Understanding AI-Powered Generative Search

This article is a transcript and a recording of the "Understanding AI-Powered Generative Search"  webinar. In this session, Marina Petrova, an expert in AI and digital marketing strategies, provides valuable insights into the evolving landscape of search and discovery and what it means for brands. 

Search isn’t what it was even a year ago. AI is rewriting the rules. Generative Engine Optimization, or GEO, is a new term used to describe how companies can have their businesses and brands be discovered in what’s quickly becoming new touchpoints. OpenAI’s live search functionality within ChatGPT, growing in popularity Perplexity, and Google’s AI Overviews are having a major impact on traditional search practices. Every carefully crafted word in brand’s content is being reinterpreted by AI-driven search engines, shifting the balance from controlled SEO strategies to the dynamic, unpredictable world of GEO. 

Key webinar insights include:

• How generative AI reshapes your content into responses you can’t fully control.
• Why traditional SEO practices are giving way to intent, context, and relevance in GEO.
• Practical steps to ensure your content remains discoverable in the evolving search landscape.

Transcript:

Introduction and Overview of Generative Engine Optimization

Thank you, everyone, for joining today. There is a lot to cover, and a lot of information. I look forward to this session today. Let me share my screen, and let's hope all the technical things go as planned. Today, we'll talk about generative engine optimization. To be honest, I really don't like this term - generative engine optimization. This is something that probably will become as common as SEO, but in reality, it is so much more than just optimizing your content, your information, your brand, your voice for the new search. 

Today, we'll talk about what this new AI-powered generative search is, and I'll share as much information as I can. It is not going to be a technical session. It is more about truly understanding the essence so that you can be ready for what's coming next. We'll talk about search in general. We'll talk about traditional search and what still matters [at the time of the webinar] in November 2024.

Emerging Trends and Intentful’s AI-Powered Solutions

We'll talk about how you’re probably already seeing a decline in organic traffic, and if not, heads up—you will, and I'll explain why. We will talk about the zero-click search, what it is, and what to be ready for. We'll, of course, talk about AI Overviews and compare the SEO with GEO. And, most importantly, [we'll talk about] where to get started and what to be looking at.

There is a lot of information. We all come from different backgrounds. We all have different contexts. For some of you, a lot of this may seem known and obvious. For others, there is a lot of new stuff and discoveries. I do hope that you will be able to get some learnings from this session and apply them in your day-to-day work. A little bit of introduction: our company is Intentful, we started in 2021 in the AI space. And what we do is connect companies with customers through our next-generation intelligent products.

This is not a presentation about our products, but there is a reason I want to talk about them for a second. All of our products are AI-powered. We have Generative Response Ads, a new product that we just recently launched, where a user can have a conversation with an ad right within the ad space. There's also an AI Assistant— an intelligent bot where you can have a conversation in natural language, and a Data-Driven Content Platform. All of these are powered by AI models that know your business, your content, your destination, whatever your organization is.

The reason we are having this webinar today is that all of our products are powered by AI search. It is very similar to the generative search we will be talking about today—the generative AI search that is now used by Google, OpenAI, and everyone else. Over the years, we have deployed multiple projects and continue evolving this AI search component, so we have this hands-on knowledge of how it actually works. How AI knows what to respond, how AI knows what exact piece of content to take.

That's why we want to share this knowledge with you today and are happy to continue the conversation. We are just getting started in the AI space; I mean, we are—as is the world.

Another reason why we are doing this session is that since day one, we made it our mission at Intentful  to advance the practical use of AI and share the knowledge that we have with the world, and see how this knowledge can help improve businesses. As the world keeps changing, this becomes more and more critical. Our company was acquired by GMS, which has been at the forefront of global communication solutions for almost 20 years. GMS works with over 250 telecommunications companies worldwide and 700 enterprises, and we are happy to be a part of the larger GMS family.

Evolution of Search Experience and Zero-Click Search

So now to search. Before we get to the generative search, I want to give you a little bit of an overview of how search works in general because a lot of this information will help you understand how search will be evolving and, of course, what still matters in organic search, in traditional, so to say, search.

For those of us who remember when Google just launched, Google was a window to the Internet. It was the very first interaction to get to any website. Of course, there were some exceptions, but Google was the first point of entry because it was the desktop, and everything started with Google. So, of course, every piece of content had to be optimized to be discoverable on Google.

Then, with mobile, that was the first time there was a significant change. The reason for that is we go directly to an app. We don't go to Google when we are on the phone; we go directly into an app. And that was the first change when Google traffic started changing, when Google had to adjust what they do. Search has been migrating to the channel where the customer is. How people interact with information has changed because we no longer go through that first point of entry. We search right within an app, be it social or the messenger, or if it is an image search, or we start using Siri and Alexa, and shop on Amazon, and then apparently ChatGPT.

