Episode 11 – Lazarina Stoy

Filipe Santos
00:00:00 – 00:00:58

From the rise of AI-powered automation to the evolving SEO landscape, the marketing world is undergoing a seismic shift. But with so many companies racing to integrate AI, where are they stumbling and how can you avoid their pitfalls? Rethink everything you know about SEO and digital marketing because the game has changed, and it’s time you did too. This is phraseology plus a I with your hosts, Felipe and Miguel Santos. As a digital marketing professional specializing in SEO for over 20 years, I’ve worked with companies from startups to the Global Fortune 500. Learn how to gain an unfair advantage with AI as we uncover tips, tools, and strategies. So we’re here today with Lazarina Stoy. She’s an organic search and ML consultant, speaker, and educator. She’s also the founder of ML for SEO, which is a training platform for organic search marketers, upskilling, and machine learning, which right now is super critical.

Filipe Santos
00:00:59 – 00:01:34

Lazarina has worked with countless teams in b to b, SaaS and Big Tech, to basically improve their organic positioning and has also trained dozens of teams and professionals specifically in harnessing the power of machine learning for task automation, which is quite a big deal. Lazarina is also a trainer and mentor, helping fellow marketers kick off their data science journey. And we all know that we need to up skill constantly, and this is a battle for change. So with that, I would love to pass it over to Lazzarina to mention anything I might have missed.

Lazarina Stoy
00:01:34 – 00:01:39

No. I think you captured everything. Thank you so much for the wonderful intro. I’m very happy to be here.

Filipe Santos
00:01:39 – 00:01:57

Great to have you. Again, like I’ve mentioned, watching you on LinkedIn is nothing but a marvel because there are so much great information that you put out there for everyone to learn from and some really good stats, reports, and information about things that are changing. So, Lazarena, guess the first thing we can talk about here is how has automation really changed SEO?

Lazarina Stoy
00:01:57 – 00:03:05

So I think recently we’re seeing a different wave of change in SEO with the generative automations that people are doing. So I think right now it’s becoming more and more noticeable that machine learning and AI in general are about to or are already changing our industry. If we were to look at how these tasks, models, automations were used in the past. It was mostly by enterprise organizations or really, really technical niche marketers that were building massive websites. Right now, I think people have a lot more access with generative AI to large language models and they are becoming dependent on them in ways that might not serve them very well. So I think long story short, automation is changing our industry whether we want it or not. It’s becoming a lot more readily available to implement AI. There are a lot more tools and developments that are happening that are harnessing machine learning technologies that we can use now that we weren’t able to just a few years ago.

Lazarina Stoy
00:03:05 – 00:03:14

So most of the tasks right now that we are doing can be automated, and our focus is becoming more on strategy as opposed to on task execution.

Filipe Santos
00:03:15 – 00:03:38

That’s really awesome. I I was wondering, so, like, obviously, AI has kind of gotten embedded into a lot of products including Google lately, and that’s been a a sort of contention. Why do you think if there’s any other reasons beside those kind of implementations with OpenAI also being competitor? Why the certain sudden surge in AI and machine learning specifically in SEO? Has that been kind of prompted by other things?

Lazarina Stoy
00:03:38 – 00:04:29

Yeah. That’s a very interesting question. I have my opinion. I’m happy to share my opinion on the topic. Specifically in machine learning, there are several periods in the history of this field and AI in general where they are defined by really big breakthroughs in the technology. And in the past, whenever there is a breakthrough in the technology, there is also a lot of investments on the company’s side. And this spans across multiple different industries, not just in our case, marketing, digital marketing, or our very small corner of the Internet that is SEO. So if we were to take a step back and look at how AI is implemented in marketing at the moment, it is very much a result of the competition between Microsoft Open AI, Google, and AWS as well.

Lazarina Stoy
00:04:29 – 00:05:37

This competition really enables the democratization of technology that has been developing very rapidly, and companies are eager to be part of this wave. And the more the technologies are implemented, the more we are seeing also, cases where it’s not really performing as well as we want it to, like in the AI overviews, places where it doesn’t really make sense to have this technology or it’s not ready yet. And, yeah, I can talk a little bit more about what happens typically whenever these implementations hit the market. And like we’re seeing right now, people are not really happy If this continues within the next year or so, and the technology doesn’t reach the level of what the audience expects, then we might see the opposite side of the spectrum where companies are starting to pull back their investments. The same thing like what happened with NFTs for instance, even though it’s not really comparable to compare the 2. And then we might see that the industry is kind of revert reverting back to, practices that are a little bit more manual or that they don’t really utilize this particular field of AI as much.

Filipe Santos
00:05:38 – 00:06:17

You know, to me, that that’s a fascinating point because watching this evolution over the last, I’d say, year and a half, it’s been a race, an arms race to get AI and integrate into everything. But in that, even the automation prior to AI’s big announcements, like with open AI’s products, I felt like there was kind of like a struggle between, hey, we need to scale and create a lot of content to address organic search and also to get propagated against other channels in marketing. And now it’s more actually about quality because AI can start to deliver a lot of information that is kind of easier to put out there, but that also scares people because, of course, they’ve been working so hard on this. Right?

Lazarina Stoy
00:06:18 – 00:07:17

Of course. And if you look at not only SEO, but also performance marketing as well, they have a very similar struggle with the implementation of the AI systems, like performance max, for instance, that Google is pushing. And again, it’s kind of like a black box algorithm, but a lot of money is dependent on it. Not only in terms of performance marketing, where we’re talking about advertisers money. And when we look at SEO, we’re talking about the money of independent site owners, big enterprises, all of these investments that are going into content marketing in all forms, not only, written, but video marketing and so forth. So we are seeing now that our roles are becoming a lot more dependent on, AI models. I don’t really like the word AI, so I’m going to be referring back between AI and ML, machine learning models essentially that are not ready to perform at the standard that we need them to. But at the end of the day, they are implemented at scale.

