Edward Sturm: a common thing that you see is people posting to social media saying, "Fire your SEO agency. You can do SEO entirely 100% using Jev, using Claude, using ChatGPT," whatever. You see them saying this nonstop. And so this episode is what you should and should not actually use AI for in SEO from SEOs who are experienced enough to know how to spot good advice from bad. And I'm joined by Harpreet Singh. Harpreet is a legend in SEO. Thank you for coming back to the show. Harpreet, what are you showing us? Harpreet: It's great to be back. Um, so what I'm showing you is how we can use AI in SEO without ruining, um, without ruining our, ruining our website. So as you said, every time I log into LinkedIn, I go to Instagram, I go to TikTok, no matter where, there's always videos around, you know, fire your SEO agency, did this, and then there's like a super fast video, super fast screenshot of just Claude like working in the background doing some SEO. But the reason, you know, I wanted to talk to you about this was lot of the recommendations these AI tools give you, they're not for businesses, and they don't have your nuance. So I just have a simple prompt here. Um, I had the same prompt, so give me SEO and AEO, GEO recommendations for this page. Prioritize them based on impact. So what I did was I put this prompt into ChatGPT, into Claude, into Google Gemini, and um, I just wanted to show you how the responses differ between the three LLMs, but also the recommendations it prioritize might actually not move the needle at all. So if you decide to use LLMs for SEO, you just have to be careful. At the end of the day, you still need someone who knows what they're doing to take those recommendations and put them into practice. Now, for someone like me who does this all day long, you know, I can use LLMs to speed up my workflow, and I kind of have an idea, or I have an idea of what I'm gonna take and use and what I'm not gonna use. But if you're a-- if you're new to SEO, if you're a beginner, or if you're just getting started in SEO, the way LLMs present their findings, you might think schema markup is a ten out of ten priority, whereas for your business you might be better off doing something else. So that's what we're gonna look at. That's what we're gonna look at today. So I just took a page from Nike's website, their men's running shoes page. I put the same page through three different LLMs. We're just gonna see what has the best answer, where the answers overlap, and really what is AI saying, um, what is AI saying about this? Edward Sturm: Oui Harpreet: So as I go down the page, you know, Chat-- This is, this is ChatGPT's answer. So ChatGPT says that this Nike page has a solid technical SEO foundation, but its opportunity is to become a better shopping decision resource, not just a product, not just a product listing. Okay, that's, um, that's great. But let's just look at the prioritized recommendations. Let's go through them now. So add a model comparison and running shoe selection guide. SEO impact is shown as very high. AEO impact very high. Effort is medium. Make category and intent-specific navigation more prominent. What does that mean? Validate crawlability, rendering, and indexation of products. Again, SEO impact very high with an asterisk. Again, what does that mean? Add concise question-led FAQ. So as you can see, it's given a whole bunch of stuff we can do. But as you can see here, improve the title tag, meta description, and introductory copy. Also an SEO impact of, of high. Now, how high the impact of the meta description is gonna be, um, you know, that, that's up for debate. Title tag, yes, it can have an impact. Introductory copy, yes, it can have an impact. But does meta description, for example, need to be grouped there? Um, maybe, maybe not. But let's go down and actually go through the, um-- Let's go down and go through some of the, some of the recommendations. So the first one, add product comparison and shoe selection guide. So ChatGPT found that on this page we've got descriptive selections like move with confidence, soft landings, responsive strides, strides, fly through the distance. So we spoke about this before in another video where Nike use a lot of branding language, which I think here ChatGPT did pick that up quite well. But again, they don't help consumers compare individual shoe products or help them make a buying decision. So in this case- ChatGPT just has a simple recommendation of find the right Nike running shoes and, um, you know, have a product grid with links to relevant collection and product models. Now, from a UX point of view, I actually think this is a good idea. How much impact is this gonna have on SEO, AEO, GEO? I think that is something that, you know, we don't know. I don't know if this is a specific SEO recommendation. Um, but for UX and for, for, um, for customers browsing the website, I think adding this kind of comparison would be, would be, um, would be good for, would be good for users. Um, the next thing that ChatGPT