There are three key streams to keep in mind in terms of how search has been developing and how it will continue developing. It's how users interact with information. As you all know, this has been changing over the last 10 years, especially during the previous couple of years. How technology is evolving. AI is not new. Google has been using AI for ages. But now, with generative AI, this is changing significantly. And, of course, there is always a commercial component—what it means for businesses, what the revenue models are, and what makes sense and what doesn't.

The reason I'm bringing this up is that a year and a half ago, there were already signs that search was going to change. Google said that, you know, it's actually changing. It's different from what it was before because it feels like, at a high level, people are asking follow-up questions. Because of this, a lot of information was discovered on other channels. So, Google became a follow-up channel. And then there was a mind-blowing stat that I have continued including pretty much in every presentation for the last couple of years. According to the SEO company Ahrefs, almost 95% of all questions people ask on Google are asked 10 or fewer times per month. Keep that stat in mind once you start thinking about how you optimize the content further for generative search and any type of content you work with because people want to know different things.

Another important thing here is shopping. While it may not be directly relevant to your organizations, to your companies, it is relevant to understanding the overall search ecosystem and how it will keep evolving. 

So again, remember that moment when Google was the first entry point to the Internet, then mobile took a part of it, and then Amazon because now most of the search queries that are shopping-related start on Amazon, at least in the US. It is similar in other countries as well, with their own versions of the key retailer. 

This also had a significant influence on Google's revenue, and therefore, they had to start looking into how organic and paid searches are displayed.

And please bear with me because we are getting to generative search. It is all important to understand what you need to do to make your content discoverable. You need to understand the ecosystem. At least, that's what I think is important.

Remember those days when it was easy? You just tried to get into the first three positions.  You know this is 75% of all clicks; ideally, it's organic, or it could be an ad, but you know the target. However, because of all of those changes, Google has started improving the user experience. Let's put it this way: because user experience is always at the core of every product, including search engines like Google and, of course, the new players like ChatGPT search or Perplexity—we'll talk about them later.

What they started doing is introducing featured snippets where the answer is just shown at the top part of the page. You all know this; you all experience this, and this usually means that this does not lead to a click. It may seem counterintuitive because then it means that Google is not really making money on this. But it is way more complex than just a direct link or a direct app that is placed on the homepage. For Google, the most important thing is to make sure that users can find the information they are searching for, and if this means providing a quick response like this, then so be it.

I don't know if you have heard about this term called zero-click search, but that's basically what's driving this trend because people get the answer immediately in this answer box. And that means it does not lead to a click, and what I was showing just now is one of the examples, but there are many ways this content is served. And this is in early 2024. For those of you who are interested in search, I really recommend following this company, SparkToro. They have a lot of interesting information.

Source: SparkToro

The zero-click search trend, I believe—I may be wrong—but I believe it started even before COVID. It was like 2019, probably maybe even at the end of 2018. But it kept growing, it kept changing. And so, almost 60% of all the content displayed on Google gets zero clicks. And this is for the US, but it is very similar [for other markets]. Maybe there is some variation in other countries, but it is the trend, it is the essence of search. This means that your content is still likely to get discovered. Your content is still likely there, but no one clicks on it.

Impact of AI Overviews and Search Evolution

So, when you look at your Google Analytics, where do your users come from? You will not see this because there was no click. And then, let's look at what happens when there is a click. So, out of this remaining 40%, a significant share goes to unpaid organic results. Then, a rather significant share, almost 30%, goes to other Google properties: YouTube, Maps, News, and so on. And then the ads, which are in search—we're only talking about Google Search here—right, go to paid ads. 

When you think about where your users are coming from or where you think they are, what you see in your reports and attributions, Google Analytics does not always show what the real picture is. This is something that I would like you to remember as one of the key learnings from the first part of our session, and it will continue being important as research evolves.

Then we get to AI Overviews—that's what Google calls them—and we'll get to other players, not just Google, in a minute. AI Overviews are how Google refers to the answers that are being generated. 

In August, it was just less than 8% of all searches that triggered AI Overviews. It all made sense, so it was secure websites, .edu, .gov backlinks, and the usual SEO-valid content. Most of these were also leading to the top 10 organic search results, and the ones they were using were the ones with high authority. But this has been changing incredibly fast. And there is an article—happy to share the link afterward, or you can just Google it—where, at the end of October, Google said that they are rolling out... Let me play this here. They are rolling out AI Overviews globally. I believe it's 100 countries at the moment and that AI Overviews will reach more than 1 billion global users every month.