Lazarina Stoy
00:07:17 – 00:07:43

And we’re kind of seeing, this issue across multiple different, not only in marketing, but across multiple different industries where the technology is being implemented because of the AI hype and companies not really wanting to get behind in this trend. But at the same time, we’re also seeing that it’s not really performing to the standard in terms of what it was promised. So there is a gap between, the expectation versus delivery of these products.

Filipe Santos
00:07:43 – 00:08:09

So far. And and you’ve got a, like, a real point there that everybody’s racing. This is nothing new. When something new comes out, the competition must be there, and you must go out and produce something. But, yeah, the quality has been kinda lackluster, especially given certain, like, the first models and how that was working for people in terms of just discovering new ways of using AI. Yeah. But I would love to know from you then. Obviously, people are interested in automation for many reasons.

Filipe Santos
00:08:09 – 00:08:24

Businesses, for example, because it’s efficient and it creates a little bit less overhead, maybe it’s a little cheaper, and it also kind of makes things consistent in terms of quality. But why are they getting so much more interested in automation and kind of where can they be going wrong in this process?

Lazarina Stoy
00:08:25 – 00:08:59

Yeah. So I think to answer the first question is quite easy. Everyone is doing it. So everyone wants to use chat gpt somehow, and we are constantly bombarded with post, like, 10 ways to use chat gpt to do whatever. Whether it’s a content strategy or write a blog post or automate your LinkedIn, like, it I’ve I’ve just not seen, like, that it cleans your kitchen or something. Everything else that I’ve seen. So, obviously, there is the trend there, and there’s nothing wrong with that. I don’t think that’s wrong to be interested in this field.

Lazarina Stoy
00:08:59 – 00:10:30

I I think that this is an amazing field for marketers. It it’s literally what I’ve been doing since the start of my marketing career and I can absolutely say that it is a fantastic area to upscale if you are a marketer. Where I think a lot of people get it wrong is that they, start and stop with chat GPT or any other tool that you have that is kind of like a repackaged, maybe slightly upgraded version of chatgpt, like a content writer, like, Jarvis or whatever else. I don’t want to name names because it’s not the the problem of these tools. It’s more like the, the mindset of people trying to get into AI or get into data science. So what I always suggest to do is to start with different projects that you’re working on and try to understand a little bit more about the different algorithms that you can use to achieve a certain task, how they work, how they have been built. And it that’s regardless whether you’re using an API or a service, just to understand what goes behind that. Because once you start repeating this process for several projects that you’re working on, where you want to harness some sort of automation, you will slowly start to understand a little bit better where you can implement some AI models or machine learning models and which tasks, what kind of data do you need in order to make the implementation successful, where you can maybe start experimenting with training your own models, especially if you’re working at an enterprise organization.

Lazarina Stoy
00:10:30 – 00:11:16

You might have access to really good data that you can actually use for training, and it it will make the whole experience a lot better. Otherwise, you might, find that starting and stopping with chat DPT is probably going to work in some instances, like maybe giving you a slight productivity boost or anything else. I don’t know. Maybe, like, as I I use it, for, tasks like brainstorming or assisting, with code writing or troubleshooting and things like that. But beyond that, I haven’t really seen fantastic implementations where it can completely replace a task that you are doing. So the earlier that you start experimenting with other machine learning models, the better your experience in terms of learning machine learning is going to be. That’s what I would say.

Filipe Santos
00:11:17 – 00:11:47

Yeah. That brings me to a a really solid area that I wanted to ask you about, which is basically a lot of folks that are listening and a lot of folks that are trying to get in and understand AI and not be, like, intimidated by it. You mentioned here, like, you know, that you need to know how they work in order to best implement them. So what’s the easiest way for folks to kind of get familiar or kind of research or figure out what the innards of an AI algorithm are? Or is it more about, like, following folks like you to really get a sense of what that is because you’re breaking it down?

Lazarina Stoy
00:11:47 – 00:12:52

Yeah. So I wouldn’t toot my own horn here, and I would say don’t follow just folks like me, but actually branch out and follow folks that are machine learning engineers, data scientists. There are fantastic examples perhaps I can share with you after the call so that we could put it at the description of the video or wherever for people to visit. But essentially, in terms of choosing the right resources, any kind of task that you are doing, let’s say for instance, just to give a few examples, you can have a task with 3 different types of data. So there are 3 things that you need to think about. What is your solution? What is the type of machine learning task that you need essentially? And what data do you have and what kind of output do you need? So let’s say for instance, we have images and we want to caption them. We need a captioning algorithm that is essentially image recognition that outputs text. That’s essentially our problem for writing alt text, but, translated in terms of machine learning language, so to speak.

Lazarina Stoy
00:12:53 – 00:13:56

Let’s give another example. If we want to give, someone like an algorithm or whatever to read the page, read the title, and write a meta description based on that. That’s text to text transformation. We have a limit in terms of the characters that we want the output to be, and it’s a a summary task. So that’s kind of how you would think in terms of, tasks on how to translate as your projects into machine learning, tasks and how you go about researching what kind of algorithms you could potentially use. If you want to transcribe videos, for instance, you need audio to text transformation. If you want to, let’s say, forecast, you have forecasting algorithms, like now you’re in the field of machine learning researching for these, the next task would be to understand for your particular project, what are the different approaches that you can take? So there might be, unsupervised machine learning, supervised machine learning. It might be deep learning like neural networks and so forth.