said, surface intent-specific category links. So it said Nike already has crawlable links to other, other relevant categories. However, many of them are filters and not prominent editorial discovery paths. So now ChatGPT is saying, again, introduce a visual prominent shop by running goal section and link to these other URLs. And I think this is a good recommendation because, again, of the previous episodes we were comparing Nike, Adidas, and New Balance, and New Balance had this specific section. And at the time, New Balance were outranking Nike, and they were outranking Adidas. So this wasn't the reason, but this is one of the things, um, New Balance did well. They internally linked really well from the page. So this is a recommendation, um, this is a recommendation that I would, um, I would agree with. Speaker: This method of marketing is so effective I had to make sure it wasn't against Google's rules before I kept using it. It's a form of SEO I call compact keywords. Whereas most SEO focuses on putting up articles to answer questions, how, what, when, compact keywords focuses on putting up dozens of pages that sell to searchers who are actually looking to buy. These pages rank on Google and convert so much better than normal that when I discovered this years ago, I couldn't believe this was allowed. It's less work too. The average compact keywords landing page is only four hundred and fifteen words. Compact Keywords is a thirteen hour deep course on getting sales with SEO. A customer said, "Compact Keywords contributed to a four thousand dollar sale within the first six weeks." Another customer said, "Give it to a junior employee, have them follow it exactly as Edwards laid out. You don't have to do anything, and you're going to gain a six-figure SEO level employee just by having them go through this course." Compact Keywords is about setting up an SEO funnel that brings you sales for years and years and years. It works with AI, it's less work than traditional SEO, and it makes way more money. You can get it now at compactkeywords.com. to the podcast Harpreet: The next one on the list is verify search engine and AI crawl accessibility. So in, in this case, ChatGPT is saying the saved HTML contains product titles, price links, and images in the document, which is positive, and it contains an index, follow, robots, and self-referencing canonical. Okay, great. But then it wants me to validate all of the following on the, on the live, on the live website. So in this case, ChatGPT is not actually giving us to do. It's-- What it is giving us to do is telling us to go and validate because it can't do something himself, something yourself. So if you are new to SEO and you upload one of your pages, to do this step, you're gonna need to talk to someone who knows what they're doing anyway, or you're gonna need to do more research in this. So actually, in this case, I don't like this recommendation because it doesn't take into account, uh, it doesn't take into account the person asking for recommendations in the first place. It doesn't take into account their, their, um, their skill level. But as we scroll down, and this is one I Edward Sturm: also I don't even understand why, like if, i- if everything's in, in the HTML, why do you have to go and validate things within Google Search Console? Like, if, if it's all right there in the HTML, you shouldn't even need to do that unless there's JavaScript that... But, but even then, that's code and, and, uh, so it should, it should be like if, if it's JavaScript, if there, if there is too much JavaScript, ChatGPT should just say right there, "You add too much JavaScript." Like, Harpreet: Yeah, and when we get to some of the-- When we get to Claude and we get to Gemini, you'll see that some of them actually do that. Um, in this case, ChatGPT didn't, um, did- didn't, didn't do Edward Sturm: and that's an area where you're gonna go and like s- and like, okay, I did, I, I was doing something last night with my site and with Zapier. I thought it would take me 30 minutes to do. I was, like, having... I, I was gonna add in an, a extra field into a contact form. I added it in. I decided it wasn't a good idea. I went back, and then I needed to figure out how to use, uh, how to... And then I, I wanted to add in an AI step within Zapier to automatically sort these emails and automatically respond to them using AI. And so th- it was this thing that I literally thought would take me 30 minutes, and it took me three and a half hours, and I was working until 2:00 AM doing this. And the, the point of sharing this story is ChatGPT tells you g- to go down a rabbit hole with something that, that could seem like a very innocuous check. But if you're inexperienced, you're gonna spend four, six, eight hours understanding what different things mean, saying, "Is this a problem? I came across this." And your ti- and then by the end of it, you're gonna be too exhausted to actually do the things that actually make a difference Harpreet: Yeah, exactly. And that, that's