Optimizing for AI-Driven Search

CSP Founder and CEO Chris Rodgers shares a presentation discussing AI-driven search including SEO and GEO. Watch the video below or visit the CSP YouTube Channel for more.

Below is a written copy of the video transcript:

Happy to be here. My name is Chris Rodgers. I am the founder and CEO of CSP.

CSP is a specialized SEO agency serving regional, national, and enterprise organizations. We provide our clients with revenue-driven and business-aligned SEO solutions. I have a small but mighty team of 14 full-time employees with over 100 years of SEO experience.

You can follow me on Twitter. My handle is @SEOdub. I have my LinkedIn link up there as well.

And if you want to visit our website and learn more about my company, it is csp.agency.

I have been working in the SEO space for just about 20 years. I’ve worked as an analyst, a strategist. I have overseen and built teams and really filled all the different roles within an SEO agency.

Today, my responsibilities are really leading my team, making sure they have the resources they need, assisting in strategy and some analytics and things at times when needed.

And a big part of my role is establishing the mission and the vision for CSP. Part of that is very much looking forward at what is happening in the SEO space. And that also crosses over into what is happening with AI, specifically AI when it comes to search engines. Now with large language models and generative AI.

So there’s quite a bit of of reading and learning that I have to do to stay the curve. So I’ve prepared a presentation for you today on optimizing for AI-driven search. So hopefully you find this interesting and also find this valuable as well. So today we’re going to talk about generative AI.

AI chatbots versus search engines.

Oh, one last thing. We are going to try to do questions at the end.

I’ve got a lot to cover, so it will be a bit tight, but please hold questions towards the end, and we’ll try to get in those at the end of the presentation.

We will cover ChatGPT and search, AI-driven search engines, what is GEO and a recommended approach, geo research and ranking factors, Google’s E-E-A-T and CSP’s M-E-E-E-A-T framework, and the three pillars of PR influence as it relates to GEO.

So first of all, what is generative AI?

Most of you probably know what this is, but it is a type of machine learning that uses AI to create new content. That content is multimodal. It is in all different forms, so that can be text, it can be audio, it can be video, it can be images. And ultimately, this generative AI is trained on various data sets. It looks for patterns and then creates new content based on those patterns.

So let’s talk about AI chatbots and text responses. A lot of what we look at in SEO, much of it is multimedia. There’s a large focus on text and text generation. You can see the image that I have on this slide.

We’ve got our three main chatbots, OpenAI’s ChatGPT, Bing’s Copilot, and Google’s Gemini. I created this with Copilot, with DALL-E, so this is generative AI.

And it’s representative of, I think, where generative AI is at in that I could give it a lot of descriptions. I could make this animated and give it direction. And it came up with this original image. But despite that ability, it could not get OpenAI right. So we’ve got a, I think a miniature O or something in there in the image. 

So AI chatbots leverage LLMs or large language models to provide natural-sounding responses to questions. So this type of AI does not know anything.

It doesn’t know facts, it doesn’t know truth. It essentially once it’s trained on these large data sets. It is guessing the next word in a sentence using probability based on the data set it was trained on.

It does it really well, it comes across really convincingly, but the process that it’s using is, while it’s quite advanced, it’s also somewhat simple.

So, let’s talk about search engines for a moment. So search engines have long been supported by AI. We have another image here that was generated by DALL-E. This one’s a little bit better.

So with search engines, specifically Google, they have had a number of AI systems integrated into their algorithms for quite some time. They use natural language processing, NLP, as well as other machine learning systems to really understand queries, to understand the intent of searchers, what they’re trying to achieve, whether they’re trying to know something, go somewhere, buy something, et cetera.

And Google has access to the Knowledge Graph. And the Knowledge Graph is this large database that has all of these facts and entities within it, right? So think people, places, and things. And it understands the connections between those people, places, and things.

So the Knowledge Graph will understand that CSP is an SEO agency, and the founder and CEO is Chris Rodgers. And Chris Rodgers did his undergrad at UIC, a master’s program at Full Sail University, et cetera. 

So, you know, in terms of the relationship between things and how things are linked, it’s it’s fairly sophisticated. I would argue that in some ways, Google’s traditional search engine with AI is more advanced than what’s happening with these chatbots and large language models.