Source: Google

What we just talked about—those zero-click searches—so get ready for your traffic to continue dropping, because while there is a link that is included here (points at a link in AI Overview response), it is very similar to the featured snippet or that answer box, because a user is getting the answer more or less immediately. So when you see your traffic drop, that's fine. It's just the way content discovery is evolving.

In the AI Overviews, from what we understand now, a lot of this comes from our understanding based on how we work with AI search in our products. The core SEO principles remain the same. Google will always focus on what's important for the end user, for the visitor. So, they will continue updating the algorithm to ensure response relevance. What changes is how this information gets presented. Instead of a list of links, as you know, it is a generative AI that is used to form a response.

So, how exactly does that information get into that response? There is a little bit of a technical thing here but, still oversimplified. Think of it as there is a lot of information. So, what AI does is break down the information into chunks and analyze those chunks for relevance. It's no longer about keywords. I think that keywords will exist for some time, but really forget about the keywords. It is about the context; it is about the topic, so AI analyzes how close it is to what the user is interested in. And there are multiple things: it's intent. We'll talk about intent as well. It's intent, its relevance. And there is also some other data that goes into this. But it is not about the keyword. It's also about the context and surrounding text.

And then the next thing, as you probably know by now through multiple presentations about AI, is the way it works; it predicts the next best word based on the data that it was trained on. But it's also important to know that it is not just algorithmic probability but also context-driven prediction. It will look for the context in terms of this prior word to ensure a logical connection and that this gets better and better every single day with all AI models.

Once it has done these first three (points at example), the answer is provided as an aggregated response from multiple sources within the right context. I understand that's a lot to digest. I've been working with search for many years, and with AI for almost 4 years now, and it is still something that I learn every day. But it is so important to start understanding this new way because otherwise, your brand will never get discovered.

Core SEO Principles and Adaptation to AI

If you look at SEO 101, it has always been about crawler accessibility, and I can't tell you how many companies do not think their website needs to be open for the robot to discover it. It needs to be accessible to the crawler. It needs to have relevant content that usually matches the search query, keywords, and user experience. The website needs to load fast and needs to have a user-friendly UI. This is still key. Google is about users being happy. And that means if I am Google and showing a website that does not load fast, then the user will be mad at me, Google, because I gave them a website that is just not good enough.

You all know the backlinks currency and how important it has been to have as many other websites linked to you as possible over the years. And then you need to have all those beautifully optimized titles, URLs, and all of that, and then still relevant, by the way, are snippets and schema. Schema makes it easier for the AI crawler to find the relevant information. If you don't have a schema on your website, please check it. It will really help discovery.

We're still talking about Google AI overviews. What's changing now is, yes, you still need to have relevant content. Don't try just what you know was stuffing keywords. But keywords get changed into topics and context; this is super important. 

Evolving Content Strategies and the Role of AI

Then, before we go into the next one, many years ago, I was thinking, why do we need long-form content? I really couldn't understand why Google wanted to have an 800-word article or a blog in a world where we get information in seconds. If we want everything on the go, why an 800-word length? Only when we started working with AI did I realize that AI needed more context to get a better idea to understand them better, that it became clear that the longer your content is, or at least was, in the old search world, the better it would be. It would be much easier for the crawler to understand what your content is about. This is slightly different now, but you still need to have more than just keywords. It needs to be topics, and it needs to be context. 

The next one is mentions. So I can't really say if backlinks are going away or not, but now, your brand, your content, your destination—whatever you're promoting—needs to be mentioned. It doesn't need to have a link. But if some other website mentions you and provides some relevance, they will be able to pick it up in the world of new AI crawlers. So this is something great, especially if you have a PR agency or earned media programs, and this is something to pay attention to. 

Clicks and traffic matter for Google in 2024, including paid ones. Everything else stays the same, at least for now.

ChatGPT Search

A lot of people used ChatGPT when it launched from day one as a search, but that was wrong. I was very concerned about this because it was not for search. It was a large language model, and that had cut off at its training data. But we don't need to spend time on this now. At the end of October, ChatGPT launched a live search. It is, for now, still in the Plus and Teams users. However, that is already a Chrome extension that changes so many things because for those who do have this paid version, and again, we all know that this will change soon to non-paid as well, this changes the search experience as well, and you don't need to go into Google or any other search engine, but you can use ChatGPT through a browser extension.

Source: OpenAI

There is a lot of information on the OpenAI website about how it works. A lot of it is rather technical, but it is similar in terms of how I was describing how it looks at different information and the relevance and context. I'd like to pay your attention to the first areas where OpenAI is looking at search are shopping and travel, and this is especially important for DMOs and travel brands who are on this webinar today. I would suggest continuing to pay attention because travel will likely be one of the first industries to be disrupted here.