Lazarina Stoy
00:13:56 – 00:15:39

Some of these approaches are surface level, a little bit easier to understand, and easier to implement if you’re building your own algorithm from scratch. Even if you’re not, if you’re just trying to use an API, which I think would be good for most beginners to start by using different APIs just to get a feel of the field, that it would be good to understand what algorithms this API, is using so that not only you feel more comfortable whenever you are implementing this API, but also feel comfortable explaining how this particular task was completed with the help of the machine learning algorithm. Because most likely, you will have a team that you’re reporting to or stakeholders and so forth. Like I mentioned, the more of these projects that you do and the more of these mini implementations that you yourself are involved in and you’re overseeing and you’re learning about these different algorithms, the more you will find differences between them and you will be able to articulate when to use one algorithm versus the other and so forth. So just to summarize my tips, follow people outside of marketing that deal with data science and machine learning engineers, try to translate a particular SEO project into a machine learning task, then research the different algorithms that you can use to complete the task and find out a little bit more about the methods and the ways that those algorithms have been trained. Or if you are training your own, obviously try to understand it to the level that you can articulate it and implement it. But the good thing is that everything that you’re doing most likely has been done before by someone outside of the marketing field. So it’s just about finding ways to implement these things.

Lazarina Stoy
00:15:39 – 00:15:48

And then if you’re a really advanced technical marketer harnessing machine learning, then you kind of branch into the field of training your own algorithms, working with data scientists and so forth.

Filipe Santos
00:15:48 – 00:16:30

And I know that it’s extremely helpful because the way you define all of those elements and how it all plays together is exactly how it works. But the thing is with folks that are really trying to figure out how like in a business, let’s say a small business, an entrepreneur or a startup founder. How can they really kind of start to probe, like, how their business can best work? Because, like you said, chat gpt is more generalized. There are more, proprietary or smaller models that can probably be helpful in businesses, machine learning, for example, getting the right mix of internal data, first party data. So I guess we can kinda stem that into what companies or what types of companies would be good for AI or ML automation?

Lazarina Stoy
00:16:30 – 00:17:50

To answer the question, I think all of the companies, like, every company can benefit from it. And I think it’s more like every person that works in a digital environment, especially in marketing, has ways to boost their own productivity by has ways to boost their own productivity by harnessing machine learning. So in marketing, I think what makes it a really good field to work with machine learning models and to automate tasks is because we have a lot of data at our disposal and not only numeric data, but we also have textual data that we have to analyze in order to create better strategies, especially in SEO. We work with digital content. What was important before and is still really important, of course, is to understand natural language processing and how to integrate insights from that into our SEO strategies. And now because we have amazing models that can transform video content to text, We also have another avenue that is opening to us to integrate insights from video strategies of competitors or the landscape in general and create a better SEO strategy focusing on Google search, but also YouTube strategies and improving organic search in general. So that’s just one way that marketers could use it. So I think for every marketer, you should think about how to integrate it.

Lazarina Stoy
00:17:50 – 00:19:22

And when it comes to the companies, obviously enterprises have a bigger advantage implementing these technologies because you already have engineering teams, data science teams, and they have a ton of data that they can use to analyze and to improve their, content and whatever strategy that they’re working on to also analyze user feedback, because a lot of them have a ton of feedback from users that is just sitting there, not really utilized or incorporated as part of strategy to also analyze sales data and so forth. But marketing agencies, but I’m also thinking at the same time for business owners, the sole owner of a business related to a product or a service. So in both of these situations, you have routine processes that you need to do related to upkeeping your website, related to updating blog posts, related to improving your online strategy, analyzing competitors, making sure that content is also transformed from a blog, post to social media posts, or maybe you’re also targeting TikTok or YouTube. So thinking about all of these components and the little ways that you can make all of that work, to be more streamlined and a little bit more automated, every little minute will help these people. I think regardless of the size of the company, I think machine learning is definitely something people that should be looking into.

Filipe Santos
00:19:22 – 00:20:12

The world that opens up for every company size. I mean, small businesses that are probably right now, like, so lean that they’re unable to operate in in many different channels now have an opportunity with the right kind of structure and right kind of time spent in automation and in machine learning can, you know, drive, some of those results that they probably couldn’t have before. On the same token, I think you’re right. The quicker companies get into this, the sooner they can advance those motives and get the benefits of the resulting work, while also beating a lot of competitors to the punch. Because a lot of folks are not gonna be jumping in right away. And those that do will at least be able to take advantage of the technology earlier and then refine it for their own purposes. Right? But I’ll let you kind of see if there’s anything else that I’m missing here in terms of how that can help smaller companies.

Lazarina Stoy
00:20:13 – 00:20:59

Yeah. No. You’re absolutely spot on. I think it’s important to start early. I also think that it’s important to do it at a pace that is comfortable for the company while also keeping track of the developments that are happening in the industry. So I think a lot of people, especially smaller companies, are intimidated to provide that initial investment in terms of time needed to set up processes or to train their team, maybe even investments in terms of the courses that they would need, like their team members and things like that. The reality of automation is that you need that initial investment in order to save time down the line, kind of like compounding interest scenario. So it really, really makes sense to, start as early as possible.

Lazarina Stoy
00:20:59 – 00:22:27

I always say this like 10 minutes per day machine learning practice for everyone at the company, because at the end of the day, if we were to imagine the future, maybe 10 years from now, a lot of the people will be skilled in data science and machine learning. So you wouldn’t want your employees or yourself or your business to be catching up on like 10 years worth of advancements as opposed to, you know, being on track with all of these developments, throughout the time. And and just to say most of the things that we are doing right now in marketing, we have a ton of APIs, algorithms that are available at our disposal, and we’re not really utilizing them as much as we can. Massive shout out to pioneers like Hamlet Batista, who unfortunately is no longer with us, but he has been a massive inspiration for a lot of people in the SEO community in particular to actually start experimenting with Python, to start doing machine learning. And I wanted to mention it because it’s so important for people to know that it’s never too late to start. It’s never too late to be inspired by someone and to really start implementing the technologies, even if it’s something that, you know, has been done, it’s better to do it now. Slowly and surely, you will start seeing the opportunities for automation everywhere if you just remain consistent with it. So sorry for going on a long tangent.

Lazarina Stoy
00:22:27 – 00:22:29

I just wanted to mention it.

