Here is a new interview published yesterday by The Economic Times Podcasts with Elizabeth Reid, VP Of Google Search. She spoke about Google's progression of AI in Search, content quality, SEO, blue links, traffic, the ecosystem, agentic features and more.
She said the AI won't be replacing search but rather augmenting search. One quote is:
And so we don't view AI as replacing search in the search experience. We view it as augmenting as enabling us to reinvent search.
As a reminder, Sundar Pichai said AI Mode will be incorporated into main search a while ago. Liz Reid, the head of Google Search, said AI Mode is the future of Google Search.
Here is the video:
Hat tip to Gagan Ghorta for spotting this:
🆕 Google's Head of Search, Elizabeth Reid interview with The Economic Times!
— Gagan Ghotra (@gaganghotra_) September 30, 2025
Full interview on YouTube - https://t.co/3o2RMMd3MM
Some key points and quotes from interview👇
AI in Search
"I don't think there was ever a single moment where it was like oh suddenly we needed to… pic.twitter.com/ZuPi9PWVaF
I asked Gemini to summarize the video (although, I think it got some parts wrong, like "SEO is changing" - I don't think Liz said that):
- AI's Role in Search: AI has been part of search for years, but its role is becoming more visible to users. Liz says there was no single "aha" moment; rather, it's been a gradual evolution based on technological advancements, trust, and speed. Google is still in the early stages of using AI to reinvent search.
- Changes in User Behavior: People are asking longer questions with more context and constraints. The ease and confidence of getting a good response from Google's new AI-powered features encourage users to ask more questions they might not have otherwise.
- AI Mode and Overviews: Google's AI mode has surpassed 100 million active users per month across the U.S. and India. The company is happy with the early reception and continues to see the "blue link" (traditional search results) as far from over. Liz emphasizes that people still want to hear from trusted sources and connect with human voices, especially for high-stakes decisions or personal topics like fashion.
- The Future of SEO: The SEO playbook is changing. Publishers should focus on creating high-quality, detailed content that offers a unique perspective rather than shallow, manufactured content. Since users are getting quick facts from AI Overviews, they'll click on links to go deeper, so the goal is to optimize for a great user experience on the landing page.
- Addressing Competition: Liz states that LLMs and Google Search can coexist and grow together. She notes that many people use LLMs and then come to Google to fact-check the information, which contributes to overall growth in searches. Google also aims to match key features like conversational context and personalization.
- The Role of Agentic AI: Google views agentic AI as a way to remove "grunt work" and effort for users while still giving them control. Liz mentions automating tasks like filling out forms or making phone calls, always with an eye toward not disrupting the relationship between the user and the partner/service provider.
- India's Importance: India is one of Google's top markets and its biggest for visual search (Lens) and voice. Google considers it a critical market for innovation, bringing new features like AI mode to India rapidly, often right after the U.S. Liz also highlights the potential of AI to make information universally accessible by overcoming language barriers and differing levels of literacy.
Danny Goodwin pulled out a number of good quotes and posted on Search Engine Land.
Here is the full transcript (excuse typos):
Our guest today is Elizabeth Reed, Google's global head of search. Elizabeth or Liz joined Google in 2003 as the first woman engineer in the company's New York office. She joined a team called Google Local, which has become the foundation of what you see on Google Maps today. Today, of course, Liz is steering how billions of people in the world discover information on its most popular search platform. In this conversation, we talk to Liz about how AI is reshaping search. Is it with search or is it replacing it? We press her a little on whether more and more people are shifting their questions to standalone LLMs. How companies are rethinking SEOs in an AI first world. How Google is reinventing it search to become more friendly and more conversational while keeping trusted sources up front. And of course, we ask her how and where India features in the big Google search story.Liz, very warm welcome to the latest episode of corner office conversation.
Thank you so much for having me today.
I've heard you and also Mr. Sundar Pichai repeat the fact that the customer is Google's northstar keeping that as a focal point. When did you first realize that AI and search had to marry?
