The State of AI in SEO – 2025 Edition

Join us for our upcoming webinar, “The State of AI in SEO,” where we explore how technology is reshaping digital strategy. We’ll cover everything from the latest AI algorithms influencing search rankings to the significant impact on content creation, sharing practical insights to help you stay ahead in the world of SEO.

Joining our host David Bain for this live podcast episode will be Nitin Manchanda, Pam Aungst Cronin, Katarina Dahlin, Tejaswi Suresh and Garrett Sussman.

Watch On-Demand

Listen to the Podcast

Transcript

David Bain 

The state of AI in SEO 2025 edition. Hello and welcome to the February 2025, edition of the Majestic SEO panel, where we’re discussing the state of AI in SEO. I’m your host, David Bain, and joining me today are five great guests, so let’s find out who they are. I’ll start off with Garrett.

Garrett Sussman 

Hey everyone, I’m Garrett Sussman, the Director of Marketing at agency iPullRank. We offer SEO content strategy and generative AI services and been kind of neck-deep in all this AI Gobbledegook. So thanks for having me today David.

David Bain 

Also with us today is Pam.

Pam Aungst Cronin  

Hi everyone. I’m Pam Aungst Cronin of Pam Ann Marketing and Stealth Search and Analytics and coming this year, Pam Ann AI as well. First time I got to introduce that. So super excited about being able to offer AI services now too.

David Bain 

Intriguing – thank you so much for joining us. Also with us today is TJ.

Tejaswi Naidu 

Hey David, thanks for having me. Yes, I’m TJ. I Head of SEO at Mayflower. We do all things next generation entertainment and tons of AI related topics. I’m super excited to share a couple of insights from my learning so far.

David Bain 

We’ll be super excited to hear that as well. And so also with us today is Katarina.

Katarina Dahlin 

Hello, my name is Katarina Dahlin, I’m a senior growth hacker at Genero, a leading growth marketing agency in the Nordics. Very excited to be here.

David Bain 

Very excited to have you here as well. So first time here, hopefully we won’t put you off, and you’ll be excited and pleased to join future panels. Also with us today is Nitin.

Nitin Manchanda 

Thank you, David, super excited to be here. Looking forward to this conversation. Thanks for having me. I’m Nitin, I have been practicing SEO for last 12 years, and have been spending a lot of time on AI these days, like everyone else. And yes, looking forward to this conversation.

David Bain 

So let’s start off by asking Garret the first question. In fact, we’ll ask everyone this first question, but we’ll get Garrett to begin. And that is, how would you summarize the state of SEO in 2025.

Garrett Sussman 

There continues to be a lot of noise and very little signal. Or, I should even say, a lot of signal. It’s very overwhelming. Still, there’s a lot of changes every day. We’re getting new models. We’re getting changes in search behavior. I think there’s a lot of volatility. It’s never been more exciting. There’s a lot of tools at our disposal.

However, there’s also a lot of misinformation out there, and so I think as SEOs and marketers, we’re doing our best to keep our heads above water, to leverage this tech and not be, not succumb to just the overwhelming nature.

David Bain 

I always look at the panel faces to see who’s nodding or shaking their heads and Nitin you were doing a bit of nodding. And when it came to, when it came to Garrett talking about the volume of tools and options out there. At the moment, how would you summarize the state of AI in SEO?

Nitin Manchanda 

Oh, yes, I 100% agree with everything that Garrett mentioned. I think since I started SEO, it was always changing. And we were hearing like, you know, this year is probably, you know, where we have the most volatility. And now, if you talk about last couple of yours. Everything has been changing every, every, everything, right? And if you talk about SERP as well, right? So I was working with brands where, ranking one, they were getting 16% CTR, and now at the same position, they’re getting 5% and they’re happy about it, right?

There’s a lot of things changing, and we as users, our search behavior has been changing a lot as well, right? We are not looking for that chain of, you know, keyword searches, or, you know, like, search queries to find what we’re looking for. We just want one prompt and a perfect answer. So a lot has been changing, and I think it will keep changing, like I mentioned. And I’m super excited about what is happening and how it’s evolving.

David Bain 

Katarina, how would you summarize the state of AI and SEO?

Katarina Dahlin 

I would say that, well, if we last year have everyone has at least experimented with ChatGPT and maybe found their own ways of working, but this year, I think we will move more into automating things, and taking those things that we are doing in each ChatGPT or another tool and just trying to automate as much as possible.

David Bain 

That’s the key, isn’t it, knowing what to automate, knowing what to actually add value to and use your human hands for.

Katarina Dahlin 

I think everyone is just trying to find their find their own ways of working and things.