the whole point of this video, just to make people aware of, yeah, you can go down rabbit holes with AI and actually not end up achieving anything when the whole point of you using AI in the first place was to either speed up your workflow, to get efficiencies, you know, just, yeah, get things done, get things done faster. And this next point, um, FAQs, thing. So ChatGPT thinks FAQs for, for Nike are a high priority, um, high priority piece of improvement. Now, before I go through the answer, why, why do you think ChatGPT is saying this? It's because AI is trained on years of content across the internet, books, and whatnot. And for many, many years, um, SEOs have been saying FAQs are important for SEO. Um, even before AI, people used to say, "We need to add FAQs on the page." Um, you know, before... I'm going back a few years now. Google used to support FAQ schema. Everybody had FAQ schema on their, on their, on, on their, on their pages, and Google used to show FAQs in the search results. So there's so much content that talks about the importance of FAQ from an SEO point of view that is old. But now let's move forward to 2022 and beyond, so just recently over the past few years, FAQs have been seen as something very important for AI search, um, the same thing with FAQ schema. You can go on the top blogs right now, the top SEO news company websites, you can go on AEO vendor websites. FAQs, they're, they're everywhere. And for this reason, ChatGPT thinks adding FAQs that target purchase-intent questions are a high priority improvement. Um, and they've got, you know, things like these, "What are-- Which Nike shoes are best for long resting, uh, long distance running?" "What's the difference between Nike Pegasus, uh, Vomero, and Structure?" and so forth. Now, I think adding FAQs to improve user experience is a, is a good thing. That's not a bad thing, especially because they have so much branded content on the website. Someone might not know the difference between Pegasus and Vomero. Like, I don't know the difference, so it would be nice to know. But is this a high priority SEO item? No. I can tell you as someone with experience in SEO, there are so many other things that should be fixed on the Nike website that should take priority over this, that where if I prioritize this over something that's actually gonna make a difference, you're just prioritizing the wrong place and, and, and, and, and wasting time. So this ties exactly what to, to what you just, um, to what you just said The good thing is, in this answer, that ChatGPT does say here that don't assume FAQ schema will generate Google FAQ rich results for a commercial retail query. But at the same time, I think something better would be... Now, if you're a beginner, again, reading this, what do you mean don't assume this FAQ schema will generate a Google FAQ rich result? Now maybe you're gonna go down the rabbit hole into thinking, okay, what will? But we know that, you know, Google has depreciated those, Edward Sturm: Yes. Harpreet: so only an SEO knows that. But if you're new to, if you're new to SEO and you're using these type of prompts in, in ChatGPT, again, this is another example where you can legit highlight it and I can ask ChatGPT, and I can go down another rabbit hole of okay, what, what, um, what will? And, um, I'm just gonna quickly go down the list now. The next one we've got improve structured data. Again, like, how much of a priority is this over some, some other SEO items which we'll see in, in Claude and Gemini? Like, you know, the-- it's out there, optimized title tag and meta description. I'd say title tag is more important than the, than the meta description and, um, yeah. So as you can... Edward Sturm: I'm surprised... And title tag was below the FAQ? Harpreet: Yeah, TikTok Edward Sturm: That's crazy. Harpreet: Yeah Edward Sturm: crazy. That should be, that should be number one. It should... I Harpreet: And Edward Sturm: I don't know if it was good what they have. I mean, it's... What is it? What is their current... Their, it's Harpreet: I'll show you. So r- this is... Remember, okay, this is the Canadian page. The title tag is Men's Running Shoes and Trainers. Now, I'm not gonna give something away until we get to Claude, Edward Sturm: Okay, okay, okay Harpreet: you will see what Claude, you will see what Claude picked up. But one of the things ChatGPT has not done yet, what about the H1? referenced the H1 yet, and title tag is all the way sixth. If you're Edward Sturm: Yeah Harpreet: right, do make title tag H1, your H2s, make those the first things you work on because they're so easy to Edward Sturm: And those are the first things that users see. Those are literally the first thing that a searcher sees is your page title in the SERP. Or maybe if Google picks out the H1, it might be the H1, but these, those should be all the way at the top Harpreet: yeah, exa- exactly. Ex- exactly. And then when you scroll down to, okay, the last one is, uh, correct or validate Canadian localization, which is very