Sure. Well actually AI coming into search has been for many years now. If you take the first examples BERT and trans and Mwere introductions of transformers in there. They were more what I would call behind the scenes. So they were the use of AI more to affect ranking. As the technology proceeded and starting with mom but really more with AI overviews. The the role of AI became much more at the feature level. It became much more visible to users. But it's really been an evolution. And that evolution is partly based really on where the tech was. What could you trust? What was fast enough? How was the information high quality enough? And so as we went along we could bring it further and further. So I don't think there was ever a single moment where it was like oh suddenly we needed to do it. We just gained more understanding, we gained more conviction about how to bring them together. And despite all the several last years, I think we're still really at the beginning. I think there's still huge way to think about how we use AI to reinvent search. And it's really one of the most exciting times in search ever.
Would you say big tech companies like Google take a little longer are taking a little longer to respond to the innovation than more nimble-footed you know startups or or slightly smaller companies like say a Perplexity do you do you do you face that problem?
I think, in search one of the things that Sundar will talk about is this idea of bold and responsible so we want to move quickly but we also have you know multiple billion people who trust us every day. And so figuring out that right balance I think is really important. I think also if you go back in time in Google, people sort of think of Google in regards to Google search, Google maps, Chrome. We were not the first search engine. We were not the first map application. We were not the first browser. We really want to come in and innovate early but also do so in a way that produces a really high quality product. And that is really thoughtful to the trust that people place in us. And I think we're really at the beginning of the journey, as opposed to at the end and really excited about what the future is for search.
Let's go a little deeper. You know, I've heard and seen multiple interviews of yours, and what you've written saying that the one big thing that has changed now in terms of how people search is that the questions are longer and much more. In fact, I think they are 2 to 3x longer. What else has changed in the way people are approaching Google search?
Yeah. So there's a few different aspects. One if you think about the 2 to 3x longer then the question is why are they 2 to 3x longer? and they're longer because people are giving more context. They're putting multiple constraints. Okay? If they're going to go find a restaurant, they're not just searching for restaurants. They're not just saying kid-friendly restaurants. They're saying I have a four-year-old and a seven-year-old. I need to ensure it's family-friendly, but I need to be outside. Can you please find me restaurants? Right? So, they're they're expressing more of what they want. They're also just asking more questions, right? There's this sort of idea that many people have that the number of questions you have is sort of fixed and that you're already bringing them all. But that's not actually what we see. What we see is there's sort of an implicit calculation that users have which is like is it worth my time? Am I confident that a response is going to come? Is it going to take too much effort? How do I think about that? And so when you lower the bar to actually ask that question and get a good response, people ask more questions. And so they ask longer questions, but they also just pick up the phone and they ask a question they wouldn't have otherwise asked because it just feels like now it's worth their time. They start exploring an idea or they ask a question and then they ask a follow-up on AI mode because it's just so much easier to do a follow-up. If I had to go repeat my entire context, all everything I said before, it wouldn't be worth my time. But if I can just ask a quick follow-up, then I'll go ask it. And so we see people just asking more questions, not just harder questions.
Got it. Is it possible to quantify or give me some sense of what the adaptation of AI mode has been like?
So we're still very much early on in AI mode at this point. So far with India and the US, we're already seeing over 100 million active users per month. Okay. We just started to roll it out in more countries. And I think we'll continue to see it grow. But we also think about AI Mode and AI Overviews really as two different sides of the coin. And so we have both a lot of people using AI mode but also a lot of people using AI overviews and exploring how the two of them go together.
Right. Right. And are you happy with the 100 million users that you've got in India say or the US?