So then we have the culmination or the combination of now chatbots with search, right?

And we’ll start with ChatGPT because, frankly, OpenAI is really what started this big movement recently in terms of these changes. So you know, I would imagine most people on the presentation have used ChatGPT before. Some of the main limitations, right, we’ve got inaccuracies that are built into that. We’ve had a fixed data.

In terms of time frame, some of the newer models on ChatGPT actually are updated data from April 2023, and now ChatGPT is integrated into Bing has the capabilities to use Bing search.

So you’ll notice if you asked it certain types of questions that are related to, you know, timely things like news or weather, it’ll actually show that it’s searching within Bing.

You can hack this a little bit by putting in the words in 2024, and that’ll typically bring up, it’ll search in Bing’s database. So the, I think the idea here is that we can combine, you know, search and these chatbots together for a more enhanced experience.

To be clear, ChatGPT is not a search engine.

Open AI has spoken about creating a search engine, but they haven’t released one yet. And I think it was really smart with what we’ve seen, you know, kind of happened with Google and in some other areas. I imagine one will be coming at some point.

When we look at AI-driven search engines, we’ve got our legacy search engines like Google and Bing. Google created Bard which became Gemini.

And Gemini is now a standalone chatbot that you can actually engage with, like you do ChatGPT. And you can also engage with Gemini through their AI overviews.

And you probably started to notice that, especially if you’re logged in into an account in Google or if you’re logged into the Google Labs, you’ll see those AI overviews a little bit more.

And Bing, having their relationship with OpenAI is now integrated with Bing’s copilot or their chatbot.

Same thing, you can interact directly with the chatbot, or you will start seeing you know that same type of AI responses within Bing search as well.

When we look at Perplexity or you.com, these are AI search engines that were built from the ground up with a different experience.

I’m a fan of Perplexity. I like the way that they’re doing things, and then Brave is also integrated with AI search. And I imagine you know there’s others as well.

Here’s an example of an AI overview in Google. How should I select a PR firm for my company?

So this was Gemini that created this text response. And, you can see some citations down there at the bottom.

So it’s its citing, not necessarily the sources where it got this information, but it is backing up that response with some included sources that you can visit. You will notice the bottom line there, generative AI is experimental.

So when you get false information or inaccurate information, you know they’ve kind of you know covered themselves with a disclaimer.

Here’s an example of a search result with perplexity. Similar, right?

But you do see the sources are listed at the top, but a similar experience in terms of the interface. All right, so let’s talk about real briefly retrieval augmented generation.

So this is the process that AI search engines largely are using, and it is where you’re taking a large language model and connecting it to an external base of knowledge to improve the accuracy or quality of responses.

Again, the idea is that it is the best of both worlds, right?

You can be interacting with a chatbot and get these direct responses, but supposedly it should be based on, you know, more accurate search results. The steps that you see over here on the left are actually from the Google search generative experience patent docs. So we believe this is how Gemini is working today with AI overviews.

You know Disclaimer, we don’t know that for sure. They haven’t verified it, but we have every reason to believe this is the process.

So you went to query into Google. Google then takes that query and says, Hey, you’re looking for this. We know that people that look for this also have searched these five things and found it really useful. So we’re also going to search these other five things. They go out and complete all those searches.

They have all of the indexed pages, so all of the web pages, and now they have their smaller, more focused data set that really should be more qualified and more trustworthy because that is being based.

It is using the capabilities that Google has to validate web sources, right? Should be more accurate. Then they create that summary that you see, they create that summary, and they do something interesting. They actually take a snippet out of that summary, and then they go perform another search with that snippet. And then from those results, it’s going and picking its citations.

It seems to me a little convoluted as to why they’re they’re doing these extra steps on the end. 

I don’t know if that’s trying to prevent gaming of the system or if there’s other rationale for it. It may be for just diversity of information resources, so it’s not just coming from the top-ranking websites. But that’s how we’re, you know that’s how things are working today.

Okay, so let’s talk about GEO.

GEO stands for generative engine optimization. it is the marketing practice of optimizing websites to appear more prominently within AI-driven search. That is my definition, so you’ll get a different definition if you go and search.