Filipe Santos
00:22:29 – 00:23:05

No. No. That’s a really good point, and I think we should never forget. Hamlet Bautista was and I remember, like, he was putting out scripts. He was kinda getting people into Python and hey, look at what we can do with Python to actually automate or get more better data from the SEO side of the house. So really critical to mention that, and I’m glad glad you brought it up. That being said, there are these folks out there that are putting out really good information that are getting data science married with SEO a little bit better. And taking us away from the traditional SEO stuff that sometimes is flawed or based on myth and putting us back into the testing and trying arena.

Filipe Santos
00:23:05 – 00:23:26

And now with AI in the mix, it just gets us a little bit more advanced in terms of scale and understanding the quality and how we can tweak things. So from that perspective, La Zarena, how about your story? Like have you everybody’s got a different journey. Have you seen any really good real world applications lately in terms of challenges and successes of implementing AI and SEO?

Lazarina Stoy
00:23:26 – 00:24:38

Yeah. So I can talk, about a few experiences that I’ve had. I’ll try to keep them generic so that we don’t, you know, name. So in terms of, rule based automation, one area where, machine learning really shines and there are several different, algorithms that you can implement to build a better system, is within internal linking. So I’ve worked with a couple of companies where we were able to really scale the auditing of internal links and to recommend a lot more targeted and tailored suggestions for how the pages should be linked together with one another. Another model that is quite, commonly used with really big enterprises is, enterprises is, entity analysis. Depending on the type of data that you’re doing the analysis of, it can be extremely powerful. So I’ve been involved in several projects where we’re doing entity analysis and kind of text classification with the aim of understanding better, maybe like a historic catalog of, pages that are related to news or maybe even not really well categorized side catalogs of blogs.

Lazarina Stoy
00:24:39 – 00:26:38

Whenever you take the same, like, entity analysis process and you try to implement it in the context of internal links, then you’re able to understand which articles, speak about different entities that are important. And then imagine also pairing that with entity analysis and keyword research, where you start to really understand the structure of keywords, how different entities are referred to by users, how queries are structured, and then also incorporating concepts like sequential queries and query paths to better understand the whole user journey. You know, how Google pushes users to different searches and how they present information to them. Then also taking that same entity analysis and all of this data that you already have from your own content, from the keyword research, the queries, what Google, suggests to users, and also pairing that with how your own users speak, converse, like how they refer to different products or entities when they are talking on social media, when they’re leaving comments, when they’re giving feedback and user feedback forms, how they’re referring to in the calls that they have with your sales team and so forth. This is a very good example of how even with something as simple in terms of machine learning, like entity analysis, When you really understand how to implement this concept in s, you can switch the input data that you do the analysis of and then extract a ton of different insights in terms of how you integrate these insights and, understanding of entities into the strategy. It kind of feeds back into all of these different aspects of how the brand performs. So these are successful implementations that I’ve seen. Lately, I’m seeing a little bit more of the unsuccessful implementations that are related to generative AI and producing content at scale.

Lazarina Stoy
00:26:38 – 00:27:54

Very recently, I had a conversation with a partner of mine where we talked about what is the best way to integrate generative AI because, of course, I do believe in the power of scaling content automation. I am a big proponent of programmatic SEO whenever it’s done right or otherwise, like database driven content systems as opposed to programmatically generated content, which these two things are, a little bit different in my eyes. One is only focused on the method of producing the content, and the other is focused on the method of creating pages, organizing information, managing SEO processes. So programmatic SEO does not mean generating content with chat GPT. It could involve that maybe in some scenarios, but it doesn’t equal that. So what I’m seeing is a mistake that companies are making in terms of their implementation is, first of all, they are relying on chat gpt to do the on page structures or to do the research. So it’s important honing down on the point that I said a couple of minutes ago that it’s important to understand how generative AI works. Just a very quick shout out to Brittany Mueller.

Lazarina Stoy
00:27:54 – 00:28:57

She has a course on introduction to generative AI and a bunch of different resources. So I think she’s the best person to learn from in our industry right now on how it works. And it’s important to know what it excels at. So it excels at text to text transformation. It excels at generating text from a prompt. So relating that back to content production, it’s not wise to rely on chat GPT or any LLM to research. It’s not wise to rely on them to do on page SEO because on page SEO is strategic analysis and also should be informed by the content that is already ranking in the SERP and also our human brains that, should tell us what information is missing from this content that is ranking well. Because in order for pages to be, endorsed by Google, they should also present unique information, and ChargeGPT really cannot give you that because LLMs in general, they’re only giving you information based on what they’re trained on.

Lazarina Stoy
00:28:57 – 00:30:15

They don’t have an understanding of the real world. So this is a big tip that I can give to companies. If you have processes involving chat gbt, creating on page SEO and structures, then don’t do that. Ensure that that is done informed by a SERP analysis, informed by competitor research and informed by an SEO consultant that can actually do entity analysis, understand the entity landscape and the information that is missing and how your brand can stand out. If you have all of these components, you might be able to develop detailed enough prompts for the different sections of your page where you can get drafting support. And I do mean drafting with an LLM. And I I say drafting, not writing because the draft requires an editor, and you do need a human editor to go in and improve the content and to actually expand on this information. And what you might see in projects like this is that it actually makes sense to go the traditional route as opposed to incorporating an LLM or generative AI because at some point, if you do have a very skilled copywriter that is doing all of this research for you, it would make sense for them to do the draft and the editing and so forth.

Lazarina Stoy
00:30:15 – 00:30:44

So in a lot of cases for companies that I’ve, consulted, they actually revert back to their old practices because it’s more cost friendly and it’s a much higher quality result as opposed to incorporating chat gpt and spending a ton of hours to edit the output and to make it better. So hopefully, I’ve provided enough examples of both the good and the bad implementations, and the people listening can have some takeaways from my experience with clients.