f- hreflang again. If you are a business owner, and, you know, many businesses do sell into multiple markets, hreflang, it can be very, um, it can be very complicated, and it can be, it can be, um, overwhelming. In this case, it Edward Sturm: Overwhelming. Overwhelming is a Harpreet: Yeah. Edward Sturm: great word, yeah Harpreet: And even as SEOs, w- okay, one of the reasons not to, um, uh, dump on SEO conferences, but go through all of the talks that happened in the last two years at the major SEO conferences. find so few talks just talking about international SEO. It's because it's complicated. It's real proper technical SEO. Um, one of the, um, one of the disadvantages of ChatGPT search right now, and I've seen other SEOs complain about this, is that ChatGPT, um, just gets localization wrong. Um, people see that don't belong in that country they're searching from show up in their country, and a part of that is because of, um, is because of international SEO, where ChatGPT is not respecting hreflang or ChatGPT is just pulling up the wrong site. So I think moving forward, this is gonna be very, um, very important. But again, you see how this one is number seven, so this one actually in my opinion should be higher 'cause the language is set to, um, en-GB in the back end of the Nike page. So this is the Canadian website, but the language is, uh, en-GB. And I said we'll mention this in Claude, but in the page title, um, Canadians don't use the word trainers. This is more of a UK thing, right? Canad- it's men's running shoes, and I think the term in Canada is sneakers. So in this case, even the title tag in itself is not using the that Canadians, um, it's not using the language Edward Sturm: Wow. Harpreet: So Edward Sturm: Wow Harpreet: if you have a website and you're selling into different markets, what l- what language do people use? So for example, in England, they don't use-- they don't really use the term sneakers, they use trainers. So you'll find that this is-- I'm just talking about normal like Air Force 1, uh, Air Max, like those type of pages. They use trainers, so the pages should be optimized for that. But whereas the Canadian version, they use sneakers, so you need to use the, the Canadian version, or you need to use the US, um, you need to use the US terminology. So again, so I think this number seven should have actually been a lot, lot, lot higher. Um, so as we scroll down, then we get to some medium priorities. And again, you could argue that some of these medium priorities, maybe they could be in higher priority, maybe they could be in a lower priority. Connect editorial content to commercial cra- categories. So Nike's long form category copy discusses so and so, but they don't inc- uh, they don't appear to include contextual links to matching categories. So we've actually spoke about this before. So when we scroll down the page Uh, we scroll all the way down the page. We keep going, we keep going. You see that Nike have this text at the bottom of the page. When we did the pod a few weeks ago, Adidas and New Balance, they had contextual-- Actually, Adidas had a big block that linked to other pages, but New Balance did this well, where within these copies, just had links to other pages. And as you can see here on the Adidas page, it's just a block of, it's just blocks of text. And again, I'd actually argue that based on some of the, um, based on some of the recommendations we've seen in the priority section, this could have been, um, this could have been bumped up. This could have been, um, this could have been bumped up. Number nine, adding verifiable expertise and evidence. So it says, you know, the page includes broad performance statements, but and I, and I agree with that. There's lots of broad performance statements on the page, but there's no supporting fa- there's no supporting facts, there's no supporting figures. So, you know, maybe we could add some of that. And then the last medium priority is just improving image and page experience signals. So it says the product images, they already have descriptive alt attributes, which is good. However, the Open Graph, Twitter meta image, and they used a generic Nike icon rather than category-relevant image. Not, not a big deal. It's not really for SEO. Edward Sturm: You know what it's- do you know what it's doing? You know what it's doing? It's just writing all this stuff that's gonna be overwhelming. Like, if you were... It, it should be, it should be, like, I don't know. This, this should be no more than, like, five brief points. Harpreet: Yeah. X. Yeah. And Edward Sturm: that's it. Harpreet: yeah Edward Sturm: the, uh, like, SEO, here's something that people don't realize. SEO, like everything in life, falls by the Pareto pr- by the Pareto principle. 