David Bain 

Pam, how would you summarize the set of AI in SEO?

Pam Aungst Cronin 

Well what I’ve been thinking a lot about is zooming out to kind of the 30,000 foot view, and just looking at the big picture of how traditional search engines keep adding more and more AI. So like Google adding AI overviews to the top the search results. And if we could say traditional, they haven’t been around that long, but traditional AI tools like ChatGPT are now adding in web search functionality so and then. Now there’s also AI search engines coming out, like Perplexity. So basically, like, with search integrating so much AI and AI integrating so much search, it’s just all morphing into one new thing, like aI powered search. People are like, how? I mean, even, even the title of this, like, how is AI affecting SEO? AI and SEO, I don’t think for very much longer, we’re going to be talking about them as two independent things. I think it’s going to be more like SEO is now AI or AI now searches the web like it’s just one thing.

David Bain 

It’s incredible how fast it’s evolving and how quickly SEOs need to evolve with that. And I guess it’s always the question, what is SEO? And is SEO still the relevant term for what SEOs do as well? TJ, just before we get onto specific topics, how would you summarize the state of AI and SEO?

Tejaswi Naidu 

Well, it’s been a year since I last spoke with you about the State of AI, and back then, it was all about, you know, using traditional models that OpenAI gave us, which was the 3.5 back in the day. And now you have for all, and you also have, like, the latest model, called the o1 and now there’s also the o3 which is the reasoning model. I feel like this is kind of groundbreaking, because it allows the AI agent to take its own time to, you know, understand implications, before it spits out an answer.

I think this is definitely going to be groundbreaking for us from an SEO point of view, because when we bring in the aspect of AI agents into the picture, you can, you know, fine tune a lot of AI to be tailor focused for your SEO work. So let’s talk about, you know, product descriptions for E commerce sites, right? The traditional way is just ask for ChatGPT or whatever, to spit out an answer. But now you can actually give it a list of constraints. You can give it a list of your own data for it to learn, and then it can learn and reason you know within itself before it gives you an answer, which makes it a lot more smarter, a lot more intuitive in terms of the output that you get.

So I feel like the true power of AI has gotten significantly, a lot more better in terms of overall output. And the true power, if I may, lies in how you prompt. So I feel like the state of AI in 2025 and 2026 going forward, comes, or boils down, to how you structure your prompts. How do you structure your thinking process in laying out how you want to execute AI around you. You can automate it. You can use it for content. You can use it for log analysis. There’s, there’s plenty of use cases that we can, you know, use AI for.

I feel like 2025 and 2026 has significantly improved versus 2025 and I feel like this is just going to, like, accelerate and get a lot more awesome, if I may, in the in the coming months and years.

David Bain 

Katarina, you were talking a lot about automation, and the key is to know what to automate. So how do you know what to automate, and what tools do you use to assist with that automation?

Katarina Dahlin 

So I have an e-commerce site, and I basically started to automate the optimization of product descriptions with Make, because that is annoying to do manually and you just need help of AI. So yeah. But also what TJ was saying that the prompting is the most important one.

David Bain 

That’s make.com you’re referring to. So what is that, if anyone hasn’t used it, and how do people best use that?

Katarina Dahlin 

It’s kind of a tool where you can connect different applications to each other without having to know any any code. So it’s a no-code platform. So basically I have all my product data that I would need to optimize, or like, yeah, to optimize with ChatGPT. I have that in a spreadsheet, and then with make, I connect ChatGPT to the Google Sheet, and, yeah, build a workflow so then it will just read the data in the spreadsheet, and the ChatGPT will write an answer and just put it back in the spreadsheet. And yeah, but you can also, like, connect other, basically, almost anything, like Shopify or WordPress or Search Console.

David Bain 

So it sounds like Zapier, but with AI and web development in there as well.

Katarina Dahlin 

Yeah, you can connect many different apps. I have connected and used the ChatGPT.

David Bain 

So how do the prompts work? Then with the ChatGPT, do you still have to actually make the prompts within ChatGPT itself? Or is because the fact that it’s connected to another app using make.com It means that it’s getting the prompt prompts directly from the other app somehow?

Katarina Dahlin 

It basically works like the same way that it’s easier if you first go to the ChatGPT and just build your prompt and just test it there and see how it works and when it works. You can build the same thing, but you do it straight in make so that you can connect, connect it to its spreadsheet or directly to the website if you want, but it’s basic, basically the same. It’s just in the in the makeup.

David Bain 

And TJ, you were talking about prompts as well. You’re talking about the importance of getting prompts right. So what? What is good prompt practice in 2025?