I am, you know, personally trying to push the marketing aspect of this so that, you know, we don’t have a bunch of people just spamming tactics like we have had in SEO, but I’m sure we will. Anyway, some important caveats before we jump into this. It is extremely early. This stuff is changing very, very quickly.

So by and large, we recommend an approach where you’re aligning with current SEO and marketing practices. So no big hard right or left turns, because if things change, you may end up in no man’s land. The techniques are likely to change in terms of how we optimize for this. And there are no established best practices. It just has not been out long enough.

We do have some data, there has been some testing, there have been some studies. We’ve seen that there are some things that we can do that work. And we have a good enough idea that we can start you know taking steps and executing some strategies, but we have to be mindful of of how things are changing.

And the last thing is that large language models have big challenges with accuracy. They are just not set up to understand what’s true or what’s not true. It is purely a language technology. So I think inherently there are some big challenges with getting things accurate.

You know, I do expect that it’s gonna get better, but for all we know, there’s a whole nother type of AI that’s gonna be coming that we haven’t even seen that’s gonna serve the same purpose and make LLMs obsolete.

Okay, so a recommended approach.

SEO, at least modern SEO and GEOare very much aligned. And I say modern SEO because modern SEO to me is more in line with marketing. It is looking at audiences, personas, creating content with a purpose to reach people at at a stage of the buyer’s journey, right? This is to me, modern SEO marketing.

It is not your SEO checklists and scores and updating your, heading tags, those types of things. 

Those are still important to do because you know, those practices are still going to be important to populate these search engines.

But when it comes to GEO, those are not the things that are going to drive visibility in generative engines. I would avoid, you know, technical hacks like there are people that are already hacking open the AI overviews, there are ways to do it. But those are not good long-term strategies, right?

Content originality and quality is your content easily replicated. This is a really good approach, right? Good content is not going to go out of style, there is going to be a place for it. Google will find a place for it on their search engine or they won’t get it and it will go someplace else and we will be optimizing someplace else.

Okay, so I did want to first cover optimizing for ChatGPT. There was an industry study by NP Digital, Neil Patel’s agency, where they looked at 82 different ranking factors for ChatGPT, and there were some significant ones.

So the number one was relevancy, right? So this is how we build websites and web pages. If you’re creating you know a content for PR, how closely does that content match what people were specifically looking for?

So think about that depth within each individual topic, what you’re covering. We’ve got brand mentions is huge. The PR tie-in there is obvious, right? We have outside reviews. We have the age of the domain or the brand or the business. We have recommendations and we have authority.

One important thing to mention is that almost 30% of the answers in this study came back as inaccurate or off, right? They did not present a real solution. That is why today ChatGPT is really not a great source for for working as a search engine, and that’s not its strength.

But there’s a lot of usage on ChatGPT. There’s a lot of, you know, early adopters in terms of the larger market. So it does make sense to pay attention to this and, you know, make sure that you’re doing things to be visible.

Okay, so some of the strategy and tactics.

Keyword research still applies. You should not be doing keyword stuffing, right? This does not work the same way it does with SEO. Strategically placing keywords and using keywords in the context of your audience and a top makes complete sense and you should be doing that research. When you look at content for like businesses, laying out their solutions really clearly, right?

So what is it that you do? Who do you serve? Are there certain industries or sectors that you serve? All that needs to be built out on the website.

And then things like case studies and testimonials, reviews, all things, all those traditional, you know, confidence factors, those are gonna be important on your website to show ChatGPT because it doesn’t have the same capabilities that like a search engine does where they can look at all these other factors, right?

ChatGPT is very much looking at your site and they’re looking at other sites that they find that might speak, might talk about you. When we look at off-page or off of your website, those brand mentions are obviously huge. So anything around PR is gonna be impactful.

And then reviews and recommendations, so review sites or different roundups, those kinds of things that are referencing your brand. Still important, and it’s going to be important that your site is able to be crawled by search engines as part of this whole process with AI search.

So technical is going to be important and using schema, we expect, is going to be very important as well. And it’s a good time to be paying attention to Bing and optimizing for Bing, right?