Adapting to the New Search Paradigm

If we look at the old search, and yes, SEO was a huge industry, you had significant control in terms of how you can have visibility in search. It's content optimization. It's keywords. It's backlinks. In the new search, you really don't have much control in terms of what appears in the AI-generated responses because they come from multiple sources. And while that is still less control, you can start working on those content programs now to ensure you are represented in the new search. 

In old search, we had owned media like websites, blogs, paid ads, and earned media like PR mentions and reviews. The user's journey was to click, visit the website, refine the search, and repeat. In the new search, which is key, we have aggregated sources that come from real-time data now, not something that was in 2022. It understands context. Then, there is the ability to continue the conversation. The user journey is to receive a direct answer and refine it with follow-up questions. While we cannot be sure that the user will act immediately, most likely they will. So it's very different from the way it was before. 

The Role of AI in Brand Representation and Customer Experience

I mentioned to some of you at some of the conferences we met during the last several weeks that about 3 weeks ago, I was traveling to North Carolina to The Outer Banks. I was arriving in Norfolk, VA and had to rent a car. And when I was searching for the car, that was the day when ChatGPT announced the search. And I actually started [to search] in ChatGPT just to see what they show. When you look at Google, you get the usual search results we know: locations, brand websites - the owned assets you can control and optimize. It took me a couple of times to get to the AI Overview, so I was not yet in that 1 billion users a month. But I finally was able to get there. The way the information was put in there is the same as here. So, it is similar to the approach. It pulls it from different sources and suggests it as an AI Overview. 

However, when I did this with ChatGPT, I honestly did not expect to see this. I did not ask about reviews. I just asked, tell me about Hertz in Norfolk Airport. It started with “generally mixed experiences.” Then, it talks about what customers are talking about and the issues. So, this is combined with multiple sources on the Internet. So, the way they collect information is different from how Google does that, and it combines it into a response like this. And as a brand, as Hertz in this case, I have no visibility. I don't know the type of response that I would emphasize at the moment, but as a brand, what I know and think about now is I need to make sure that my content across multiple sources that can be used for generative search represents my brand. If there are any issues, I need to be able to fix them. I think the PR element will be incredibly [more] significant than this, but again, we can talk about this later.  

I then did a similar search on Perplexity. So, Perplexity is changing. From what I know, they have been using OpenAI models, but it looks like they are shifting to another model now. The response here also tells me that the customer experiences have been mixed. This is the same or very similar to what I saw in ChatGPT. 

So, the approach is similar. They look at multiple sources, and they include the sources here (points at links). And then you as a business can still influence this. But it is very different from the controlled SEO environment that we are all used to. This becomes a truth serum for brands, companies, and organizations, and something where the focus shifts from controlled owned channels to ensuring that your organic content is present in the broader data ecosystem for this AI-driven discovery. This will likely change as everything is changing. But that's where we are right now.

Ads in Generative Search 

And yes, this is very important. When we talk about search, we at least know, okay, these are the links that we are getting. We are position 3, position 5. And this is what people see. In Generative Search, no one knows what you will see because the essence of being generative is because it generates, and every time, it is different. So, I mean, the essence and the context will be the same, but the exact words will be different. 

Also, if we think that, okay, so it generates something, but we can maybe influence it with ads?t's probably not as direct and straightforward as it used to be. OpenAI said that initially, they don't plan to have ads in search. We'll see if that changes and what their commercial model will be. 

Perplexity has just launched an ad program. It is a mix of traditional search ads that allow users to ask questions. But those answers are not specific to the AI model, like in our Generative Response Ads. They are what Perplexity knows about that brand or the business, and Google Ads will no doubt evolve to compete effectively

Building Effective Content Strategies

All of this means that there is so much content work ahead. If you want your product brand organization to be discovered, you need someone whose job is basically almost full-time to start planning and thinking about what it could be and how to approach this.

There are three key pillars around which it makes sense to build content programs. One is that you need to understand your visitors, users, customers, and intent. And it's all doable. It just requires time and effort. So, the first is intent, the other is context, and the other is relevance. Let's talk about each.

Understanding Intent

Intent. And you know there is a reason we are called Intentful. It's understanding the type of customer intent. What is it that they are usually interested in? Is it informational? Is it navigational? Are they ready to buy? Are they comparing? Are they trying to solve some problem? Or they're just searching for inspiration. When you think about your brand or your product in this context, most likely, you can relatively quickly create a spreadsheet where you would put each of these types of intents and then describe for picturing different types of target audiences that you have, what are those intents that they have, and that becomes the sort layer of the foundation for your content plan.