Tejaswi Naidu 

I feel like in order to get the best possible output from AI, you have to be a subject matter expert on that topic which you’re working on, because AI, by itself, is not smart by definition. It can only be smart powered by somebody that’s, you know, pushing the right prompts to it so it can give you the right information.

If you come from a point of not having enough knowledge, but asking AI to come up with advanced topics, there’s plenty of opportunities for it to hallucinate, break down, give a lot of misinformation, which you can which you then treat it as potentially true information, and then you just have a mess in your hands, right? So I feel like in order to get the best possible output when it comes to prompting, you also should define how you want to like, lay out the grand the grand vision, so to speak, when you want to build out any topic, when you want to use AI.

Just to give you an example, right? So I wanted to understand how I could use AI for my keyword research, for defining a content brief and producing content. If I came from a point of view as a newbie SEO, there’s plenty of opportunities for AI to define how it should potentially be done, which cannot be as effective as possible. But because we have enough experience among ourselves in this finalist, we can, you know, clearly define what, what the constraint should be, what should you know the sequential flow of prompting should be so that you get the best possible output that you can from AI.

So it’s really important when we use AI, that the person using it knows what prompts to write up, so that you get the best possible output. So I feel like anybody who, who you know is the prompt master, will definitely win the AI game for sure.

David Bain 

So to get the best possible outcome, as you said, there is it necessary to use a custom GPT, or can you do that, as long as if you prompt well using a fairly generic tool, like OpenAI?

Tejaswi Naidu 

Yeah, it really depends upon the case, right? So if you want to create like a persona to chat with you, right? You can use a relatively simple LLM to like, define how the chat should be. Like with a, you know, with the AI person, mimicking to be a personality.

But if you want to use LLM for let’s say how you should pull in the stock market data in real time and then put it onto a chart.js to visualize it, and then make sure that you get an alert on slack when it anytime there’s an anomaly in terms of like Nvidia stock performance, for example.

So in that scenario, you definitely need to have the best possible prompting with an expert in finance, for example, along with you, so that you can make sure that the level of output you get is quality in nature.

David Bain 

So I know that you want to share a lot about AI agents as well. TJ, but just before we get to that, maybe let’s go to Nitin first. Nitin, I know that you think there’s been a bit of a change in search behavior because of LLM so how would you summarize that?

Nitin Manchanda 

Oh, yeah, massive change. I think if I talk about, you know, my personal example as well, maybe three years ago, when I was searching on Google, I had a different search style, but now it’s like, I’m not thinking about two or three word search, especially when I’m looking for something which is more on informational intent, right? My prompts are rather big.

If I’m planning a trip to Paris, I would just go there. I would say, planning a trip to Paris with my family. I have two kids, eight and four. Right? Plan to cover things which are, you know, must do. Plus, I want to keep it like, you know, very family friendly as well. So want to also do explore some nature and all of that. So when I do this right, the itinerary I get is very close to what I would actually want to do, right.

So there not just this. I would probably add something more, like, also give me recommendations about where exactly I should be staying. I get everything that I want right. So it is working as my personal assistant, more than just a search engine, or, yeah, you call it like, LLM search engine, whatever you call it right? But it is now behaving more of, you know, more of my personal assistant than anything else. So that is the biggest change I’m talking about.

We have been working with a lot of brands which are getting a lot of visibility from the US, and because right now, it’s pretty aggressive in the US versus the other markets. So we see a lot of changes. Well, massive increase in zero click as well. But our clients are happy, because if, for example, they were getting 10,000 sessions before, and that was converting 200,000 US dollars as business right now, they’re getting, for example, 8,000 clicks from that cluster, right? But the conversion is even better. It’s not 100,000 maybe 120,000 because all those clicks which were just getting just landing on the landing pages and for some information and then bouncing off, now we are just cutting on that, because now it’s more qualified clicks that you’re getting on the website, because people who are looking for basic information, maybe they’re on the top of the funnel of their journey, right? They are getting all the answers there itself, and when they’re in concentration phase, or rather bottom of the funnel, right? That is when they’re landing on a website, getting more information, checking reviews, ratings, all of that, and then converting.

So this change is huge, and I also did like a quick survey to understand how people are consuming llms, right? It is consistent pattern. 99% of the cases, there are one person cases where they still believe they have questions about data privacy, all of that, right? But 99% of the users are pretty comfortable with how they’re searching, and the applications are also pretty similar.

David Bain 

The point that you came up with really interested me was LLMs and actually people perceiving it as a personal assistant. And does that mean that you think the future of search is actually morphing into being a personal assistant, and Google perhaps as well, morphing into being a personal assistant, as opposed to search engine?