20% of things that you do make 80% of the difference. And, and, and, and so this is giving you, like, everything, where, yeah, if you're... Dude, if, if you're just starting, you might be overwhelmed and quit. You know what else I'm gonna do? I was thinking, like, how do we, how do we make this episode, like, more useful to people who are using AI? I'm gonna take the transcript from this episode after, and I'm just gonna host it on my server, and it's gonna be linked to in the description for this episode. And so you can take this transcript and say, "Don't think like a stupid generic AI. Think like Harpreet and Edward on this episode, and here's a transcript for this episode." And we're gonna have that Harpreet: That's, um, that's fantastic. And, you know, even when you see this recommended implementa- implementation roadmap, which is a lot smaller than all of the recommendations we just saw, you're still gonna get overwhelmed. You're still gonna end up wasting your time 'cause you need to figure out, like... You need Edward Sturm: And it's funny that it says phase one, it says phase one is update metadata, but that didn't say that, it, it didn't say that w- in, in the directions above Harpreet: Yeah. E- exactly. And then when you scroll down, it's like, okay, how do you measure success? Again, you know, this is a really good one for this episode 'cause, you know, how do we measure success in SEO now that clicks are going down, there's more zero-click answers? We're saying impressions and quick clicks for queries such as men's running shoes Canada. Um, yes. Yes, yes and no. Impressions I think are bound to go up. Clicks, I don't know if clicks directly to the page will, will go up or not. But Edward Sturm: I actually think all, I think all of that is nonsense. I'm making an e- an episode on this. It's gonna come out, Harpreet: Mm-hmm. Edward Sturm: It's, it's about... 'Cause we, we just made our own rank tracker for the tool that I'm funding. Uh, and, and w- we thought, like, a lot about KPIs, and honestly, th- it's the s- the most important KPI is your rankings for the keywords that you are actively targeting. It's th- it's that simple. It's the k- it's the rankings because you, if you're targeting keywords, it should be for a good reason. The keyword is valuable to you in some way, and you want to be able to see that you are ranking well for the keywords that you care about, and that you can then group keywords by, by maybe the subfolder, by maybe the intent, by something, something that is valuable to you. And then maybe you also see the estimated monthly search volumes. Um, and then you could see the clicks just to know that the monthly search volume is, Harpreet: Yep. Edward Sturm: the right track. But... And that's it. And you don't really need anything else other than that Harpreet: No, exactly. And you know what I don't like from this, uh, measurement success? So we gave ChatGPT this page, and we ask it for optimizations for this page. Now what I did was I turned off my memory so it doesn't have, like, my history. Like, n- this does not take into account my history with the AI tools. So m- uh, memory is off. It's in private. It's in-- It's, as you can see, it's a temporary chat. It's private mode, so it doesn't know I'm an SEO. But I gave it one page. Why didn't it ask me to upload more pages to actually see the overall health of the website? We c- Let's just say that these optimizations were amazing, these tips were amazing, and we did them. We could have a terrible website, like we could have terrible other pages on our website, and this page would be impacted because the rest of our site is, is terrible, and these KPIs would be meaningless. So one of the things it left out here was just the overall health of the, um, just the overall health of the website. As we scroll down, um, is the last recommendation, again, another rabbit hole. If you could only do three. So why are you giving me all of these high priorities? Then you're saying this is how you measure, but then if you could only do three, do these three, and they're not the three that are gonna make the biggest difference. But this is just ChatGPT, and you can see how, you can see how the recommendations, they vary between technical SEO, content, and so forth. But Claude, what does Claude do? Um, Claude, straight away, it's a little bit different. So Claude just started analyzing the page right away. So it said the highest impact was, um, the missing href lang clo-- h- href lang cluster, which I kind of, um, uh... Sorry, the locale signals and the missing href lang cluster, which I kinda agree with because as a technical SEO, it would be one of the things I change. Because if you have a page specifically targeting Canada, and I know this men's running shoes page for Nike isn't ranking well in Canada, and it's sending UK signals, I'd wanna, I'd wanna, I'd wanna localize that. Same thing with the language and vocabulary. That'd be one of the first things I fix because it's targeting the wrong keyword. And as you just mentioned, it's so important to target right keywords and then measure success against those keywords. So I think in this case, Claude actually did, did, did well. It did a good thing, and it says, "You know what? You need to change the, you need to change