This is the index that ChatGPT is accessing. That relationship is very strong, so, uh you know, it’s a good time to pay attention to Bing. 

Okay, let’s talk about the research that we have around GEO.  There was a research paper titled GEO: Generative Engine Optimization. It was authored in late 2023 by researchers from Princeton, Georgia Tech, the Allen Institute of AI and IIT in New Delhi.

In this research paper, they claimed that they were able to drive a 40% lift in visibility from the techniques that they explored. I do have the link there in the bottom. We will share the slides.

You obviously won’t have time to read this whole abstract, but we’ll hit the high point. So these are the factors that they looked at in their research study.

So authoritative content, they’re really talking about language. And this is in the context of optimizing a single page on a website. So using authoritative language, within the content keyword placement, we’ll talk about that. But again, this is not old school, you know, keyword stuffing.

We have statistics, including statistics in your content. Including citations, actually citing your sources was another thing they tested. Including quotes, right, from subject matter experts, other authoritative people. Increasing the understandability, right? So actually keeping the language simple and conversational.

Fluency optimization, right? Making everything flow, be clear and engaging, avoiding complex sentences. Using unique words, right, that may make things more interesting. And then the use of technical terms, which obviously may be a juxtaposition with understandability and fluency, but again, they looked at those things.

So of the factors that they tested, the top three were citing sources, adding quotes, and then including statistics. So those are the three leading factors.

The next two were fluency optimization and making language easy to understand. And then also authoritative, making things more authoritative, that that gave a lift as well.

Keyword stuffing had a negative influence. So you gotta balance a line when it comes to keywords, right? Even today, you should be using keywords and SEO as clues to help Google align audiences for a particular piece of content, right? Or even for an entire website. Keyword stuffing is going to send you the other way and really hurt things.

Here’s some examples, right? So literally, this is from the research study. You know, it is including sources, statistics. There’s examples of how you would make, you know, language more authoritative. Combining geostrategies yielded the best results.

So, you know try to do a lot of these things in tandem if you can, and it makes sense. We obviously don’t wanna, what I would call spam. We don’t wanna be just making up statistics and doing quotes and my fears, that’s where this goes. And you’ve got people just kinda making stuff up and trying to put it in there and putting out a bunch of false things.

So hopefully we can actually do this the right way.

Okay. I’m going to shift gears a little bit here and talk about E-E-A-T. 

E-E-A-T is a content framework that Google provided a little while back. It stands for experience, expertise, authoritativeness, and trustworthiness, right? And Google tries to look at a website or a piece of content, understand these factors.

Experience, if we use the example of like an attorney, how long have I been in practice, right? How many cases have I actually handled?

When we look at expertise, am I a general attorney or do I have specialization? Am I a personal injury attorney? Am I a personal injury attorney that actually specializes in truck accidents, right?

We have authoritativeness, right? Is my law firm well-known? Are people online looking for my law firm? Are they looking for my brand? Who are the other people that are on that are practicing in my law firm? Are those notable people? Do they have LinkedIn profiles that show all of their experience?

And then lastly, trustworthiness. Am I, you know, associated with trustworthy organizations? Have I gotten, you know, news coverage from reputable media outlets? So this this is a content framework that very much we believe is going to be a big part of GEO as well.

As we move forward, Google is not going away. They are going to be a part of the AI search experience moving forward. And we know that these are important factors for Google at CSP.

We took this a step further and actually Jesse Farley, our VP of Growth, came up with a concept called M-E-E-E-A-T, and it really is just adding a few factors and building on E-E-A-T.

So adding in was multimedia and engagement.

So when we look at the way that these AI search engines are working, they are multimodal. So it is going across video, it is looking at audio, it’s looking visually, text, all of those things. So that is the world that we need to play in.

It is also a time when we need to elevate our own content. We’ve got AI-generated text generators everywhere that can create that content. It is much more difficult for them to create and hold a piece of content that has video that also has a podcast or audio that has unique visuals or infographics.

And when we do these things, we also increase engagement. This is a time when we need to increase engagement on our actual web pages, right? At a time when Google is trying to have all the answers and keep everyone on Google, this is a time when we want to have reasons to attract people and keep them on our websites.

So we can achieve engagement by really becoming more diversified with our content and with multimedia.