Nitin Manchanda 

100% and I think the good part about this personal assistant is it’s not limited to certain, you know, vertical or expertise. If you give a right prompt, you get the right answers. So another, for example, right if I talk about Botpresso, when I started, I was the only one who was from engineering background, like I started my professional journey as software engineer, and now, four years after, I have a team of 25 and I think almost everyone can code in Python, which is crazy, and they’re writing a lot of micro crawlers to solve their problems, which is fascinating, even on weekends, you know, I get you know things that, hey, you know, we’re working on this, and we saw this this way. It’s, it’s, it’s awesome.

Something where you were spending probably 10 hours before now, it can be done in 10 minutes maximum. That is also, you know, including the time that you’re spending with ChatGPT to find a solution. So it’s fabulous the way it’s evolving. And I see it evolving further. So, yeah. Exciting times for sure.

David Bain 

Pam, we briefly talked about DeepSeek earlier on in this discussion. It obviously shocked many people with the quality of what it can actually produce. What are your initial thoughts on that? And should Western countries, if we can call we’ll call them that perhaps consider banning tools where we’re not sure where the data is going and who’s learning as a result of it.

Pam Aungst Cronin 

That’s a loaded question. Where to begin? Well, so, yeah, okay, so DeepSeek. I think the number one, let me just start with like what I want to nail home. The number one, most important thing to recognize about DeepSeek is that there’s an all AI, that there’s a stark difference between a model or an LLM and an application that you use the model or the LLM in.

So that difference, especially in the case of DeepSeek, is enormous, because if you separate the model, which is open source, they released it for free. You can download the entire thing yourself. You better have, like, you know, 50 terabytes of space available. But you can download the whole thing if you want, and run it on your own computer with, like, no internet. You can, like, totally wall it off.

Or you can use it on the DeepSeek website or the DeepSeek mobile app, that’s when you’re using it through a foreign controlled application or website is when your data is being is exposed to said owner of the application and whatever government their country is in.

So everyone’s freaking out over DeepSeek stealing data, but what most companies that I’m seeing doing so far, Perplexity, surprisingly, are doing is integrating the model, not to the application. So if you take the model and apply it to your own application, it’s from what I understand, and I’m not a software engineer, but from what I understand I’ve been researching, and maybe someone with a software engineer background could correct me from wrong. But it that the risk, then, of your data going to somewhere else, some other country, is if you download it yourself, like say, you can run it without the Internet.

So once you’re using the model as a standalone product, that risk is, I don’t want to say gone, because, again, disclaimer, I’m not software engineer, but that’s, I think, the most important thing to know. At least you can be sure that if you are using the DeepSeek mobile app or the DeepSeek website, then you are exposing your data to a company based in China, and therefore, because of their laws, potentially the Chinese government, because they do have the option in their laws to demand that any corporation based in their country has to turn over data if they say so. And that’s where the whole Tiktok ban thing was spawned from, and here’s where I could rant on all day and night about how short sighted that law was so turning over to your point about banning things.

If Congress wanted to do what they said they wanted to do, which protect our data, they would have looked at the bigger picture and not written the law to apply only to social media apps. If you read it, it’s like, basically the definition of a social media app. And it says, like, I don’t I don’t even think they could ban DeepSeek under this Tiktok law, because it says in the law, like, it has to meet this criteria, and the criteria describes like, posting social media content, consuming social media content, connecting with other users like it’s not at all what DeepSeek is. So I don’t even think they could ban it, because that bill was spectacularly short sighted and actually, in my opinion, intended for another purpose. Because, like I said, if they really wanted to protect our data, knowing that AI is as pervasive in business and life as it is now, they should have taken that into consideration. But right now, our hands are tied. So as far as banning things, should we be perhaps, can we absolutely not, not under these conditions?

David Bain 

Pam, you mentioned the fact that if you use a tool like DeepSeek on their website, then you have less control over your data. So possibly a more pertinent question to get fuel for whether or not someone should be doing that is to ask you, would you use DeepSeek on the website?

Pam Aungst Cronin 

I think I’m falling into the category, especially because of the world I work in that I just have data privacy fatigue, I guess you would call it like, I’m so accustomed to my data being everywhere all the time, and just, I don’t know, I guess I just doesn’t bother me. I’m also not doing anything like sensitive obviously, you know, the government needs to worry about this a lot more than I do. I really don’t care if they know what recipe I’m making for dinner, because I looked it up in DeepSeek. Obviously, if I’m using the application on my phone, then they might be getting more than what I’m making for dinner. So I downloaded the app to test it and play with it, but I will not be using it on my phone.