You know we always seek to try and answer more and more user needs. So we'll continue to do that but we're only a few months out if you think about it. We just started rolling out AIM mode to the US in May. We started rolling out late June to India. So we've been pretty happy with the reception so far. So unlike what a lot of people say, the story of the blue link is far from over. I do think the story of the blue link is far from over. I think it is the case that many many people want to hear from other people. Okay. And we see this true, both with our younger users and older users, right? AI can help make quick questions easier. It can help pull information from different sources together. But whether it's your fashion advice or a high-stakes decision you're going to make on a purchase, they often want to hear not just from what the model has to say, but they want to hear from other people. And so, um, what we see is that one of the things people really like about AI overviews is that ability to pull together the AI response. And the web links. I do think we see an evolution in some of what they're interested in and that they want that depth of experience. They want that personal view. So they're more interested in web content that is that is rich that is not sort of like just the shallow basic fact but brings in the unique perspective or that creator view. But to the extent that like people want to connect with humans and that human spirit and the fact that the world is very diverse, right? one answer isn't the same for everyone. I think there's going to continue to be a huge demand to hear directly from people. I like I like the the value of AI responses, but I also don't like to everyone when they're like, "So, you know, I'm going to tell you what so and so said." I'm like, "I want to hear what so and so said directly, right? Don't tell me what so and so said. Connect me to that." Right? And so to that extent that you don't want to hear the interpretation of someone but you want to hear from them directly then I think web content short form video user created content will continue to flourish, right and is that is that happening now? Do I directly reach forums where I get to hear actual experts speaking instead of the summary or is that still work in progress? So as a user one of the biggest things you would have seen in search over the last few years and it shows up in in AI overviews and AI mode too but certainly throughout the search product is that there is more and more content from user generated sources right you see more demand for forcontent you see more demand for short form video and we respond to your previous point about the customer the user being our northstar so as we've seen more demand for that we've integrated more of that content in and I think It's an area for continued exploration. We're also been experimenting with things like inline links within AI overviews and AI mode. So, say you know according to the following site or according to this creator and then linking directly to that creator so you can hear more. Um, but I do think that's really essential that we continue to highlight the voices of the creators of the trusted sources of these folks with deep expertise and experience in spaces.
Right. Right. Do you see a gap in knowing the intent of the user? For example, when I go to an LLM, I say I need to email my boss or I need to email my listeners this. But when I have always done Google searches, the intent is never very clear. I think there is a gap there in synthesis. Do you is Google still trying to sort of fill that gap?
I think on on the good side to the point about two to 3x queries that that you were mentioning two to threex longer queries as you were mentioning we do see people are adapting but there is there is definitely room to go there's a lot of people have been using Google for years and they have an expectation about what's possible and so we're going to continue to experiment in how does the user experience change not just as a result so that people know they can ask longer queries and so people can provide provide more of the context today in a way that if they tried to do so two years ago, it wouldn't have been useful, but they don't do it right. They don't realize they can do it. And so that is an interesting user challenge to figure out how you can teach people both if you love expressing this query this way, great. But if you want to share more context, you can. And so we'll continue and then also with personalization, right? How do you maintain that statefulness, that personalization? So a user doesn't even always have to express that context.
Liz, one thing that I realized is that a lot of people when they ask questions to standalone LLMs, they don't seek provenance. I mean in the sense of where the information came from etc. I do of course but even so your search quantof searches or the number of searches you say have really gone up. How has that really happened?
Well, so I would say a couple of things. Sometimes people don't seek providence on the LLMs. They come to Google and they seek their providence. So we see a lot of people go ask the question and then they come to Google to double check it. So you see that. But I think it really goes back to this idea that we're really in this expansionary moment. Okay. That the number of questions that people ask is not fixed. So there's no sort of zero snature here. Okay. Actually as the tools get more useful people are just asking many many more questions and so it's actually possible that everyone can grow at the same maybe not everyone but lots of people including Google can grow in this time because it's just much easier to ask questions and it's just more rewarding and you know if you can just ask a basic question about restaurants in your town why bother but if you can share all of this information about where you're going then maybe it's worth it if you feel like to go learn a new subject is very hard then you're not going to necessarily spend the time. If it feels really easy then you can take five minutes out. Okay. I used AI overviews to learn a bunch about cricket. I still don't know very much about cricket so don't quiz me on my cricketut. But I learned various facts over the time because I had friends and co-workers who were big into cricket and they would make these statements. I'd be like I don't know what you're talking about. And so I would just type this question because I didn't want to show up just how dumb my cricket knowledge was. So I just ask this question. I get a response. And so it really just is this moment where the pie is growing in a pretty tremendous way.
But I love how I think AI overviews has seen a 10% increase in usage and also your clicks have grown. So maybe many of those clicks are coming from other LLMs when they're trying to sort of fact check what they've got. I love that. But but is there are there features from other LLMs that internally Google would want to match you know be it context. So when I ask a question today on one conversation versus when I ask a question on that same conversation tomorrow I don't have to give context because the LLM usually knows. then there's of course inbuilt other apps. If I if I want to for example generate a picture on midjourney, I can ask chat GPT to give me a proper prompt. Which of these features do you think that you really want to match and which ones would you would you not take because that they would sort of maybe dilute Google's brand and the kind of structure that you have.