And then lastly, we have the three pillars of PR influence.

So these are not all related to optimizing content, but it is important in terms of the influence and relevance that PR has. When it comes to you know AI search and generative engine optimization. So certainly brand building, brand exposure, the ability to help brands establish trust and authority.

And there’s direct way that this happens in PR in terms of establishing subject matter expertise, right? The depth, the focus and detail provided within a singular topic, the ability to go in deep and really share expertise, and then helping make meaningful connections with an audience, right?

Targeting a specific audience with valuable information, things like topic alignment, unique content, easy to understand and highly fluent. So if we do all those things, I think we’re going to be very much in line with where things are going with generative engine optimization.

So we’ve got, we actually did really good here.

I think we’ve got about five minutes.

So I am going to go over to the chat and see if we have any questions.

I’m not seeing anything at this point, so no questions from the group.

Q: In what ways does gen-AI harm SEO?

A: You know The biggest threat right now, I think, is to content creators and top of funnel content creators. For SEO specifically, SEO is going to have to continue adapting. It’s always adapted, but very much Google is trying to keep people on Google and within their paid ecosystem, right?

Google’s biggest revenue generation is from their ads. So if they can keep people on Google circling around and and clicking on ads, then that’s going to be better for Google. But no, for SEO, we’re just going to have to keep adapting.

I will tell you that the spammy SEO guys, the guys that are doing checklists and scores and are just like really process driven, they’re probably going to be in trouble because there’s there is a lot that’s going to have to happen here on the marketing front, creating good content that’s well researched. That goes far beyond just, you know, SEO data.

Q: How do these tools best align with the other PR tools in the box? (You mean like how do the SEO tools align with PR tools?)

A: Unfortunately, I’m not, I am not familiar with maybe the PR tools you’re referencing. Well, SEO AI is just one more tool. Oh, I see what you’re saying. So yeah, how do these tools best align with, I mean, I think being able to just adapting the PR space to what’s happening within AI search and staying on top of this.

I do think that it can be aligned, right? When you look at this particular slide, this is this is how these things align, right? It is the way that you’re approaching content.

It is how you’re using your ability to influence how brands develop and how they you know make themselves trustworthy and authoritative. These are gonna be real challenges that businesses have. specifically when it comes to SEO or GEO.

You know We’ve got a handful of citations that are at the top of search results now, as opposed to the 10 links. So there’s expectations that there may be more traffic going to fewer people.

So it’s going to be more important that brands establish themselves than maybe they have been able to before. There’s going to be challenges with some of those smaller publishers and smaller businesses.

Q: What sources do you read?

A: You know, I follow Search Engine Land and Search Engine Journal. I follow Rusty Brick. Let’s see if there’s any other ones. Honestly, a lot of it I find on Twitter and random LinkedIn, just people I’m connected with. 

Q: With uniqueness, quality, facts, so important to content, how will AI-generated content be ranked by AI search?

A: It’s really interesting because we just went and looked at some of those factors and some of those factors were fluency. Right? And readability. So you could argue that AI could be used to actually improve the readability and the fluency of content. So there’s there’s a place for it.

But you’re you’re going to have to find ways to create truly unique, engaging content. And that’s not going to happen by being automated with ChatGPT or with Gemini. You can use it, right, you can use it in the process. But that’s going to be commoditized content. Everyone’s going to be doing it.

And if you’ve used ChatGPT enough, you’ll see that in certain subjects, it starts saying the same thing in different ways over and over again. So you’re only going to be able to get so far with the automated AI content using AI as we create content in original and creative ways, I think is is where it’s at.

Or hey, go all human and actually spend the time and you’ve got a chance at making something that’s truly unique that AI is not going to replicate.

Guys, thank you so much.

Q: Will it be dinged for rank?

A: If you’re just automated, yeah, if you’re just automating massive amounts of content, there was a big update here in March that rocked a lot of those websites that were mass-producing AI content. It may still work in some circles, but they’re trying to stop it.

I think we’re over time, so I just wanted to say thank you to everyone for attending.

I hope this was informative and good information.

If I can help in any way, feel free to reach out to me or my company.

And yeah, you guys have a great day.