Ultimately, if I decide I really want to use it, I will use the model alone, by itself, through a tool like make.com or Zapier. This is another world that’s kind of morphing into one with AI is automation. I’ve been using Zapier, actually, for over 10 years to automate my business, but now they do have the app of ChatGPT in there, and you can, through an API, connect anything you want to it. So, you know, theoretically, I would not use DeepSeek’s API on their server, but if I wanted to use it as a model, I would use it in an environment like that, where you can select the model itself and you trust the company that’s hosting it for you.

David Bain 

So if Zapier now have ChatGPT, then is there any reason to use make.com over Zapier. So

Pam Aungst Cronin 

I’m a decade plus long user of Zapier, and I tried make.com and it seemed like a foreign language to me. I see it does the same thing. It does the same thing, but just how it does it feels so not intuitive to me. So I guess it’s just like, kind of like, if you’re an Android person or an Apple person, like you just kind of can’t operate in the other world effectively. I feel like that. I’ve just born and raised with my automation life in Zapier, and I just make.com a lot of people say it’s easier to use, and I think the graphical interface is better, the user interface is better. But, yeah, no, I fumble around in there, but essentially they both do the same things. There’s some slight differences in what apps they offer and the functionality you can do through those apps, but it’s it’s the same thing for anyone wanting to start out, I’d say, play with both.

David Bain 

Garrett, Nitin talks about search turning into a personal assistant. What does this mean for conversational search? What are the differences that you’re currently seeing in terms of the way that people are interacting with LLMs, and how does this lead to changes in the SEO strategy that someone should have?

Garrett Sussman 

Yeah, there’s a lot of big picture kind of strategy we need to think about in the world of SEO. So couple of important points. One is, Google still has 90% market share of when it comes to search. So when we’re talking about channels like ChatGPT, Perplexity, you know, Copilot, things like that, it’s still very minimal, right? That’s not to say that it can accelerate extremely quickly, where we’ll see search behavior change in that respect, I’d say, in terms of the people who do use it, it extends beyond the typical way that we use keywords, you know, when we go for Google search.

As Nitin really well explained, we are typically used to just putting in the keyword email marketing, you know, pizza near me, things like that. With conversational search, our user behavior changes, where we start to be able to talk in natural language, either conversationally, where we either start with a longer prompt, or we can go back and forth to refine the output to get what we’re looking for.

Now, as Pam mentioned, ChatGPT, including the search feature that you can turn on with your searches, gives this retrieval, augmented generation, so you’re not getting output just based on what these models are trained on. It’s going out, finding relevant sites, getting that information and including it in the output. So we have more of this real time experience now.

All that to say is there’s an interesting study actually put out today by Semrush, talking about how, interestingly enough, through Clickstream data, people, when they turn on the search on ChatGPT, are still reverting to our traditional search behavior of like, putting in the very short, not natural language types of searches. Even though they’re using ChatGPT, when they’re they don’t have search turned on, they’re talking to it like they’re talking to a person.

I anticipate over time, we’ll move more towards that assistant, sort of, you know, if you ever seen the movie, her being able to converse back and forth. These assistants and these environments are going to have better context of who we are. It’s going to have more information as Pam said. Right? There’s this trade off, like your data is everywhere, to some extent. If you want to engage in this tech and provide more information, you’re going to get better results. If it knows where I am, what I do, what my interests are, I’m going to get more personalized, better results than you know if you go and search on incognito, and it has no idea we know this now they’re inter they’re integrating more tools, like the AI agents, as TJ was mentioning, like recently, ChatGPT just put in operator, which allows, basically, ChatGPT to go to a browser and do things on the browser for you. Right now, the text not great, right? It takes a really long time to do a simple task, and it messes up constantly. So while there’s excitement there, it’s hard to really go all in.

Plus, like ChatGPT is like, you need to spend $200 a month to do it. And I don’t know if the average person is going to be dropping that kind of cash, unless they’re getting it, you know, from their business, or they’re like us, and they’re experimenting. All that’s to say, a lot of exciting things are happening. User behavior, search behavior is inevitably going to change. You know, when we have that conversation a year from now, we might see a lot more people using these types of agents, like these personalized search experiences, not 100% guarantee, like, the tech might hit a ceiling, and we might just be like, this isn’t good enough, or it’s good enough, we’re going to use it.

But the implications for brands is now, how do you think about if you’re not getting as much traffic from search, it’s more of a brand visibility play, which we’re seeing in conversations around SEO, is how much availability of market share, share of voice, are you showing up if you’re not generating traffic, are you still being the recommended brand for consumer based choices or consumer based searches?