So I think about this a little bit differently. One is sort of what are the the capabilities that LLMs provide and how do we unlock them for search and then sort of what is paramount to what makes a search product right so so things like context and personalization we talked at IO and started to roll out some of the features and we'll continue to roll out more of those this question about like how do you do follow-up questions easily both like in the session and on on from session to session right has been pretty critical to how we think about AI mode. We've talked at I/O and started to roll out in the US and hopefully soon to India more of some of the agentic capabilities, right? So these fundamental things where where a capability unlocks either new questions you can ask or the ability to ask the questions more easily. We talked at IO about the goal of making it possible that you can ask truly anything on search and that you can do so effortlessly, right? So what are the things that you bring in from LLMs that make it effortless? So I think you know anytime we feel like it does make it effortless like then we want to take a look at those underlying capabilities. On the other hand I think there is a few things that are really essential to how we think about search. We want to do everything we can to maintain that trust and to nurture that trust. Okay. So if we feel like we have to make a trade-off that breaks the trust then we don't want to do that. A second thing that is essential is we feel like people do come to search often to connect to other voices to hear from other people. And so we don't view AI as replacing search in the search experience. We view it as augmenting as enabling us to reinvent search. And so how do we bring those voices forward and allow you to get started on Google but to make it easy for you to click out to the web to go deeper throughout. And so that that sort of web forward ecosystem for part of an AI search experience as well as trust are things that we view as something we can't break.
So rookie question here and I may be asking you to repeat yourself but when you pull your deep research from multiple sources. So the ranking of those search results is this the same kind of process that you use on in the blue link format? I mean just from a trust point of view obviously.
Yeah. So, mostly yes. Okay. So what we do with, something like deep search is we are querying web search underlying ranking. And so in that sense it's the same. It's different in the sense that the the a question a user is asking with deep research is often more complex than any single web search will go. And so we'll take that search query and we'll break it into several queries. Right? Okay. But once you issue those queries, we're basically hitting our underlying core web ranking the same way we would if you had sort of gotten the traditional blue links. And so we are really building on top of that long tradition and expertise and trust in web ranking but doing it to answer questions that are not possible as a single query. I still think we're going to we'll continue to innovate in that space. How do you get better at doing the query fan out? How do we get better at the model, understanding how to take advantage of those different links to give the most high quality response and pull it together. But it is really built on the same underlying web ranking.
Okay, let me flip this a little bit and talk about companies that feature on those blue links. Okay, or entities that feature. How is discoverability changing? You know, how is the SEO playbook changing in the era of, you know, AI search and AI old view and AI mode?
So I think there's a couple of different things for companies to think about. In some sense, the guidance is very much the same and in some ways it's different. So the guidance that we've given that we would continue is people should really produce content that users care about and not think about building content for search engines. Okay. Um, but that only becomes more pronounced now. Okay. If you produce content that's very shallow that you're just going to hope is ranking at the top but doesn't really have much to say, then your content really doesn't have much more than like an AI overview would give in the first place. Okay? So, people aren't going to want to click go check the same fact that comes. Okay? Um, but if you produce really high quality content, you bring your perspective in, you're bringing your experience, um, you're doing something that you want to read for a minute or something, okay, then users are going to want to click in further, okay? Because they're not clicking on the link to get that 5-second response. They're clicking on the link to go deeper, okay? Um, so there's more push on sort of niche content, more push on sort of the detailed understanding of the space that you're coming in across. So less manufacturing of content and more like actually putting yourself in the space. The other thing I would say is we would encourage publishers and creators to think about it in a different way in that there's what we see in some of the traffic is that there's less of what we call bounce clicks, okay? Where a user clicks and they immediately click back, right? They're like, "Oh, I thought this was going to be useful. No, it's not." Okay? And because we're giving enough contacts, people kind of know and they don't have to do as many of those bounce clicks, okay? So, what you do see is a lot of excitement on the the deep clicks. Okay. And so, I think that like as a publisher, you want to really think about like what is the user's experience when they get to my site, okay? And so sort of be optimizing for a great experience when you land on that page. Because people are going to want to go there not to just get that 5-second thing, but to actually go go deeper on that. And so in some sense that's always the content we wanted people to create and it's always the content that users really wanted to see but you kind of can't get away with content in the old way. Now you really do have to build that content for users.