Ensuring Relevance

The second one is relevance. There are multiple things here. One is that it needs to be up-to-date and reflective of what's happening now. As an example, Summer destinations are already outdated because summer is over by now. The content needs to be up to date because otherwise, it will not be served.

And then exclude all the unnecessary fluff. And I know it sounds obvious, but, oh, my God! Sometimes, when we look at the copy, what is this? What did they really want to say with this? It needs to be more factual and more informational. And this is number one in terms of that I really suggest you focus on. It needs to be informational because when that chunking process happens, and when AI does on how close the answer is to what the query is about, it will be more about the fact. It will be more about the information than the beautifully worded phrase because it may just not be clear enough for the AI to understand. 

They [AI search engines] will adjust that information into a beautiful copy adjusted to the style of the user. But when they find the information, they need facts. It needs to be accessible to the crawler. You just keep it open [to the crawler], have a schema, and make it accessible. 

The content or information that you have needs to be easy for AI to understand. We just talked about this; I cannot emphasize this enough: forget about keywords. It's not about keywords. It's a different way of searching. It needs to be about topics and about context. It should have structure because for the AI to understand, it  [a technology] is getting much, much better. I think that AI understands me much better than most people or many people, but still, I think it's just finding a way for AI to understand what you want to tell it. So, there needs to be structure.

Establishing Expertise and Authority

Expertise. I cannot emphasize this enough. Remember how Google had, as they still have it, E-A-T - expertise, authority, trust - this is still true and relevant, evolved into E-E-A-T (Experience, Expertise, Authoritativeness, and Trustworthiness). If you can include the source of information if you're writing an article, let's say, going back to travel and destinations, if you are writing an article about an attraction, include that it was written by a tour guide or a museum guide, or so that adds value, and that adds expertise, authority, and trust when AI decides whether to include it or not. If this content is published somewhere other than a reputable website and has no relevance or expertise, it is likely it won't be included.

And that goes to the source authority. That is similar to the case with Google, which will continue with AI engines - fact-checking signals. If there are ways you can include and reference that the information that you showed is factually correct, this is an additional benefit. This is similar to the backlinks. Back many years ago, the idea of backlinks was that the more backlinks refer to your source of information, the more it appears that you are a relevant source of truth.

This is now changing. And it's more about the facts. So, if you can find ways to include official websites or anything that adds that additional trust and value, that would be helpful. We talked about the time, so it needs to be today and not 2020 or earlier. That's also the case with many websites because a lot of information goes back to 10 years ago and just needs to be updated. And you don't want that information to be included in the AI-generated responses. 

Content Consistency and Multi-Format Integration

The last one is content consistency. So once you have that content plan, a skeleton based on intent, context, and relevance, make sure that it is consistent across the channels where you will be distributing it.

In terms of the types of content, text is still super important. We're not going into how LLMs work, but text is a major player. So, text content is back, and images [as well]. AI can now read images easily, the same as video and the same as audio. So, all of the types of content you have are good for AI ingestion, and then they will get shaped into AI-powered answers. 

Understanding the Shift from Traditional to AI-Driven SEO

I also asked ChatGPT—of course, I did it—to summarize the difference between old and new SEO. I agree with most of this, except for the optimization. It says that it's manual. I know that a lot of companies did not really do it manually. There were quite significant automated programs in place. But pretty much everything else is accurate.

But back to when we started about the zero-click. Now, with AI searches, it all gets a new meaning. Clicks just don't happen because a lot of answers will be received, whether it's Google or Perplexity or ChatGPT, or we'll have other players. But that information will be happening without getting to your website. It doesn't mean, though, that your traffic is going down, or it doesn't mean that your content or information is not discovered by your visitors and users. It's just adopting this new mentality that discovery and distribution are getting completely redefined.

Preparing for the Future of Content Discovery

And we are just at the very beginning of this. I think next year will be different. I think it requires a lot of effort for this to sink in, but don't ignore content distribution because you need to be the source of truth, and you need to have your content, your information, all of those things that we just talked about intent, relevance, timelines. To be across not just your website but across another partner, publisher or other reputable sources because you cannot just rely on the fact that it is content from your website that AI will use to provide the answer. As we saw with Norfolk's example, the car rental example, information was really pulled from multiple sources.

That's pretty much what I wanted to cover today. If you have any questions or want to connect on LinkedIn, I am always happy to. I'd like to return to the core steps that you need to start doing now. Start building content programs to be ready for the new era of discovery. It is completely different. We can be in control, but this requires way more effort than the traditional search.