Filipe Santos
00:30:45 – 00:31:29

I I definitely feel a little bit smarter just from hearing that. That that being said, you’re probably breaking a lot of hearts in terms of use of generative AI at this moment, but you’re right. It’s all about the implementation. The programmatic pages does not mean just churning out stuff from chat DPT, per perplexity, or anything else. It’s about coming up with ways to structure a page that addresses a concern in the best way possible for SEO. Right? So I love that. I love the fact that you’re you broke it down to the the components and what actually works with versus what doesn’t. We’ve seen that maybe in the first couple minutes of AI’s release in terms of generative AI, folks were able to get away with a lot and there were there were some experiments that were a little insane and they worked very well for a little while.

Filipe Santos
00:31:29 – 00:32:05

But again, that wasn’t a long term strategy. With that, I guess, that’s a perfect segue, Lazzarina, into maybe your considerations or thoughts around why this sudden surge of artificial intelligence, market, machine learning, and why it’s so significant for digital marketing, and how companies from every size, small startups to large companies, can use these technologies for automation and efficiency. We’ve kinda covered some of those topics, but I would love to know a little bit more about why you think or how you think people can move along this tangent as new technology updates occur and as AI and and machine learning a little bit better understood.

Lazarina Stoy
00:32:05 – 00:33:26

So I think this search is significant for digital marketing because it’s important for us to pivot our strategies a little bit in some aspects and to also become a little bit more in tune with these technologies so that we could hopefully potentially have a voice in terms of what works and what doesn’t. Also advocate for this technology in the organizations that we work in or with the clients that we work in. I will expand a little bit on each of these points. So in terms of understanding what works and what doesn’t, we’re all seeing examples of where this technology is implemented and it’s not really giving a very quality output. So I think it’s important to understand the pitfalls, why the results are like that, why the company, like, for instance, Google, why are they doing these implementation, what is driving them? And to also be able to advocate for change in terms of what’s happening with our industry. That’s one thing. So in order to be insightful with our recommendations, we need to be educated in terms of what this technology does. The second thing is to also advocate for, this technology to be implemented for productivity improvements in the processes that we work and this is partly selfishly for us to be working on more exciting things.

Lazarina Stoy
00:33:27 – 00:34:12

Like, I’ve seen a lot of people recently say something that, of course, the technology has been there for years, but I’m happy the industry is waking up to it that nobody should be writing captions or the descriptions by hand anymore because there’s so many algorithms that can do that for you and we should at least try to get it to 90% of the way automated and then have a human review. So whenever we advocate, for these type of implementations, we are freeing our own time to work on projects that are more strategic and that are more advancing for us as consultants and, you know, are developing our skill set in ways that is not yet able to be replicated by a machine.

Filipe Santos
00:34:12 – 00:34:40

Absolutely. Well, let’s talk a little bit about the automation. Right? So, obviously, there’s a lot of routine marketing tasks that are out there that have been there that have caused stress for, many a marketer and many a leader. With machine learning specifically, I guess, could you talk a little bit more about the entity analysis and content transformation part of that? Like, what in the simplest terms can you say would be useful for this audience to understand about it and about making use of that?

Lazarina Stoy
00:34:41 – 00:35:52

Yeah. So, I think in terms of entity analysis in a previous response, I I gave a few examples of where you can implement it. The the quickest takeaway in terms of entity analysis is to understand what entities are, what is the EAV model, like entities, attributes, and values, and to understand how that relates to queries and keywords. So in terms of actually implementing entity analysis, I have a resource on doing this in Google Sheets, so you don’t really need any coding skills in order to get started. My suggestion is to use an API that is focused for this task as opposed to relying on chat gpt to do entity analysis because based on my own testing, it’s not really a very reliable source. And with a focused, API like Google’s natural language API, for instance, you can get, a much better output in terms of the entities that are recognized, what type of entity it is, what is the reference to this entity in Wikipedia if it has it, and so forth. So it’s a lot more robust in terms of the data that you get. It’s also important to know whenever you’re doing a task like this, how to implement it in different projects.

Lazarina Stoy
00:35:52 – 00:38:01

So during this podcast, I reference things like implementing it in internal linking analysis or in keyword research when you’re organizing topical maps, when you’re, analyzing content, to extract the main entities, and maybe you can use that for tagging purposes and so forth, and you can even go beyond that. So that’s as much as I I think it’s okay to say for entity analysis, and then your other point was for content transformation. So when it comes to content transformation, I think it’s very important for people to start thinking about this because not only the changes from Google lately are alerting us to the importance of having an omnipresent brand and not just relying on Google search as the main source of traffic or brand awareness or income for many websites, but also the user journey and the user preferences are changing and they have been for a long time, but now we are seeing platforms like TikTok, YouTube, and even social media communities and online forums becoming a lot more present in the user journey. And I think that there is room for all of the different components of the search journey, but, we don’t yet know what is going to be the pivotal component that is going to make people buy or make people complete the purchase or whatever else it may be that is important for your brand. So in that line of thinking, it’s important to start diversifying the content channels that you pursue. And for that, there is a ton of machine type to another, or to at least help you to get, like, 90% of the way there. Or to at least help you to get, like, 90% of the weight there through automation. And one approach that I think works really well and it’s very personable and is based on the human element is to start from a video, transcribe that into blog post, transcribe that into a social media post for the different platforms, and then also chunk the video up into shorts, snippets for TikTok, and so forth.

Lazarina Stoy
00:38:01 – 00:38:56

This actually allows the human content creator or brand expert or whoever it is that is responsible for that content creation to be present for, let’s say, 1 hour. And from this hour, you can create maybe 2, 3, 5 blog posts depending on how organized you are with that 1 hour. You can create 50, 60 different shorts and 5, 10 different, social posts for different platforms. And, I think this is the way in terms of the mindset of how we should be thinking about content creation that, you know, we need to have the human element. We need to have a very robust research. We need to ensure that we are not replicating information that is already out there, so we’re not populating the web with junk. And we also need to harness automation in a way that amplifies human voices as opposed to creating robotic sounding videos and podcasts and so forth. I I think for these two things that you mentioned.

Filipe Santos
00:38:56 – 00:39:48

Yeah. I I like how you mentioned atomized content because for, I I think, the last 3, 4 years, folks have been trying to do it, and their workflows became so much more complicated and difficult because they’re trying to break up a piece of an initial theme of content into many parts and across many channels. And now I think with AI with the right kind of AI implementation, you can do so much and so little with even a small team. So that’s a great use of that technology and a great use of kind of being able to be out there and being broadcast in the right way with the right information. With that, we go into the dark side because you’ve mentioned some good examples of, you know, AI and machine learning at use. But a lot of folks are really perplexed and annoyed by AI overviews. So why do you think they’re so bad right now? Why do you what’s your opinion, or what have you been seeing that has led to this poor execution?

Lazarina Stoy
00:39:48 – 00:41:02

Oh, well, I think a few things. So first of all, back when I was in university, which was almost 5, 6 years ago, whenever we were learning about machine learning and the approaches, ethics was a really big part of it. So we were talking about how can machine learning be trained ethically. So I think that’s the biggest pitfall right now that not only the training is not done in an ethical way, the information is not sourced ethically and it’s not sourced in alignment with the standards that we have in terms of data protection, but it’s also not quality checked in an ethical way if you read a little bit more about how actually the human review systems work whenever there is a generative AI training. And then that lack of accountability and that lack of ethical implementation follows through to the actual implementation of the technology. We’re seeing very unresponsible practices specifically from Google. Not only OpenAI also had very big mishaps in terms of the text that it was creating because it’s very easy to gain generative AI. In a nutshell, that’s the biggest problem.

Lazarina Stoy
00:41:02 – 00:41:59

It’s very easy to gain generative AI because the way that it works, it’s based on a model that essentially provides what it thinks as the logical response. It doesn’t have a real understanding of the world, of the entities it’s mentioning, and so forth. So it might someday, whenever this technology is merged with, like, fact checking algorithms, with real world entity analysis, or whatever other, models can be included. Right now it doesn’t. So in terms of the AI overviews, I think specifically they are very harmful because they are first of all being presented for queries where they shouldn’t be, like medical, your money, your life queries. I’ve seen examples that are extremely harmful. That’s something that Google could have guardrailed. We’re also seeing just a lack of fact checking for these responses.

Lazarina Stoy
00:41:59 – 00:42:44

Also, just a shout out to Lily Ray that has been covering all of this on Twitter. She also highlighted something else that the information that they’re pulling is being pulled from several different sources, but they are in the top 100. So most of the time you’re seeing actually really low quality sites that are informing this. So similarly to other SERP features that Google has, they’re not actually collecting, reviewing, and investing into ensuring that the information in the data that they have is actually good. They’re not investing in this data at all. They’re just surfacing and kind of repack repackaging the information. So I think that’s where it falls. There’s just lack of accountability, lack of ethics.

Lazarina Stoy
00:42:44 – 00:42:48

It just results in a bad experience and very, very harmful information presented.

Filipe Santos
00:42:49 – 00:43:20

Yeah. Good, good point on, Lily Ray. I’ve I’ve seen a lot of our posts with examples, and they are pretty catastrophic and and scary in some cases. I’m like, wow. Imagine if people actually trusted those results. That would not bode well for anyone. And the fact that they’re not using the same, at least some of the logic that they’ve used in previous algorithms is really strange to me. I I guess that’s why we’re we’re all complaining about it and why folks are trying to return back to the old, non AI generated search results because that at least is reliable.

Lazarina Stoy
00:43:21 – 00:43:22

Is it? I don’t know.

Filipe Santos
00:43:24 – 00:43:26

More reliable, hopefully.

Lazarina Stoy
00:43:26 – 00:44:09

Yeah. Especially after what we’ve seen from the past update last year. I don’t know how reliable it is. That’s the reality of our industry at the moment is that it’s very messy what’s happening. And that’s why I think it’s important for us as consultants to remain as vigilant and as educated as possible because it’s very difficult to have conversations like this whenever you know that the performance isn’t really as much in your control. And over the past couple of years, we have seen a lot of examples of brands doing everything right and still getting the boots. So it’s really hard to remain focused on what really matters. And for people that are wondering what does really matter, it’s the users that matter.

Lazarina Stoy
00:44:09 – 00:45:16

It’s not Google search or DuckDuckGo or Bing or YouTube or whatever. It’s the users that matter. So I think we as marketers might evolve from SEO the way that we see it being focused purely on Google search, and we might evolve more into addressing search behavior and finding more easily, more quickly the users for our brand or service, whoever we are working for, wherever they may be. So that’s how I think the role of SEOs might change in line with all of these developments that are happening in the industry. Because at the end of the day, what really matters is users, how they search the particular potential clients for the brands that we’re working for, and how we can actually get better connected with them regardless of the platform where they’re at. And I think a lot of SEOs are now realizing that, and Google is just no longer that much of a reliable platform for us to be purely focused on, and we should actually diversify our skill sets. But a lot of the consultants that I know have been doing this for years. So it’s not really a surprise to a lot of people, I guess.

Filipe Santos
00:45:16 – 00:45:49

Regarding the the con the quality of results, for example, we’ve been here before though. Most listeners won’t remember this, but way back in the early days of Google and SEO, it was all keyword spam. There were some really, really terrible results. People didn’t know how to research and the second thing is that, you know, unfortunately, a lot of practitioners were just keyword spamming to get to the top of results. So this is kind of to me, it’s reminiscent of that. It’s just more complicated now, but it’s the same thing. So it’s kinda site it’s coming in cycles. That being said, let’s talk a little bit about the future.

Filipe Santos
00:45:49 – 00:46:00

So are there any insights that you can share regarding trends in machine learning, data science, with regards to marketing, and perhaps maybe even the next big breakthrough that could come our way?

Lazarina Stoy
00:46:00 – 00:47:38

So I’m going to be very pessimistic, and I’m going to say that I think that we’re headed to another AI winter, especially if the issues with current generative AI models and how they’re integrated into systems are not fixed. And my main predictions so just some context for the people that don’t know what an AI winter is, is essentially where you have a period of, this technology is adopted by businesses very rapidly. There’s ton of investment for it into businesses and AI research. The technology continues to grow, but as much as it is implemented, it’s not meeting the expectations of businesses, shareholders, clients, and then the investments start to dwell out. That firstly impacts companies, but then it also impacts AI research as well. So how much funding AI is getting as an industry stops or at least diminishes for a certain period, stops or at least diminishes for a certain period until the next big breakthrough. So far in the history of AI, been 2 major AI winters. And I think right now, because this breakthrough is so massive and so misunderstood in terms of how companies are, implementing it and also the ethical issues that are related to how data is sourced and the lack of legislation currently in terms of who is protecting the data that AI is being trained on, all of these factors when they collide, if they collide, and I think they will, will probably lead to a lot of companies pulling back from their implementations that relate to generative AI.

Lazarina Stoy
00:47:38 – 00:49:30

Just to give you a very recent example from Google’s I IO, a couple of weeks back, they mentioned the incorporation of Gemini, their AI model, and system into a lot of different Google workplace tools. And whenever you start thinking about the access of data that these systems are going to have, in order for Gemini to be able to read your email history and write an email for you, starting to think about all of the sensitive data, not only on a personal level, but also on a company level that exists in these domains, in spreadsheets and documents and so forth. We know that they have access to this data obviously because it’s hosted on their platforms. But now when you start thinking about how the models are are trained, how poorly they are performing in terms of guardrailing all of this data, then you start to think, okay. What’s going to happen if, let’s say, a company like Microsoft, maybe that’s a bad example, but let’s say Zendesk or g 2, all of them incorporate Gemini as their AI assistant, and it has access to their emails. What will happen if someone penetrates, you know, this database or even finds prompts or engineering tactics to make, Gemini give that information away. So here we’re kind of going into very unfavorable scenarios, for these organizations, but they’re also very real scenarios that, can happen if these implementations continue at the pace that they are without having proper security, without having proper guardrails in place, without us ironing out the ethical details and the copywriting issues and things like that. So my prediction is that there will be a point where these factors are going to collide and they’re what we might see is actually, a reduction in the development of these technologies as a result of that.

Filipe Santos
00:49:31 – 00:49:51

Interesting. Well, that might actually appease a lot of folks that are sick of the hype and sick of hearing about AI. That being the case, it’s very interesting how the technology might shift then in that scenario. So, Lazzarina, let’s talk a little bit about the ML for SEO platform. What and why can marketers and SEOs find this particularly impactful? How can they use it?

Lazarina Stoy
00:49:51 – 00:50:56

Thank you so much for asking about this. So first of all, it’s very early days for this platform. I’ve taken, some time until the rest of this year to hopefully create kind of like a one stop shop where organic search marketers can get bite sized tutorials, guides on several different areas of their work. What is important for me is first of all provide them with bite sized tutorials to implement APIs and to do machine learning in a way that is not intimidating. So most of our tutorials, if not all, will have a Google Colab template so that you can just execute the code as opposed to writing code from scratch, or that they will have a Google Sheets template. They will also have Looker Studio templates so that you actually are equipped with all the resources that you need to get started. Hopefully, the way that I’m designing the tutorials, courses, and so forth are to allow you to do this while you’re also doing your main role. So it’s not intimidating.

Lazarina Stoy
00:50:56 – 00:52:12

It’s not a lot of work. It’s just something that you can do and slowly learn to understand how it works and why you do it. The second thing beyond tutorials and templates is implementation guides. So what I see a lot of people in terms of my private, training practice, and that spans across individuals and also in agencies, is that they can implement a certain technology in their day to day, but they don’t know what to do after they implement it. So in the example that we gave previously, okay, you know how to transcribe videos or you know how to do entity analysis. What now? What are the different ways that you can implement this to make your strategy better? So these are the type of resources that we’ll be focusing on as well. And the third thing for me is that similarly to Aleda’s learning SEO roadmap, I think that there is a time right now where we need a similar roadmap specifically for machine learning and how to harness that for organic search. Like I mentioned, for me, organic search is not only Google search, but it’s also search on any different platform, like think YouTube, TikTok, Etsy, Shopify, you name it, so that, you know, you’re actually equipped with different bite size snippets of information that allow you to automate different tasks.

Lazarina Stoy
00:52:13 – 00:52:45

I do also want to highlight that we are open to contributions from different authors. I’m hoping to create a expert directory for people that are aware of, these technologies, how to use them. So if you are, one of those people, definitely get in touch. We already have onboarded fantastic people to the platform. Greg Bernhardt, Nasser, Jess Peck, just a ton more are, just coming. So I’m very much looking forward to seeing the content that they create for the platform as well.

Filipe Santos
00:52:45 – 00:53:20

That’s amazing. And I think that’s a incredible call out because, yeah, experts that are out there that are willing to knowledge share, they’ll help so many other people get in front of this and get comfortable and develop their own skills, which I think is critical. I think you’ve mentioned a lot about upskilling as we’ve talked. You know, how exactly then is the AI, the machine learning aspect of this in terms of automation spilling to other areas of marketing? And, you know, what kind of significance is there? And how can, like, ML for SEO kind of help people get familiar and comfortable with implementing strategies there?

Lazarina Stoy
00:53:20 – 00:54:24

I think definitely there is a spillover from like SEO into other areas of marketing just because the user journey is really becoming a lot more integrated. So the decision making process for purchasing something or a service or researcher is becoming a lot more integrated from different platforms. So, when you think of the typical marketing team, in my opinion, it should no longer be siloed. It should be integrated with SEO and PPC and content marketing and video marketing, all of them working together for a cohesive, omnipresent strategy for brands. So from that point of view, we will be, focused on creating content that addresses challenges in terms of, the day to day of all of these different types of marketers. Hopefully, that will enable them to become more productive and to really harness the power of technologies. Some of them are quite recent. Some of them have existed, like, dozens of years ago have been released.

Lazarina Stoy
00:54:24 – 00:54:46

So the idea here is to provide enough alternatives for marketers, but to also provide it in a way that is accessible for them in terms of how they can implement the technology, how to present it to stakeholders, how to do a project from a to zed, you know, harnessing the power of the technology and so forth. So hopefully, I’m answering your question.

Filipe Santos
00:54:46 – 00:55:13

You certainly are. I mean, there there’s a lot to it. So I I kind of just want folks to sink their teeth into this a little bit and figure out how it applies to them specifically. But the industry is changing and obviously with artificial intelligence, how do you envision the change in user behavior, search behavior, teams operating at organizations, and the needs that are required there? How do you feel like these behaviors will change? Because right now, we know how marketers know how users have behaved, but how is that

Lazarina Stoy
00:55:13 – 00:56:53

changing? So I think just some of the more practical changes in terms of practitioners are that they are harnessing machine learning and AI a little bit more in their day to day, which is fantastic. We’re seeing a lot companies integrating automation. Another great thing is that a lot of companies are now starting to pay more attention to the data that they have, how it can be used, how it can be protected, and if it should be, a little bit more. In terms of user behavior, we’re already seeing a lot of people replacing, or if not replacing, eventually supplementing their searches with chat gpt or with copilot. So we might be seeing an enhancement of this behavior, especially if the models become even better in the future. Now this is probably going to be something that we will start to understand a little bit better at what stage of the journey people are incorporating AI assistance, how are they using them, are they supplementing them with additional research or are they just stopping there, in which case there will be different scenarios that will be played out from a marketer point of view. Also thinking that user behavior will change in of how skeptical they are of the content that they see online or at least I hope it will become, especially if we have this wave in terms of misinformation and content, AI generated content that’s currently infiltrating all platforms like, forums or AI generated videos, shorts, TikToks, blog posts. I think that people will become a lot more selective with the sources of information where they try to find answers.

Lazarina Stoy
00:56:53 – 00:57:31

So what this might mean for online businesses is first of all, if your brand is recognized enough and has a big enough content library, then users might be flocking to your site directly to perform a particular search. And here, in my mind, I’m thinking in fields like the medical domain, people flocking to Mayo Clinic for instance to get answers to particular medically related searches. So that we might be seeing companies, especially with big brand catalogs, investing in improving their own information retrieval systems so that they could serve better answers to people using their internal search systems.

Filipe Santos
00:57:31 – 00:57:47

Okay. That makes a lot of sense. What do you think are the most crucial skills that marketers, entrepreneurs need to kind of pick up right now in order to be ready? It’s constant change, but what do you think would be your recommendation of skills that everybody needs to pick up?

Lazarina Stoy
00:57:47 – 00:58:48

Learning how to learn new things. That’s it. As simple as that. So I think it’s very interesting to consider learning as a skill, but it is a skill. It’s important to know how to quickly pick up a topic, how to learn what you need in order to really understand it, and how to be able to park it until the point where you will need the knowledge that you gained and continue learning new things. This is my biggest, tip. There isn’t anything in particular that you should be learning right now, but you should always be learning, improving the way that you do things, the technologies that you’re using, the tools that you’re testing, the way that you think about users, all of these things. So it’s very fortunate that SEO sits at the cross of marketing and technology but both of these fields are very dynamic and they require a lot more kind of output from us as practitioners in terms of the efforts that we put in to stay relevant.

Lazarina Stoy
00:58:48 – 00:58:57

So it’s important to learn how to learn and to be happy doing it. If not, I think a lot of people would be at risk of their skill set becoming obsolete.

Filipe Santos
00:58:57 – 00:59:26

Amazing. Yeah. Very pertinent. I appreciate that and it sounds like the perfect advice. So maintaining a competitive, edge through continuous learning and adaptation in AI marketing is the only way not to get left behind. That’s what Lazarina is saying here. And as a professional, is in our own hands to stay relevant and sharp. Learning about the tech, the strategies, and the ways that you hone your abilities, specifically to your business and to your role, are what make you valuable and in demand.

Filipe Santos
00:59:26 – 00:59:40

Working with folks like Lazarina can help you get there and get to the right space and build the right habits for the success. La Zarena, why don’t you mention how folks can get in touch with you, follow you, and kind of see what you’re up to and learn from you?

Lazarina Stoy
00:59:40 – 01:00:12

Yeah. So you can get in touch on Twitter or LinkedIn. My handle is la zarena stoy. On my personal website, I share different presentations that I’m delivering, blog posts, tips, tricks related to SEO and data science, and definitely on my website, machine ML for SEO, you will find all of these machine learning for SEO and for marketing tutorials and guides and templates that we’ve been discussing. Subscribe to that newsletter because we will be starting that soon. So you don’t want to get left behind.

Filipe Santos
01:00:13 – 01:00:14

Thank you so much, Lazarina.

Lazarina Stoy
01:00:15 – 01:00:16

Thank you so much for having me.

Filipe Santos
01:00:18 – 01:00:41

Thank you for tuning in to phraseology plus AI. We hope that you enjoyed today’s episode as much as we enjoyed creating it. Don’t forget to sign up for our email list at phraseological.us. Give us a raving 5 star review anywhere you get your podcasts. Thanks in advance. Remember, stay curious, stay informed, and keep listening for more. Until next time, take care and happy listening.