So in a recent interview Mr. Sundar Pichai said that the ad revenue from AI overview search results is the same as without. Now I may be getting into a little complicated territory here but in the simplest possible way a lot of our viewers would want to understand how is that possible so I think there's multiple different dynamics going on on the one hand sometimes when AI Overviews show up the ads are lower on the page on the other hand the search traffic is just growing overall right so we see with AI overviews people issue more queries and so at the highest level the two roughly balance out okay let's come to agentic AI how so as a big tech company you can't chase every innovation that you see that others are also chasing right and agentic AI is still you know in its in its somewhat nent stages and I'm again talking about this as a user project mariner is of course still a prototype but do you see the human agent interactions really evolving to where people are saying they would evolve and you mentioned India also in this context. So I would like you to double click on that as well. Sure. Well so I think the word agentic means a lot of things to different people as they come. I think the way that we think about in search is if you go back to this idea that you can ask for anything effortlessly. Where is their grunt work? Where is there a bunch of work that a user just doesn't want to spend time. It's too much effort. And how do we take that work out of it? Okay. And in some sense things like AI mode took out of the groundwork of sometimes having to issue many many queries and pulling the information together by doing it. Okay. But then you can think about at the next level whether it's from a productivity perspective or maybe dealing with transactions. How can you actually reduce that? Now there's an interesting part to your point which is the tech is nent and also user expectations are changing and early on and so we have to figure out how we both take out the groundwork but don't eliminate control where a user wants it. Okay. So maybe we can do in some cases if a transaction is really important, maybe we can do all the a lot of the work to get to the point where like most of the initial form filling is there, but you maybe still want to click yes on the transaction rather than have us do the final transaction for you, right? Okay. On the other hand, maybe something is repeat and you're just like, I just want to do this again and again. Fine. Okay. Um, one of the things we did, in search over time is help automate a bunch of the phone calls. Okay, so we've been expanding the ability like if you're trying to get a haircut or you're trying to get a plumber or something where actually in a bunch of cases it's not even just about the browser um, doing that like you actually at least in the US you actually have to pick up the phone and call a whole bunch of plumbers to get out basic information and then go figure out okay can we eliminate um, the work around that? And so we've done work on that. And so we think about that also from a tech perspective is there not a single solution. Maybe we have to automate some of your phone calls. Maybe we have to automate some of your browser actions. Maybe there's times where we integrate on the API side. And we should pull them all together. So that the experience feels fast. It feels like it's empowering to you. It feels like it's good for the ecosystem as well, right? So we don't want to disintermediate the partners that we're working with. There's different aspects that may matter to them. Do they still have a loyalty and relationship with their customer? Do they still have the ability to upsell? And so when we want to think about a experiences, we really want to work with them so that the user experience feels as easy as possible. The user feels empowered, but the partner also feels great about that as we go. And I think it's something that will change every six months. And it'll be easier in some industries than others. The stakes are higher in some cases than others. And we'll continue that. To my point about India earlier, India is like really such an important market for search. It's one of our top markets overall. It's our biggest market for lens, our biggest market for voice. It's a market that like we brought AIM mode to India right after the US before any other country and within basically a month of rolling it out at scale in the US. So it's such a critical market for us to innovate on. And so, you know, if I looked a few years ago, if we launched something in the US, like I don't know, maybe a year or 18 months later, it would come. Now, we're really trying to shrink it down. You know, two months, if not weeks. Um, search live is coming to India soon. I'm delighted to share. So we're going to continue to push on both bringing innovations rapidly to India but also thinking about where does India actually prove a ground for innovating itself like it's not always going to be second you know oftentimes it will be second and not third fourth or fifth but sometimes I think it'll even be first going forward India is really rapid at at adapting a bunch of the AI innovations and you know when you see the uptake with things like lens and voice it really feels like it's pushing the limit. The users are really challenging the systems and so we really need to respond and innovate in them. Great. In terms of tech talent, how is India contributing to Google's shift from information to intelligence? We have a great we have a great team in search here and and it actually cuts across our whole system. So from the underlying infrastructure that makes it possible to serve at low latency and high capacity to what is our work on innovating