So there’s a lot to unpack there, and I think that the smart brands and smart agencies are starting to think about it without necessarily jumping all in, because going back to my original point, we’re still talking 90% market share for Google. And even though they had their earnings report yesterday and said they’re investing, you know, significantly more, $75 billion more than the $60 billion they initially implied coming up in 2025 it’s still, it’s still their world, and they have a very strong mode that who knows if their market share is ever going to decrease.

David Bain 

So Garrett, when you talk about user behavior, search behavior changing with regards to interacting with LLMs on a conversational basis, are you talking about moving towards more voice search, or are we still being led by people typing into search engines?

Garrett Sussman 

I think there are a lot of directions we’ll go in in terms of multi-modality. I think when we look at search results, sometimes we want a text response, sometimes we want a video. You know, it depends on the use case. I’ve had a lot of interactions like, I like the advanced voice on ChatGPT, where I’m brainstorming back and forth and able to talk to it of how I want to build out, you know, a webinar presentation or ideas, or, you know, podcast questions, like we’ll go back and forth via voice, where that is more appropriate when there’s other times where I’ll use a canvas feature where I want to work in the text back and forth there.

The other really interesting thing to pay attention to is something that Google is rolling out probably this year, called Project Astra, which is the ability to use your video, similar to Google lens, where the AI can see all of the inputs around it, the audio, the video, you know, the be able to read and interact with your environment across different types of modes, so that you have a completely different, you know, experience. It’s almost like Google Glass, but like making it fetch, like making it actually happen where people are going to use it, I think that that’s going to be a different use case.

I remember last year when Google said they were rolling out Bard, or, I guess a couple years ago, at this point, we thought that search was just going to change. It was going to be AI, but they’re like, No, we’re rolling out 20 different products. And so we’re going to have all these different AI use cases, like we had Notebook LLM, which is, you know, this way to interact with research. We have a Deep Research product that they’ve tied on to Notebook LLM, you know, we’ve got the search product. We’ve got Gemini. It’s there’s just an overwhelming different types of tools, and mostly the best ones are coming from Google, although there are other players that we need to pay attention to.

David Bain 

TJ, we said we’d get go back to you to talk about AI agents. So what are AI agents, and how do we use them effectively?

Tejaswi Naidu 

So AI agents on a very foundational level, right? They are just a small communication layer, where you feed in instructions on what you want to achieve. You create a series of agents, and they all work together to essentially accomplish a task for you. To put it in even more simpler terms, if you want to, let’s say, find a Chinese restaurant that’s find a list of Chinese restaurants in a five kilometer radius. You can define for one agent to, you know, look up search, to do the search on Google. And then you can have another agent to identify what’s the distance from the point where you’re searching from. And then you can also have another agent to put all this information together and give you a summarized summary of, you know, the information that you’re looking for.

So is this a series of agents where you instruct them as to what their task is and what their expected output should be, one excellent platform that does this is the open source platform called as crew AI. So with crew AI, you can use their library to essentially define what those AI agents should be doing for your use case. You can assign tools to them so you create like so if you take the SEO use case, right, there is this, well, just going by personal experience, there is this company called data for SEO that collates a lot of meta data from search like SERP results, like API for search results by countries, by languages and so on. You can connect to their API. You can create a tool using Python, which is super simple to code with, create a tool, connect it to that API, and then assign that tool to a specific agent and say, like, Hey, I’m looking for the top 10 SERP results for Chinese restaurants in Georgia, for example. And then you take that information, send it to another agent, and you instruct that agent, okay, now that I have this data, create a summer summarized, summary of the reviews, of the ratings. You know, what else you you want to do, and then you have another agent, if you want, you can define to do another task for you.

So AI agents is pretty much like it gives you complete power as to how you want them to behave, how you want them to act. How do you want them to you know, output content for you. The vast majority of OpenAI agents still use OpenAI’s API for, you know, all the connections, but you can easily use any other LLM out on the market to define your use case. So that’s broadly what they do on a high level.

David Bain 

Let’s just go around the panel briefly and focus on the question, how do you know what to automate and what not to automate? Because you can do so much nowadays, it’s very difficult to actually know what to automate and how to do it effectively, and perhaps what humans should still be doing. So Katarina, shall we start off with you. How do you actually decide what to automate?

Katarina Dahlin 

Well, basically, I took this mindset into my daily work and just looking at my task and thinking, can I actually automate? This is there a way? And sometimes I try and it just works out, and sometimes it’s not. But I think you can, yeah, you can automate a lot of things, but one thing that I still don’t want to automate is to, like, just have their quality checks. Like, don’t auto publish things. I still want the human to, like, check the text or the content through before publishing.