the, um, you need to change the language." But now it picked up something that ChatGPT didn't pick up. Over half the catalog isn't crawlable from this, from this page. It says the header says 45 products, but only 24 are in the HTML. Why are 24 in the HTML? Because the rest of them are loaded through JavaScript. So it's actually fantastic that Claude picked this up because other AI bots, um, you know, they wouldn't be able to see the-- They wouldn't be able to, they wouldn't be able to find the other links. So already the prompt is exactly the same. I didn't do anything different, but Claude has now given me something that, that ChatGPT, that ChatGPT didn't, and this is something that you might wanna prioritize over ChatGPT's recommendations. Um, number four, the LCP image is lazy loaded. I think, you know, that's, that, that's okay. I wouldn't-- That's not-- it doesn't, it doesn't really matter. got no product-level structured data. Again, you might wanna think about adding in the structured data, but is it a high priority? Um, Edward Sturm: Is it a high priority? Harpreet: Or I've-- I don't, I don't know, I don't, I don't, think so. I think there's other things that we can do first. But as we go down the list, it's now it says the SEO copy is incitable, and parts, parts are Edward Sturm: But with a page like this, don't you n- don't you, don't you want structured data to see, to see the products in the SERPs and the information with them? Harpreet: Yep, you do. You do. You do. You do. But what's... Edward Sturm: But it's not gonna make you rank better. It's not gonna make you rank better. It's not... And Harpreet: want... Edward Sturm: it'll get you, it'll get you better click-through rates, and that's important. That's important. But, like, what's more important? It's the technical foundations, it's the relevance. Harpreet: Yeah. So, so my thing is, yeah, let's Edward Sturm: fixing, yeah, it's, it... The technical, technical foundation, making sure that, like, Google can understand what's on the page. It's the relevance Harpreet: Yeah. Edward Sturm: then you can think about increasing click-through rate Harpreet: Yes. Correct. Once you're ranking on page one, okay, um, Adidas is above me. They have this structured data on the page. I gotta make sure that I have it, I have it too. So let's get there first and then add this. Then it's got the SEO copy is incitable and parts are wrong. Um, you might wanna look, you might wanna look. It's got-- So Claude is saying there's-- So I don't know Nike's product good enough, but Claude is saying there's, like, issues with, uh, with Nike's, uh, product language on the page. So that's something I'd want to, that's something I'd want to, want to look into. Again, is it gonna help you rank better? I don't know. But it's something that you should, um, you should look into. Then when you scroll down, medium impact. Again, because this is an LLM giving you SEO advice based on what it's learned Edward Sturm: But it says, it says FAQ markup. Yo. Harpreet: Yeah. Edward Sturm: it, Harpreet: Yeah, Edward Sturm: this one's, this one actually has it wrong. It doesn't say that FAQ schema is deprecated Harpreet: it doesn't, it doesn't say that. Like, again, like do you, would you... like this is not-- This one's not Edward Sturm: It's, you know what? It's, Harpreet: advice Edward Sturm: says even where Google doesn't show FAQ rich results, this format gets extracted and quoted by AI. Harpreet: But how can it be so confident? How can it be so confident, though? Edward Sturm: Sure Harpreet: proven, proven that. There's correlations, but because there is so-- So this is my opinion. Edward Sturm: But Harpreet: much content out there, right? There's so much content out there about how FAQ rich results are used by LLMs, um, I think that's where this is, that's where this is Edward Sturm: Was one of the most repeated pieces of bad advice from SEO gurus and people who are big, and I know that I'm kind of a big person in SEO, but I, I w- I wasn't s- I was t- going and saying something that nobody else was willing to say, which is it's not gonna make a big difference. It's... And, and, and you would hear that. I heard that from so many SEO TikTokers who are like, "You need schema for getting cited in large language models," and everybody said it. I don't even w- I don't want to name their names, but it's, like, some of the biggest people in SEO, and it, it drove me crazy. And so that's why you see, that's why you see LLMs saying the same thing Harpreet: Yeah. And I just got a message yesterday in Slack 'cause, you know, a client was like, um, "If we add FAQ schema to our pages, are we gonna show up better in Google's AI overviews?" And even execs and, like, people outside of SEO, like, they're seeing this content and they're, and they're, and they're eating it up. They're repeating the content, and now it's ingrained into LLMs. But as we go down, you've got the title and meta description, and then look at, look at this one. Shouldn't this number nine be up