas, which can be tricky. Having an AI response that gives ideas, and then ads that provide specific places to go buy them, is highly useful. Because we have a baseline understanding of how to monetize and thoughtfully display ads or not display ads on a search results page, it gives us the ability to get this right and what the user is looking for while also creating the opportunity for advertisers and driving the business forward.

Have you been talking to brands? Are they looking at this new mode of discovery and thinking about what they need to be doing to optimize for visibility and searchability in AI search?

NF: The ecosystem is figuring it out; we’re all figuring this out together. Historically, it’s all been about clicks—someone searches, do I get a click? That’s going to continue to be important for conversions. But there’s an additional piece: Brands themselves want visibility in that experience even if they’re not the place the user is going to go to buy. There’s value to the impression or the mention, which is something search hasn’t focused on as much historically. This is something we’re going to be talking to advertisers and businesses about.

It’s interesting that the total number of searches is increasing with the introduction of AI search. 

NF: There’s been this narrative out there that the web is dying for 25 years or something. What’s interesting is that if you actually look at the data, the web is thriving. We were looking at data recently, and Google’s crawler, which crawls a lot of the web, is seeing way more content than ever before. Our crawlers are seeing 45% more content this year—in April versus April two years ago. More content is being created, more domains are being registered, and third-party data shows visits to the web increasing over time. Liz talked about AI being an engine of discovery; discovery also leads to creation. Google is an optimistic company that cares a lot about the web. We truly believe this will be an expansionary moment for the internet and the web, and the data seems to indicate this reality.

Maybe you’re just making it a little bit more fun to search, more fun to shop.

LR: It’s certainly the case that if you reduce the drudgery and the effort, people search more. People have limited time, and if it’s just easier to do, if it feels like a joy and you don’t have to do the hard parts but get to do the fun parts, then people will do it more often. We’ve seen this repeatedly, from the web overall to images, Lens, and AI Overviews. Lower the difficulty, make it more enjoyable with the response you get, and then people just do more of it because it’s worth their time.

https://www.fastcompany.com/91338339/how-google-is-rethinking-search-in-an-ai-filled-world?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss