In this episode of The SEO Show, hosts Michael Costin and Arthur Fabik dive deep into the world of search engine optimisation with a special guest, Michal Suski from Surfer SEO. The episode kicks off with a light-hearted introduction, where Michael and Arthur share their recent bonding experience with Michal over drinks, highlighting the camaraderie that often develops in the SEO community.
As the conversation unfolds, Michal provides an insightful background on his journey into SEO, starting as a PBN copywriter in 2016 and quickly rising to become the head of SEO at his agency. He emphasises the dynamic nature of the SEO field, where rapid learning and adaptation can lead to significant results in a short time. This experience ultimately led to the creation of Surfer SEO, initially developed as an internal tool to streamline processes within the agency.
The discussion transitions to the evolution of Surfer SEO, which began as an on-page optimisation tool and has now integrated AI capabilities to enhance content generation. Michal explains how Surfer SEO analyses competitors' websites and provides actionable recommendations for optimising content. He shares exciting developments, such as the tool's ability to connect with Google Search Console to keep content updated automatically based on traffic data.
Throughout the episode, Michal addresses the current landscape of AI-generated content, emphasising that not all AI is created equal. He discusses the importance of prompt engineering and the need for a robust data foundation to produce high-quality content that meets user intent. Michal warns against the pitfalls of relying solely on AI for content creation, highlighting the necessity of human oversight to ensure accuracy and engagement.
The hosts and Michal also explore the significance of topical maps and the strategy of publishing interconnected content clusters to improve SEO performance. Michal shares a compelling case study that illustrates the potential for rapid growth when implementing a well-structured content strategy, while also cautioning against the risks of ranking for irrelevant keywords.
As the episode wraps up, Michal offers valuable insights into the future of SEO, including the importance of maintaining content freshness and the role of user engagement metrics in determining content success. He encourages listeners to leverage tools like Surfer SEO to stay ahead in the competitive landscape of search engine optimisation.
Listeners are left with a wealth of knowledge about the intersection of AI and SEO, the importance of strategic content planning, and the ongoing evolution of digital marketing practices. For those interested in further exploring these topics, Michal invites them to join the SEO Surfers Facebook group, a vibrant community of over 20,000 members dedicated to sharing insights and best practices in SEO.
00:00:00 - Introduction to The SEO Show
00:00:17 - Meet the Hosts: Michael and Arthur
00:00:39 - Interview with Michal from Surfer SEO
00:01:16 - The Evolution of Surfer SEO
00:01:29 - AI Integration in Surfer SEO
00:02:10 - Excitement vs. Fear in SEO Automation
00:02:27 - Michal's Background in SEO
00:03:34 - Building Surfer as an Internal Tool
00:04:52 - Testing Surfer with Client Projects
00:06:11 - AI Content Generation and Authority
00:06:40 - The Journey of Surfer SEO's Development
00:08:58 - AI vs. Human Content Creation
00:10:08 - The Cost of AI Content Generation
00:11:20 - Challenges with AI Prompts
00:12:38 - Surfer's Approach to Content Creation
00:14:09 - Surfer's Data-Driven Content Strategy
00:16:38 - The Importance of User Engagement Metrics
00:18:07 - Shortcomings of Basic AI Approaches
00:20:16 - User Engagement and Content Optimisation
00:22:25 - Future of AI in Content Creation
00:25:05 - Case Study: EAT Issues and Content Strategy
00:26:48 - Automated Content Refreshing
00:28:28 - Handling Freshness in AI Content
00:30:45 - User Engagement Metrics as a Ranking Factor
00:31:02 - Three Key Areas for Successful AI Campaigns
00:33:25 - Building Topical Maps for SEO
00:34:50 - Standing Out in a Competitive SEO Landscape
00:37:03 - Publishing Strategies for New and Existing Sites
00:38:28 - The Role of Links in SEO
00:39:11 - Final Thoughts on AI and SEO
00:42:29 - Where to Learn More About Surfer SEO
MICHAEL:
Hi guys, Michael here. Do you want a second opinion on your SEO? Head to theseoshow.co and hit the link in the header. We'll take a look under the hood at your SEO, your competitors and your market and tell you how you can improve. All right, let's get into the show.
INTRO: It's time for The SEO Show, where a couple of nerds talk search engine optimization so you can learn to compete in Google and grow your business online. Now here's your hosts, Michael and Arthur.
ARTHUR: Welcome to another episode of the SEO show. I'm Arthur Fabik and I'm joined by Michael Karsten. And today we had a really good interview with Michal aka Michael Suski from the SEO tool Surfer SEO.
MICHAEL: And you know how to say Michał because you bonded with me.
ARTHUR: Because I'm Polish and he's Polish and we're Polish brothers now after a little bit of a drinking session at the bar down the road. After the show. After the show. We got together. Yeah, it was good fun.
MICHAEL: We're recording this intro now a few hours after we spoke. It was a great chat. Wasn't it a good chat? Yes. What did you learn?
ARTHUR: I can't reveal the secrets. The chat after the podcast was better.
MICHAEL: That's what they say about most conferences. It's not the stuff that happens in the conference. It's the networking afterwards or around it that matters. But SEO, Surfer SEO is a great tool. Fantastic. It's been around for a while now and originally it was just like an on-page optimization tool. You know, chuck your keyword in, you're trying to rank for, it will tell you how you should optimize your page by looking at other competitors in the SERPs and telling you what you should do with your title tags and your metadata, blah blah blah, all that stuff. Now it's sort of morphed into the realms of AI and being able to optimize copy based on really complex prompts with AI that a human is probably not going to write. That's exciting and what we thought was really cool is that this tool is going to like plug into your search console and keep content updated. So if it's detecting traffic from keywords.
ARTHUR: It'll plug into your website and post content on your behalf. Updated content that you don't even have to do anything.
MICHAEL: Now is SEO, is that either terrifies you or excites you? I'm excited. I'm excited too. Can you tell? You look excited. But you know what? We're excited to bring you this chat. So enough from us. Let's throw to our chat with Michal. I called him Michael because I'm an Aussie.
ARTHUR: I called him Michal because that's his name.
MICHAEL: You know what's going on. So I apologize for getting the name wrong, but this is Michal from Surfer. Let's chat all things Surfer SEO, AI, and yeah, let's bring it. Hi, Michael. Thanks for joining us on the SEO show. Could you give us a bit of a background about yourself and what you do?
MICHAL: Hey, hey, thanks for having me. It's really a big pleasure for me to be here. So a few words about myself is like, I'm not like really old SEO in terms of like the experience, because I joined SEO in 2016 as a PBN copywriter. So it's like kind of standard career starter. It turned out I really liked it, and after a year I became head of SEO of the agency I joined, so that was pretty rapid. Very rapid, yeah. Yeah, but that's why I love SEO as a niche, because you don't have to be like, you know, 10 years old, experienced, 20 years old, like I've been doing SEO in Google when there was no Google, right? It doesn't help you much because it's a rapid environment that you can get the best within a year or two years of testing. You can actually move the needle more than anyone else on the planet, so that's why I like it. After that, out of SEO promotion, We had a bit of an issue in the agency, we couldn't scale up. It was a problem with people. It was on a village, so 300-ish people village, SEO agency, so you can imagine there's nobody in there. Like on a market, really in a proximity of any acceptable driving distance, no way to hire anybody. OK, we either scale up with people or we scale up with the processes. And we came up with the idea of building Surfer as internal tool for the agency. So we don't have to scale up with people because there are no people. So we have to choose the process. And we're forced, some sort of, to choose the process as a part of improvement. And we build it for ourselves. And yeah, I'm building it ever since. It was 2017 when we started, and it's still up and running.
MICHAEL: Awesome. Very cool. So in those early days doing SEO, you know, you started as a PBN copywriter. Arthur started as a link builder. Link builder. Of course.
ARTHUR: Similar path. Except I don't know the address.
MICHAL: Name and phone. Name and phone location. Yeah. That's it.
MICHAEL: Yeah. Yeah. Do the outreach. See what comes back. Very painful. Don't remind me. Did you do much by way of site building? You said testing. Were you building sites, launching little affiliate sites yourself, or was it largely in that agency space, like services for clients?
MICHAL: Yeah, so that was in an agency space when we started testing the tool against our clients. And it turned out it's going well. We cannot live without it anymore. And I was like, hey, maybe we can get away from this place of talking to clients on the phone and why my site is not ranking yet. We just signed yesterday, and why it's not ranking? You are familiar with these. That was like a, it was tempting to leave that space because it's a painful space, especially in Poland, where these clients don't pay like, I don't know, 10k a month, right? Even in Polish currency, they still don't pay 10k, they pay like, The majority of the clients are really low tier and we are dealing with the worst clients. Leads for SEO, you get through the SEO. So you can imagine someone's Googling SEO for my business and we are ranking and he picks up, he calls us like, I do SEO for me. So they are like the lowest tier of leads. So we just really wanted to leave that space. And then we just published the product and it turned out that it helps. everybody else. So yeah, that was really refreshing to leave like a client SEO space to like SEO tools and SaaS and shit.
MICHAEL: And so Surfer, you know, started off as, you know, on-page analysis and recommendation type tool and of late it's morphed into offering, I guess, more AI content generation. And you know, you're here in Chiang Mai this week and you're doing a talk about from AI to authority and why maybe not all AI is created equal. Which is cool. It's very topical at the moment. Maybe if you could give us a high-level overview of this concept, AI-generated content isn't created equal.
MICHAL: Before I dive into it, I will just tell you that Surfer actually started as a pure reverse engineering tool to analyze the HTML of the websites. This is how we started. The idea was that the homepage is starting to work much better than it used to. So we didn't have to like buy links, like too many links because they cost, right? And we could like do some on-page work by just checking where are these keywords on these competitors' websites? Are they longer, shorter? Do they put them in H3s, H2s? Where do we miss them? So that was basically like a nice looking chart and a little bit of a table with all the data you want to look at when you are like comparing yourself against the competitors like pure manual work and the only thing with it back in the days was scrape google and crawl websites and check the html and tell you like breakdown of what's inside each of the top 50 ranking websites. And after that, we improved into giving suggestions based on that data. And then we improved writing with that data, so the content editor stuff, the audit stuff. And now, like you said, it's AI because, well, These days we can produce content that will compete with writers, literally. I mean, I'm not talking about like highly specialized content writers. I'm talking about the general SEO content writers, which, you know, they're like, you can get better than that because of data, because of information. human brain can't process like imagine yourself i ask you to write me a piece about my writing snowboard you have ideas so what do you do you google you read you check common elements differences you make some notes to figure out like outline for yourself and with that outline you start to write and imagine how many data points you can remember how many data points you can notice while reading these articles and with the AI well if you feed it with huge amount of information it will be only the matter of the cost it will process it and it will remember it and it's brain is like much more like much bigger in terms of the capacity so these articles are like you know better research basically because I can use that along the way. So that's why it became vocal these days, because now I'm kind of, you know, getting closer to the point, but right now you can really compete with that content. So yeah, that's why we are vocal about it right now.
ARTHUR: It's almost instant as well. So the whole process of research takes a lot of time, and actually copywriting takes a lot of time. Now it can take seconds.
MICHAL: It's like 50-50, right? When you were doing research, it was half of the time of writing the article was the research, wasn't it? If you want a proper research, how can you just read 10 articles? It will take you Two hours. With understanding, with making notes, it will be three hours. With breaking them down and actually segmenting the information. So what are the common questions answered by these articles? And you receive seven questions that everybody answered, five questions that half of the competitors answered, and three questions that were unique to some competitors. And if you want to do it on your own, by hand, with whatever kind of tooling in your brain you have, It's just not possible to like analyze it so deeply and with the well it's just a matter of a couple of dollars more and you end up with something that is like on another level.
MICHAEL: So when you say a couple of dollars more for AI, because a lot of people get drawn into the temptation of AI saving. It's a cost-saving center. Just churn out the content, chuck a prompt in, you know, and hope that you get rankings.
MICHAL: This is ridiculous because people already saved on having AI. Why save on AI itself? Like, you know, you get like 10% of the cost of a writer. You already got there. And instead of utilizing that power and making it better and better, you would lean towards writing for 10 cents or writing for 5 cents a piece. It's possible. Like if you generate a simple prompt chain in GPT 3.5 turbo, you can get an article that on the surface will look good, but it won't survive. And if you want to put it as a client facing content, you will end up with a lot of labor because you will have to fix it because you cannot. sign with your name a piece of content that will contain repetitions on the same queue. AI can sometimes repeat the same fucking word ten times and if you are not reading the outputs you will publish it on a website and you will be like, oh well.
ARTHUR: One thing I found with AI is a lot of the prompts I give, it doesn't listen. So I'll tell it to write an article that's 500 words, and I'll say, insert this keyword 10 times. It won't do it. It just bypasses the prompt.
MICHAL: It can't count. It can't count, really. Seriously, you have to really do a lot of analysis and put huge effort to teach it how to count. And it turns out that There's no point in doing it this way, like writing a 500 words article. Instead, you collect the facts. You collect entities and facts that give context to these entities, and you will end up with basically AI basing the length on the facts. Cover all the facts. Don't cover half of the facts because you ran out of 500 words threshold. you have to give it free hand and inspiration in terms of like the talking points instead of forcing it into a direction like use it 10 times, use it 5 times, the total length of the lead to the article should be 200 words. This way you end up in a dead end that's kind of, you know, you know these articles because you are trying to prompt them.
ARTHUR: I know, yeah, exactly.
MICHAEL: Sometimes I've seen some where it repeats the same sentence, not even the keyword, the sentence, just, you know, maybe slight variation in words.
MICHAL: We had a breakthrough actually in the technology when the GPT-3 Instruct, when it showed up on the market. So it was slightly after Jasper released their like GPT-3, like the best out there at that time outputs. There was this GPT-3 instruct that was not finishing the sentences, but actually understanding what you want and producing the answer. So back in the days, we had these sentence finishers. And when the GPT-3 instruct was there, we had an opportunity to actually tell it what to do. And it was doing it, not just finishing, continue writing. Back in the days, it was just finishing that. And after the instruct we managed to start creating pipelines that contained more than just, you know, like a basic start with this word and continue writing. That was available back in the days. Now it's much, much more complex, which is awesome.
MICHAEL: So if we're looking at it from a surfer point of view and the way like your tool will brief the AI to create content, you know, want to write for a certain keyword and it's done the analysis with the prompts, how sort of big are they and how much is going into it? What sort of credits are being used and how is that better than let's say going to chat GPT web-based interface, checking it, you know, create me an article on Labradors.
MICHAL: I mean, you can sort of recreate that pipeline with ChatGPT because ChatGPT has this web pilot, right? So you can read the web. But how people do it, as I've seen it, they think they are data-driven when they ask ChatGPT, like, hey, read this article and write me something similar and make it SEO-optimized. So they say, like, we are data-driven because we used a benchmark from the top 10, right? That's not the way to go. They should like structure all the layers of the pipeline to gather the information from all of the pages ranking the top like five best benchmarks from the top ten. Then it should segment that information into some sort of bucket. So where are the facts? Where are the entities? Which entities are associated with these facts? What are the questions answers? How frequently they are answered? which questions are must have and which questions are unique to some sort of competitors out there. And then match them with the answers and produce like a huge, I would say, knowledge base of the SERP. So with that knowledge base of the SERP, so instead of like feeding AI with content, you feed AI with information obtained from that content. It's much easier for it to consume, because, well, it's not like a huge, fluffy, fluff content, but it's just condensed set of information. And with ChatGPT, it's just crazy time-consuming to execute these tasks one by one, because it will be you clicking and prompting, clicking and prompting. You can use an API for that, But if you don't have these custom deals, custom offers, well, it's not really reliable if you want to produce content that scale really like thousands of articles or tens of thousands articles. So that's why there's like an advantage of using someone like a third party service who will execute that pipeline, who has a team developing that pipeline like 24 seven for six months, like being passionate and crazy about it. So sometimes there's no point, you know, trying to open the open doors. They are already open and you can focus on your business and you can focus on actually making making business from that content instead of like being the best prompt engineer for getting about your business. Like you choose your battles and you don't have to choose that battle because someone else did it already and there's no point to like finding it in it, right?
MICHAEL: Yeah, put your money to higher leverage activities or your time, not your money, and use tools that have sort of their time figuring it out.
MICHAL: We and probably others like we cover it so that we've been tempted back in, back in the days when we were starting to build all the services on our own, like even Google scraper, we built a Google scraper to like, you know, allow people to search from server for the, for the SERPs and for the information from that SERPs. And it turned out that maintaining that, keeping up to date with Google's HTML that is changing. So we started to pick Google Ads results, for example. So that was our own scraper. And we had to put a huge effort in keeping it always fresh, always in line with the information from Google. And it turned out there is no point in doing it, because there is a company who is doing explicitly this. So why would you try to compete with them and dedicate the developer to that freaking task while we can ask for the API and just get it? So this is one of these automations that you can actually outsource and they will deliver better results than you are doing it yourself because you are not 100% focused on prompt engineering and writing at scale.
MICHAEL: Yeah. So with AI and not all being created equal, you know, prompt engineering obviously is a massive part of it. What other sort of, I guess, shortcomings are there at the moment that you're seeing with, you know, the basic approach of just putting it into chat GPT as opposed to something more refined like you're offering?
MICHAL: Yeah, so, I mean, it's like a short-legged approach. Like, you will end up sooner than later when Google starts to, like, well, everything happens when you start receiving traffic. So you can rank shitty content and it will rank, but when you start receiving proper traffic, it won't, like, defend itself. It will be just, you know, people read and like, wow, Pogo sticking, whatever, searching for something else because, well, it didn't satisfy them. It satisfied Google because it had like entities, it has like these SEO optimized elements, but it didn't survive the test of real users. And I run, even when we were trying these GPT-2 like three years ago, I published like 25 articles on SurferSEO blog just to test it and they were ranked like crazy. There was like even an article about best affiliate niches. That keyword, we ranked for it and like tons of people started to read that crap. And I can tell you that Google taken us down because well, it was crappy content. And even though we tried to like refine the article later, it was too late because too many negative behavioral signals were sent to the site. So, well, we couldn't defend that. You will be able to run with them, but they won't last. So that's the basically the biggest difference between like a short-legged quick and dirty GPT like stride of GPT content compared to like really properly execute a long long pipeline it's bro it's three hundred thousand words long conversation between AI and competitors content three hundred words is like a freaking book of content of text analysis and processing prior to generating your original article. So that's a huge difference. And actually you can see it in the outputs.
MICHAEL: Coming to the topic of user engagement, you with Surfer SEO and the prompt engineering, how much are you able to optimize the content that's created for that? Like have you done research or testing to identify what aspects of the content will lead to more engagement on a page that will satisfy that?
MICHAL: That's actually a good, good one. I mean, we are as good as the Serb plus a little bit of, you know, a secret sauce, let's say. But in general, it relies on the Serb. So if you like get the shitty Serb without content, it will turn out a little bit more generic. But if you are competing with like, you know, say marketing automation tool, you write an article about that, and you have HubSpot, you have big brands in their MailChimp or whatever intercom, their articles will be really well written for engagement. And if these are the sources, you will end up with mimicking their approach. If the SERP is shitty, you have to provide your unique angle, your approach to writing a PSA. You can use a template for a roundup review or a single product review, but there's a little custom prompt on top before you create the article. So if you want to make it a lead generation or in the form of a lead magnet or whatever, you just provide that information and influence the outline. Okay.
MICHAEL: And do you see it largely with this sort of content generation being a marrying between the tool and then the person using it and being able to be on top of that for the foreseeable future? Or do you think, you know, in a year, two years, we're at the point where it's able to really handle all of that?
MICHAL: So I'm an SEO, and I lean towards automating everything. So you've got the topical map, and you see the gaps, so you just fill them. And well, here are my gaps. Fill them. And the gaps are filled, and I'm not afraid of publishing shit, even on the client websites. That's the goal and that I have on mind with creating the whole user journey with it. Of course, it's not ready yet. And of course, we are not there yet, you said, a year or two. Definitely possible. And we will build it sooner than later. But that's not the only part, like generation is just a first step. And well, I would love to elaborate on that because I got one case study that I have to trim off the presentation because I got only 20 minutes. If you want something like from under the table, I can share because that's a huge part. Let's do it. Yeah. Yeah. Okay. So there was a situation, we had a website with Dean Scaduto, that's a New York-based SEO, and we teamed up on some testing and building topical clusters. And we had a website, he had a website, that was suffering from EAT issues, and it was kind of flat after, like, it was good, and then, you know, flattened down. And we said, like, okay, let's try to test it and build some trust, like, site-wide, to publish, like, 100, 200 articles that were, like, big. SEO optimized, interlinked, proper filling the topical map that we built and so on. And it turned out that we got traction, like hockey stick kind of growth. But like two weeks ago, I was like, fuck, the site is getting hammered. Like, what the hell is going on? It's not all the traffic, but like a quarter of our growth. So like, goddamn, I lost the case study, right? Because something has happened. But we started to look into data and it turned out that we ranked for too much. Google appreciate our content production so much that it ranked us on the keywords we didn't deserve to be ranking. So that's a huge, huge thing because when you publish a ton of good quality content, you can be surprised of what you will actually end up ranking for because if there's like a right domain and big portion of content that is relevant and semantically connected and so on it turns out that we ranked for these keywords I would like Why the hell are we ranking for this? I mean, where are we? Because these were the keywords that lost all of the traffic. So the example is this. We ranked for a keyword, how to quit smoking weed. That was the keyword. And the article was benefits of not smoking weed. Right. Totally different. I mean, related, but totally different. When I looked in the SERP, all the SERPs were like, how to, how to, how to, how to. None of them were benefit of not smoking. And even though Google ranked us, I didn't even mention it in the article. Google ranked us, and we lost that keyword because we didn't have an answer. So if you are able to proactively look into your Google Search Console after a successful AI campaign, you look into Google Search Console and see, Oh, we ranked for it. We didn't mean it, but I will take it. And you write the piece that will actually answer that question, or you write another H2 that covers it. Then you can be like a, it will be like a AI optimization, like insurance policy that your article will be always up to date. And with that case study, it proved to me that we have to build it. Like, you order the AI article that is ranking, nice, cool, and at Surfer, we've got you covered, so the article will be refreshed every, I don't know, we'll just test it, but probably every month, right? To check, like, what is this ranking for? If there's anything missing, if Google thinks it's good for that, fine, we'll provide the answer right away. So that's kind of future of of AI, actually, to keep it up to date and keep it like… You need to be the first on the market to cover as many of these angles as possible. And with this auto-optimization with different sources, you can be always kind of the freshest content out there. Just not the, like, fix the date. Like, I just republished the article, Google pick me up. No, you need to add some extra value. Even with this recent update, right? There were some rumors, people said, like, The freshness is a thing, but a real freshness. So not the freshness of republishing it in WordPress, but the freshness in terms of the content it contains. So if you refresh the article and not change the content, you're screwed. But if you change the article, republish and add extra value that wasn't there before, that's the way to go. So that's kind of the outcome of the case study and of the ideation of the new products. That's cool.
MICHAEL: So you plug in outgoing?
ARTHUR: So it looks at the rankings and then it just, every month we will refresh based on whatever keywords that you're starting to rank for. New content updated, then you just publish it in order just to build more traffic.
MICHAL: Yeah. Like automatically published. I see it like you are either trusting it enough, like I would, but like you just like accept all by default and they will be already like refreshed on WordPress so you don't touch it. Oh wow. Yeah. That's amazing. If you are like a client facing and really strict to the policies of updated content, it will be like a checklist like, oh yeah, here are the changes. So yes, yes, no, yes, yes, yes, no, whatever, and publish. So it will be like two ways to do it, but that's how I see it. That's awesome. Yeah, I know. I'm afraid of it. A year ago, we introduced a little feature in there that was just analyzing the Google Search Console data and the keywords that you ranked but you didn't use. So like a one point of what we are talking about. And we run a bunch of tests on it on like really like a finance website. What are like tough niches and the guy with the only thing that he did it just went through the article and included baked in these keywords we suggested that you don't mention and he received like six hundred percent organic traffic overnight. well, right after that. So right now it wouldn't work probably like as easy because like cramming only the keywords, it's quite not enough for like right now to make the content fresh, like to be eligible to like republish the article. I just crammed some keywords and Google now take it, right? It's quite too cheap, but with adding some paragraphs, adding some like headers in there that will make a little bit of a difference to the content, that will be, wow, I can't wait to see it working.
MICHAEL: How do you deal with, or can you deal with freshness? Like these LLMs all have like a training cutoff date.
MICHAL: That's why we don't care about it because we have a lot of data fed into it first. So that's the problem that, you know, that's actually, I have a slide on it. Like the real perceived limitation of GPT is data cutoff point. But the reality is that it's missing optimization data. Because you can give it all the facts you want, depending on the source. We are SEOs, so we use sources from Google because we want to please Google in the first place. And with that, you don't need GPT to know who's the president of the United States right now. You don't care. Because if it's important, it will show up in the SERP. And we will pick it up from the SERP as a fact that the president is Joe Biden. We will know it. And if it will make sense to write about Joe Biden in this article, we will know it and we will write it. Not because GPT knows who it is, but Google knows who it is.
MICHAEL: So with this, I imagine it's like we spoke about user engagement metrics before people pogo sticking out of the results. So if they land on something because they've searched query that Google's rewarding you with traffic, it's testing. This is combining AI with those human elements, isn't it? And it's like.
MICHAL: You're just a source of ideas right now, because I can imagine something that's not only checking your ranked keywords or not ranked keywords, but also time spent on the website. And you monitor all of your 100 AI articles, and you see, OK, some are starting to get traffic, some are getting rewards. And of course, you optimize them every month. There is three articles whose time on page is like five seconds and you can use that information to rewrite them or like optimize for longer time on the page to like keep them working for you because these articles will be dragging you down. because people are not satisfied. So you need to take action as soon as possible on them, either delete or rebuild them because they are not… The user intent is different, probably. Probably it's different. So you want to work on them. So that's another thing that would be awesome to have. Yeah, paired with user experience metrics. Wow.
MICHAEL: It's cool. It's very cool. What else, I guess, you know, your talk is about shortfalls of AI. So there's a prompting and like real-time data. Is there any sort of other shortfalls around at the moment that you guys might be working on or sort of areas that you've tested that you're looking to bring to the tool that you can talk about at the moment?
MICHAL: Yeah, sure. There are three areas to actually have a successful AI campaign, and you mentioned the data. That is a must-have if you want to build to last. But besides that, you need a proper plan. This is obvious shit, but it's really with these days when the content production is unlimited, you really need to dive deep in the niche and cover every threatened topic and build an army of supporting URLs, sending strong internal links that are semantically put in there with proper anchors, creating a perfect chain of going through the topic. And these days you just have to have it. And that's why we've been building this clustering method and we created this cluster map 150 close to 200 million keywords clustered at once and then some semantics on top of that to just give you that map. And with that in place, if you use that map and execute with the AI, you will receive stuff I showed on the stage. You will see these like tenfold growth on a huge website. with just in this. So the strategy, that's obvious, but you need to have a plan. You can't research by the search volume, right? Right now we need to cover everything. No matter it's like 40 people a month search it, you write it and you send a strong internal link to boost your money page. That's the key right now. Plus you need the pace. That's the third one. So plan data for optimizing these articles and making them like mirroring current Google algorithm. And the third one is a scale, is pace, because Well, the ChatGPT content, you cannot scale it up enough to publish hundreds of articles every week. You can't because you will have to, quite a lot of clicking in there will be. So you need, it's hard to actually merge it all. So to have a great plan, then execute it in the best large language models possible with the biggest amount of information going through the pipeline, and then still make it with a high pace, so publishing hundreds. If you can make it happen, that will be a success that I will approve.
MICHAEL: Awesome. So the topic of topical maps is interesting, you know, like old school SEO, do some keyword research, map the keywords to a few pages, maybe create some pages for gaps.
MICHAL: It's not enough anymore, right? Actually, the main message of the press would be to publish the whole neighborhoods of content, not articles. Like we've been going through this journey from like people focusing on the keywords. So I got the keywords to run. So that's my keyword. I will run. So I write an article about that keyword. Then we had clusters. So this is not a lonely keyword. This keyword is surrounded by dozens or hundreds of other keywords that I need to satisfy the intent within that area. That was a cluster. But now we've got groups of these clusters, so groups of articles that are connected to each other. And right now you publish by the group, basically. That's the message. You should publish the whole whole set of interlinked articles like a little spider net with a spider in the middle that will be your money page so you are not like one by one or like cherry picking but if you are required to write it you write it and then it brings you it brings you value so that that should be the approach like right now that would be you'll be the fastest on the on the on the in the SERP on the planet basically if you do it this way in a year or two it will be a standard I guess
MICHAEL: And do you, some people might try to do this manually, you know, compiling all different data sources, building their map out. That's fine. Or using AI?
MICHAL: That's tricky to use AI on the keyword research, because if you do not cross-reference the information from GPT with some sort of real databases of keywords, Hallucination is a thing in here and we've been testing some tools and we tried like we asked it to make a topical map for the keyword best head shavers like affiliate keyword with like I know shaving your head bald and stuff and it created a silo of Reviews of the shavers, fine, awesome. But some of these shavers were like not existing Gillette models. Like, you know, like Gillette, whatever, whatever. It's junk. Made up? Yes, exactly. Made up, made up name. So they were faster than Gillette to, you know, create these. So if you want to truly like, because it's, it's all about if you want to read it after it's created and tweak and spend a day or two, because you cannot trust it right off the bat. That's like one approach and the other is like you create it in a way that you are 100% sure that it's true and you can like trust it and you can invest money in content based on that map without really doubting in its quality, that's the goal we have. So we just trust it. We made the math, we made the semantics, and that's the landscape of your situation. That's your situation. This is the environment you are in, and you have to trust. This is what people search. I know it's ridiculous, but this is half of your audience looking for these stupid things. And if you wanna be in a shape that Google sees you, well, that's the shape. If you like it or not, that's the situation. You wanna be the best, you cover them all. So that's kinda hard to believe in some situations like, why the hell should I write about it? And then you got this second thoughts. Yeah, maybe. Maybe it's worth it. Maybe people search for it. This is the way to actually build it. Not overlook any of the potential, sometimes ridiculous, but any of the potential topic to build another soldier in your interlinking army.
MICHAEL: Yeah. I like that way of looking at it. Soldier in the army. You love analogies.
MICHAL: I love analogies. I'm a master of this thing.
MICHAEL: You need them in the SEO world. Oh, for sure. You spoke about Touched on Pace before. Let's say like you've got a website, either in, let's say it's a new website. Are you a fan of building your topical map, creating all the content and getting it on there so that Google comes and finds it in one hit? Or do you think, let's say it's an existing site and you've got to build it up over time and you've got half a built site, like what are you seeing?
MICHAL: I build as fast as I can. Yeah, as fast as I can. As fast as I can publish, as fast as I can interlink them, I build it. If it will be a new website, probably like the first impression, this is important stuff. So if you are building like a fresh website, it will be good to kickstart it, like, you know, wait a week longer and publish it like in a sort of complete version. But don't wait like a year to like, you know, really make it perfect. Publish all the content that once you have that's that's a good approach and when you have existing website and you want to build it. I would say publish neighborhood by neighborhood like yeah basically a group of interlinked articles with the money page on top and you publish that family and that's that's how we do it like family by family ten articles dozen articles fifteen articles twenty articles like this is the average size of a group.
MICHAEL: Awesome. So a question that we had at mind at the start is, if everyone's doing this, how do you stand out?
MICHAL: You just have to be quicker than them. And I would say it's about custom sources as well. So not really rely only on the top 10, because you will be always the same set of sources. So it's really important to put Some effort into let's say finding a pdf about your. Whatever you are writing about or find a different source of different website like wikipedia page or something that is not in the serb but it is really relevant to your content and you can add to the mix something unique. And that's really crucial to make these articles stand out. And when you are talking about how do they last in the SERP, I would say the automation is the thing that can help you stand out. Because if anybody in the top 10 will publish a new fact about whatever, you will be able to repeat it in the next check. So you will always be up to date and you will never be lagging behind. Anybody from the SERP does it, you're the second fastest. So if you follow everybody in the SERP, so top one added a fact and you got it like after two weeks. then some other page added a fact. The first didn't add it, so you have two different facts added. But in a second phase, you got both of them. So you collect the information, like anybody's updating, you collect it. And this way you are always the compilation of all the freshest facts in your environment. That'll be the way to stand out. If everybody does it, I don't know. I don't know. I don't know. Everybody will be chasing each other.
MICHAEL: I guess things like links as well would play a factor of course domain authority.
MICHAL: Yeah, we never like made them not worth it. Like links are the top ranking factors that like John Miller will tell you or whoever will tell you no. Well, I mean, I can see it like there's no links in the mix of top ranking factors.
MICHAEL: No, 100% always since day one and for the foreseeable future in our opinion.
MICHAL: still probably the hardest thing to kind of, you know, for them, like really authoritative links can be a really source of information about the website. You receive links from quality websites, like organic, some sort of, yeah, it still happens sometimes, but it's a great, for Google to understand that you are real, like links will be always like the hardest one to game, seriously, because you can game the on page, easy. you just like check what's in there and you write it. And there are like specific ingredients in the mix. You use them and it's done. But in links, you can always trace, you can always find some other source of backlink and well, it will be, it will be always like super important. I can't see it like SEO without links.
MICHAEL: Yeah. What do you think about the concept of user engagement metrics, you know, maybe being a substitute for links, you know,
MICHAL: But they still have to rank you. Yes, they got to put you there. Yeah. And you can get yourself in there without backlinks, but it will take you a year or two years. So why really bother? Okay, if you can afford spending thousands on content and waiting for it to pop up while you can buy a few links here and there and boost the crawl budget, boost the authority and get them ranked and work for you earlier, why the hell no? I mean, it will be a waste of money to just put content on the website that is super optimized, really well researched and shit and no links because I don't like links. I mean, no. Why? Why? It's a stupid business decision. So, yeah.
MICHAEL: Well, awesome. Is there anything we haven't asked you about your talk at the conference that you think would benefit our audience?
MICHAL: I'm glad I shared this one from that study that we overranked it, so that's kind of something special. Good moment to have. Yeah, but if you execute it in the right way, the process I shared with you, so the plan and then the best LLM execution and then like scaling it up, you will find these problems. And people are overlooking that problem because it's a nice problem to have, so they wait till the problem appears, like we did. We published 200 articles and we're lightweight in the chairs like, oh, wow, what a case study. Man, we did a good job. But we should have checked earlier. Like, Why do we rank on these keywords? We should have an answer. Do we have it? No. Let's write it." And it could be automated. Imagine AI reading your article and comparing it against the keywords you ranked, and it can tell you, like, hey, bro, we've got a problem here. We need to write another paragraph because, well, I found there's no answer in our article. That's awesome. Yeah. Exactly. And if you can do it proactively before you drop, then you are home. We didn't.
MICHAEL: That's what I want to ask you about. I don't know if you've got any numbers data, but like, is there a sweet spot, you know, if an article goes up and it's getting this traffic, the let's call it the fake traffic, the false traffic, the, you know, it's making you feel good, but it's going to disappear. Is it one week, two weeks, four weeks for that user engagement metrics to sort of have an impact and Google to, Or is it, you know, how long does it take to paste a string?
MICHAL: I don't have data on it. I remember like some Google Search Console charts in my head, like, you know, seeing the pictures in my head that was like some of the hardest article lasted for like two months with that traffic that they didn't deserve. so the window was pretty long they were still ranking but the earlier you can discover that situation that you ranked on something you didn't really answer the better because you can react and by the next crawl actually that could be like related to the crawl budget as well like they are crawling you again and see like okay, we still can't see the best answer, but it's a so authority website that they must have it, right? So they will give you another chance, another chance, another chance. And if you miss these opportunities to actually satisfy them with the improvements, you will be taken down depending on how bad are these metrics on these keywords. So as fast as you can. I would love to have it daily, by the way.
MICHAEL: The concept makes a ton of sense. And I love the idea of having like surfer there as your bodyguard. It's your traffic bodyguard, you know, keeping on top of it. And then it's also an author.
MICHAL: I cannot imagine it not working really. I mean, it must work. It must work.
MICHAEL: Well, this has been awesome. For people that want to go learn more about you or the software, where should they go if they want to check it out?
MICHAL: I feel like the SEO Surfers Facebook group is the community to go in. There's like 20,000 people in there, super helpful. Yeah, that's the place.
MICHAEL: Great. Well, thanks for coming on the show, Michael.
MICHAL: It's been great chatting. Great honor for me. Thank you.