David Bain 

Okay, that’s a good tip. They’re doing automate quality checks. Nitin, how do you decide what to automate?

Nitin Manchanda 

I think, wherever I see any patterns, I think about automation. And this is not like, you know, in this AI era, but in general. So whenever I see like patterns, for example, you know, content production for let’s say hundreds and 1000s of pages, right? I think about automation like, what is common, what is not, and if it’s common, like, can I templatize it? Can I rely on programmatic SEO, or can I also bring in AI to some extent, right? So I think about automation wherever I see patterns, and that is, yeah, for me, rule of thumb, since I started my career, or even before that.

David Bain 

Garrett, what were your thoughts?

Garrett Sussman 

I agree with Nitin, it’s like kind of speaking to the patterns. We have to remember we’re talking about AI a lot in this context. It’s machine learning, right? And so that’s what AI is machine learning is good at, is identifying those patterns. So always needs a human anyway. But like classifications, it’s really good at summarizations. There. There are a range of different use cases. What I’d say not to use it for is to replace critical thinking. You know, I’m still skeptical about a lot of insights. It can point you in the right direction. It can shine a light on data, but it can’t necessarily do the work for you. It’s a tool, not a replacement. So I would say, really write with Nitin in terms of pattern recognition and classifications.

David Bain 

Pam, what are your thoughts?

Pam Aungst Cronin 

Whenever I think of automation, I think of the word tedious. I want all things tedious in my life like things that are just repetitive and mindless and just not worth my time when I could be spending my time on much more strategic things. So for example, my best automation I’ve built to date is my client onboarding. One this used to take over an hour, one plus hours to onboard a new client, because I have to put it in like all these different systems. We use Google Drive for storing our files, so copy the proposal needs to go there. We use Slack for Team communications. I got to create a channel there. And we use QuickBooks for billing. So you got to create a customer record there. That whole thing, I’ve used Zapier to automate a new contract comes in via DocuSign, a copy is automatically uploaded to Google Drive via PDF. A Teamwork project is automatically created for my team. A Slack channel is automatically created. A customer’s automatically created in QuickBooks and the invoice. It’s like, how long it even takes me to say all that like to do all that would take over an hour, and now it takes zero minutes and zero seconds.

David Bain 

TJ, do you outsource or automate everything that’s tedious?

Tejaswi Naidu 

I think it’s a mix of both outsourcing and tedious. But like Pam mentioned, anything tedious should be automated. And like Nitin mentioned, 100% agree, right? You see patterns you want to automate it. That’s the way to go. I mean, on a lighter note, when Pam mentioned tedious. I also thought about like folks that are now, you know, coming up with AI girlfriend services. So if you have a real person that’s tedious with you, you just go online and chat with your AI girlfriend or boyfriend, right? So I can see a lot of patterns and a lot of tedious also directly impacting our personal lives, too. So I’m not sure how that’s going to impact, you know, years from now, but I already see the early stages of it.

David Bain 

Okay, let’s finish off by asking everyone the question what we’ll be talking about when we record the state of SEO in 2026 hopefully in a year’s time. So have a think about that particular question. And we can also finish off just reminding the listener who you are and where people can find you. So Katarina, shall we start off with you again? What we’ll be talking about when we record the state of SEO in 2026?

Katarina Dahlin 

I think there will be just a lot of new tools that that we can use, and maybe it’s just becomes easier to automate things.

David Bain 

A lot of new tools and automate more. It’s hard to predict is that we were saying, Yeah, that’s the thing. So Katarina, where can people find you please?

Katarina Dahlin 

You can find me on LinkedIn. I’m mostly active there, yeah, just Google search for my name there. I probably show up.

David Bain 

Thank you very much for joining. Nitin, what will we be talking about this time next year on the same topic?

Nitin Manchanda 

I think we will definitely be talking about the core right, understanding user intent, what they’re looking for, how you know the what kind of content presentation they would be looking for, right? AI is fancy. And I think we are talking about a lot of all those shining stones, right? It’s easy to get carried away, and I think we need to really stick to the basics, to understand what users are looking for and how this search pattern is and how we can serve them the best possible content at the same time, thinking about EEAT and all of that.

So sticking to the basics, I think would be something that we would be talking about next year and maybe five years later as well. So that’s my take on that.

You can find me on LinkedIn. I’m pretty active there, so happy to answer any questions you might have.

David Bain 

Thank you very much Nitin. Garrett, will you be talking to real David or AI David this time next year? And will you be able to tell the difference?

Garrett Sussman 

You know what I’m going to be talking to the boardroom. This is the biggest opportunity for SEOs, is to reposition ourselves in enterprises as in brands, as the having a seat at the table, because as AI and marketing and conversational search becomes a bigger conversation in its own right, we understand the best way to build that visibility.