there where footer links point to the UK site on the Nike Canadian website? This is quite-- I think, in my opinion, I think this is a big deal because if you're shopping in Canada, do you then wanna go to the, to the GB website? You don't. So I think this one should actually be much higher from Edward Sturm: Yeah Harpreet: protective, uh, UX perspec-perspective and an SEO perspective. So again, this case now, just because I'm using Claude to do my SEO, again, doesn't mean Claude is 100% accurate. Like, we've already spoke about so many things in this Claude answer that, you know, to improving. And then if you're a business owner, you do all of these things, it's not necessarily gonna help you. It's not necessarily gonna help you rank better. gonna need to spend money on getting someone to implement this, or you're gonna have to spend your own time to implement, to, to implement this. And some of them, they're just not gonna, not gonna make a difference. So we have another one, um, you know, Check AI, Check AI crawler access. It's like go check your robots.txt. Now, if you're a business owner, you're gonna have to read guides to SEO 101 and what a robots.txt is to, um, to, to check this. And then it's got other lower impact, um ... And it's got other lower impact, um, impact, um, pri-priority. So was Claude, the next one I wanna quickly go through is just Gemini and is it any different. So everybody uses different, different AI tools. So in this case, I'm just comparing ChatGPT with Claude and, and Gemini. So Gemini gives his answer in a slightly different way, but it says a high impact SEO recommendation. So implement category content blocks. So this one overlaps with what ChatGPT had. So interestingly, I don't think Claude didn't have this, um, this piece, but, um, Gemini uses ChatGPT's, uh, Edward Sturm: Dude, it is crazy that their number two most important recommendation is optimize the meta description for keywords. Harpreet: Yeah. Edward Sturm: That's the number two, and the number three is remove meta keywords. Like, that's actually going to increase your rankings. That's crazy. I have a friend, one of my close friends, he uses Gemini all the time, and I try to tell him it's the worst LLM, 'cause I really believe that the, the, the output from Gemini is the worst. When Gary Vaynerchuk came on this show, like, we both agreed Gemini should win. We both agreed. It has all the... Like, Google has all the data. Google has the best data. It should win, but their output is the worst by far Harpreet: Yeah. And like, what is this? Like high, like, you know, this one, okay. But yeah, these two, I mean, come on, they Edward Sturm: That's crazy. Come on. Here's ... Look, Harpreet: impact Edward Sturm: here's, here's what I'm gonna, what, what I'm gonna put into the transcript for anybody who chooses to get this transcript and then give it to their LLM. It's, it's don't give a lot of, a lot of advice. Give a very small amount of, of advice, and prioritize the 20% of things that will get 80% of results, and then literally omit the other 80% that is just going to distract me. And, and I want this to be in the transcript for anybody to just give to their LLM. So, so the LL- It's 'cause the, 'cause the LLM will write back this essay on everything that you need to do, and then you just drag this file in and say, "No, no, no. Think about it like this," and you'll get, hopefully get something back that's much shorter Harpreet: Yeah, and I actually want to shout out, um, Lily Ray, who I remember going to MozCon a few years ago, and she did a, uh, talk on the SEO checklist. And this is what these LLMs seem like. It seems like they've gone through-- Like when I gave Claude, Gemini, and, uh, ChatGPT the pages, it seems like they went through an internal checklist, and they're just giving recommendations based on your old SEO, SEO checklist. It doesn't have the nuance in there. It doesn't, it doesn't have any other knowledge in there. Like this is other high-impact AEO, GEO recommendations, breadcrumb list. Integrating FAQ contact cont- content with schema. This is not, um, this is not gonna have a big impact on your AEO, GEO. Like you can do it. You can say you're doing this to optically look good in front of, you know, business leaders, in front of your clients. But if you do this day in, day out, you know that this is not gonna have an impact. But if you're new to SEO and you're using LLMs, yeah, you're gonna think, "Okay, I've done all these things," and then two months, two months later, you're gonna be like, "Okay, why am I not, why am I not, um, why am I not ranking for these things?" So yeah, the whole point, the whole takeaway of this was, as Edward Edward Sturm: Is don't use Gemini for, for advice. Use Claude. Prioritize Claude. ChatGPT isn't gonna be as bad. Gemini's gonna be the worst by far. Harpreet: Yeah Edward Sturm: with Claude, but still have- you have to have Claude give you the Pareto. And, and what's the P- and what is the Pareto? Honestly, the Pareto is the things that, that searchers see first. So it's, it's the, the page-level metadata, for example. Uh, it's the localization if you're a really huge company, if you're an international company Harpreet: And I've got one more tip, one more tip. So I saw you did a video on this a while ago, Screaming Frog. So the Screaming Frog, um, SEO spider, it will crawl your website and it will analyze every page. The f-- It has a free version where, you know, five hundred-- it will crawl five hundred, five hundred pages on your website. If you're a small business, probably enough because you probably don't have more than five hundred pages. But run your website through Screaming Frog and then go to the Issues tab, right? And then export the issues list because what's gonna happen is Screaming Frog is gonna give you a list of all the issues it's found. And then with the prompt you just said and the transcript of this chat, then, um, up-upload that Excel file into your, um, LLM of choice, and then ask it to give you recommendations based on the issues that Screaming Frog found on your website. Because now, instead of just looking at a, a single webpage or an export of your website, the LLM is working with real SEO audit data, and it can actually identify the things that, um, you know, will, that will make, will make a difference. So that's just a tip. If you have a small website, small business, use the free version of Screaming Frog, export the data, put that data into, into, into, into Claude or other LLMs if you want to use the other ones, and just play around with it to see what, um, see what recommendations you, you get Edward Sturm: That's fan- I'm gonna make a short on that. That's fantastic. That's so actionable, especially when you have the transcript from this podcast and you have the, the Screaming Frog Crawl. That's, that's incredible Harpreet: Yes. Yes Edward Sturm: That'll make a big difference Harpreet: Yep. And, um, yeah, it's something that I've been playing around with. If you follow me on TikTok, um, and, and Instagram, I've been making very small shorts of like how to fix your SEO with Claude, and this is the type of stuff I'm doing. So I'm just downloading-- I'm using Screaming Frog. I'm using prompts that would actually be that 20% to show people that, okay, this is how you use AI for SEO. And yeah, anyone who says AI is gonna replace SEOs or you can fire your SEO agency 'cause you can do it all with Claude, ChatGPT or Gemini, they're just, they're just lying and they're just doing it for the clicks. They're doing it for the views. As we've seen, you know, in the la- over the last 20 minutes, you can't. There's just so much more to, um, to, to the practice Edward Sturm: Also, a lot of people who are doing that are gaming, they're gaming X. So they have friends who have big accounts on X, and then they have their friends comment on it. Be- and I see th- I see this because I have an automation to look for viral, uh, viral SEO posts on X. Harpreet: Yeah. Edward Sturm: so, like, uh, every, every couple of days I'm getting one of these articles on X saying, "Yeah, fire your SEO agency." And I see, like, these other big accounts, and it's just all nonsense, and it's all written... It's written by, uh, by AI as well. Harpreet: Yeah, Edward Sturm: it's just complete garbage Harpreet: It's just sad 'cause there are-- I, I know business owners who fall for this. Like, it's different at an enterprise level 'cause they have multiple teams and, like, multiple people working on SEO. But if you're a small business trying to get ahead, there's just so much noise on the internet that you might not know that you shouldn't listen to this piece of noise. But because this has, you know, 500 retweets 'cause it's been artificially inflated, you might think that advice is true and fall for it, but at the end of the day, it's not gonna help you grow. So, um, yeah, it's really important to SEO advice from, you know, trustworthy people, and I think this podcast is one of those, is one of those mediums Edward Sturm: Yuharpreet, it's always amazing to have you on Harpreet: Okay Edward Sturm: Is there anything else that we're looking at? Or that, that's it? Harpreet: that was it. That's it. Edward Sturm: All right. Harpreet, where, uh, Harpreet, where should everybody go and, and find your stuff? Harpreet: Um, yeah, so all of the descriptions, all of my socials are gonna be in the description, but I'm most active over on, over on X Edward Sturm: Awesome. Thank you again for coming on. Harpreet: Okay. Edward Sturm: This is episode 1,227 of the podcast. This is so many episodes. Crazy. Harpreet, you came on, I think your first time on was, uh, oh, like a year and a half ago, and Harpreet: Yes. Edward Sturm: so- we've done so many episodes together. It's so awesome. Harpreet: Yeah, it's been Edward Sturm: Yeah. Yeah. Um, so everybody, if you watch this on YouTube, thank you so much for watching. If you listened on Spotify or Apple Podcasts, thank you so much for listening, and I will talk to you again tomorrow. Bye now.