in the actual core intelligence and the quality of the system all the way up through some of our different vertical experiences like sports or local and travel we have quite a meaningful presence in search. I also know that beyond just search we work closely with Google deep mind which also has a presence in India here and I would say the India team here is both great at helping us do things like localizing for the market and really understand what it is but they aren't just doing India specific work they're really building both foundational work and working to figure out how it excels in the country. Right. Right. Speaking of localizing, in fact in a recent podcast one of the things you mentioned as the huge potential of AI is how it's revolutionizing the use of language and in fact you specifically mentioned Hindi. Now in India the landscape is that a lot of local AI first companies are trying various things in terms of vernacular LLM and how to use AI or how to use local languages in in in terms of AI. But they have seen very limited success. Would Google with its huge you know treasure of data and its capabilities do something on that front in India? I think this is an area we're going to continue to invest in quite deeply. We've sort of had state-of-the-art models on language understanding and been investing in them for over a decade now. You know we brought not just Hindi rapidly to AI mode and AI reviews but actually lost launched one of our first cross language experiences with AI overviews toggling between Hindi and English last year as they come and we rolled out AI overviews in languages beyond Hindi and look to do the same with AI mode. I think one of the things that's really exciting about what LLM can do is that not all the content is available in all the languages, right? Certain content is available in one language but not in the other. And historically if it the content didn't exist on the web in your language, it was kind of locked out from from those users, right? And LLMs are allowing us to sort of take the learnings from one language and output them in a way that's understandable for the users. I think there's still a lot of innovation to go. Um, to your point about how do you get vernacular right? Um, how do you really understand, cultural topics in the in the question? How do you mix, you know, if we talk about Hindi, but it's really in many cases not just Hindi or English, but English in there. How do you adapt to that question? how do you actually make it such that you really feel like language is not a barrier right I feel like everyone you know we made this comment where Larry and Sergey set this vision of make information universally accessible okay but it hasn't really been universally accessible in many ways with language but it's also not just even with not just across languages but also even within a language what's your level of literacy on something okay Um, I may consider myself to have fairly good tech literacy. I have okay financial literacy. I have pretty poor medical literacy, right? Like you give me a medical paper and I'm like, "Oh my gosh, I'm going to have to spend five hours." Okay. Um, but if I can learn about a topic that matters to my daughter, that's really powerful to me across. And so, I think both cross language and within a language and across voice versus written. Um, we also see this with visual understanding, right? I don't know what it is. You know I come to India I see a sign in India I have no idea what it says if you ask me to type it in like I have no idea how to even transcribe okay but I can use lens and I can translate it in the one and so I think there's a lot of exciting work ahead that really just unlocks intelligence for everyone right it's not enough for access to information if you don't understand it and now we'll be able to help people understand.
Finally I'll refer to another podcast and this is a few years back and and you you told the host, "If I were to take Google away today, you would go nuts." Do you still think that's true?
I think for most people, absolutely. I think a lot of people like what we see from our data is that people are still using Google and even even as sometimes they're adopting new tools. They continue to use Google and that's because it really is much more overall an expansionary moment. Um, there's also there's a tendency for us to forget sometimes about what you use search for. Like you use search to navigate to URLs for all the advancements with LLMs. People are still clicking to their favorite sites, right? You use it to go and browse. You use it to go look up your favorite Hollywood actor and you don't even know what to ask. You just want to know what's going on and we'll bring a visual Search. All that said, it's a space that we are going to continue to innovate, right? We always want to figure out what is the next frontier for being the most helpful information to actually providing in a way that feels easy as possible. And so we're going to be relentless at that. but you know I really believe that we will continue to have search be as important for people over the next decade as it was for the last couple.
Liz, thank you so much for joining us on Corner Office Conversation.
Thank you so much for having me. A pleasure and I hope to chat with you again sometime soon.
Absolutely. That's it for today. We hope you liked this exclusive interview. It was produced by me, Vine Jooshi and Rajasthound designed by Amrid Reji. Please subscribe to Corner Office Conversation where we talk to the most interesting business leaders from India and the world. A new episode drops every alternate Monday wherever you get your podcasts. Keep listening. Have a good week.
Forum discusison at X