And so regardless like wherever Google goes, more and more consumers are going to be using these tools. And so brands should listen to us in terms of, how can we leverage all of the marketing we’re doing to make sure that we’re showing up, because when you see Google search, it’s going to be more dynamic. It’s not going to be static. And we need to change the way we think, away from this whole idea of keywords and more terms, brand visibility, topics and those types of paradigms.

So, you know, find me on LinkedIn. I’ll be speaking actually in New York at the end of April at SEO Week. But aside from that, I’m a LinkedIn guy these days.

David Bain 

Thank you so much Garrett. Pam, what are your thoughts on what we’re going to be talking about next year?

Pam Aungst Cronin 

My crystal ball says that we’re going to be talking about GAIO, which apparently is becoming a term – Generative AI Optimization, because I do think that the AI tools like ChatGPT and Perplexity are going to eat into Google’s market share. It may not be tremendous at first, but small percentages of that market share do generate a lot of traffic. Our clients are already getting a decent amount of traffic from ChatGPT and Perplexity, so I can’t even imagine how much that’s going to grow over the next year.

I mean, tech nerds like us, like we’re not surprised to hear each other say that, Oh yeah, I’ve tried DeepSeek, I’ve tried this tool, I’ve tried that tool, but like, I hear non techie friends saying, like, oh yeah, I know Perplexity. I use it. I like it. Like, that keeps happening, and I’m like, wow, okay, that’s really gonna cut into their market share. And one thing that I don’t hear a lot of people in the SEO world talking about right now is exactly, how do we optimize for getting brands found in ChatGPT and Perplexity-type Generative AI tools. And I completely agree the basics still very much apply. And we will be talking about that. I agree, but I think we’re gonna get into the more nuanced craft of GAIO.

To piggyback off the thought about sitting at the boardroom that is like, we can be early adopters of GAIO, if that’s the term it ends up sticking with. But it’s going to be a thing, whatever we call it, and executives are going to be wanting to get their brand found in these tools like ChatGPT and Perplexity. So we as SEOs have an opportunity to be first to market, per se, as experts in how to do that. So I think we’ll be talking about whether we call it that or not.

David Bain 

Pam, where can people find you?

Pam Aungst Cronin 

So you can find me pamannmarketing.com for all things marketing and SEO, and if you’d like to subscribe to my new AI newsletter, you can sign up for pamann.ai.

David Bain 

Thank you very much. TJ, what are your thoughts on what we will be talking about when we recorded the state of SEO in 2026?

Tejaswi Naidu

I’m just gonna say I don’t know, but I can tell you that my travel niche side that I’ve been working on for the last year or so, right now, I have eight times more crawl from Cloud bot, I have 20 times more crawl from Amazon bot, and I have 15 times more crawl on OpenAI bot versus Google bot. So this tells me that these AI tools are consuming a lot more information on my site than Google bot does, which leads me to two potential outcomes.

One, there’s going to be an increased Search Availability from these platforms over time, and you’ll see like Garrett mentioned, like Nitin mentioned, like Pam mentioned, you’re going to see a lot more conversational search queries coming up on those platforms as we go.

The second thing is Perplexity, I think that this tool is going to be the Challenger tool, which eats up the lions share, even though it’s a tiny percent of Google’s market share. But yeah, it’s going to be interesting, because I haven’t seen that level of crawl rates from AI bots before, and since they’ve started, they’ve been quite aggressive in how they’re consuming content.

So I’m already seeing tiny percentages of clicks coming in from these platforms, right? So is this only going to keep increasing? Is this, you know, gonna just shoot up or rock it up? I don’t know. So it’ll definitely be interesting to see how things play out in this coming year. But one thing’s for sure, nothing’s ever going to be the same. Everything’s constantly changing, and I feel that acceleration will only keep accelerating a lot more faster. So it’s just about, how do we adapt? And how do we, you know, keep ourselves abreast with whatever is changing in the in the AI SEO field.

David Bain 

TJ, where can people find you?

Tejaswi Naidu

Yeah, you can find me on LinkedIn.

David Bain

I’ve been your host David Bain, and you’ve been listening to the Majestic SEO panel. If you want to join us live next time, sign up at majestic.com/webinars and of course, check out our other series over at SEOin2025.com.

Previous Episodes

Follow our Twitter/X (@Majestic) and Bluesky (@majestic.com) accounts to hear about more upcoming episodes!

Or if you want to catch up with all of our previous episodes, check out the full list of our Majestic SEO Podcast episodes.

Back