Transcription
If you want your brand to show up in Chat GPT, you have to do the exact same thing that you do in order to show up in Google. Or maybe not. I've seen some experts say that ranking in AI search requires a rather different set of tactics compared to conventional SEO. Confusing, right?
Well, for this episode, I brought in our very own Patrick Stalks to shine the light on this argument. Patrick is one of the most knowledgeable and experienced SEO professionals that I've ever met, and he's been keeping a very close eye on all the AEO GEO hype. And trust me, he has quite a few things to say about all the advice that some of the newfound experts in this field are pushing.
In this episode, we went deep on all the intricacies of getting mentioned by AI chatbots and explored how getting your brand into ChatGpt is different to ranking in Google. Sounds interesting? Then keep listening and don't forget to hit like to support the work I'm doing with this podcast.
Hey Patrick, welcome to HR's podcast.
Hey Tim, excited to be on.
Awesome. So, let me give a bit of context for anyone listening. Uh, we recorded a podcast episode with Ryan Law, uh, our head of content marketing, where we went through all the kind of fundamental principles of what we call answer engine optimization, or AI search, or how to show up in Chat GPT and Google's AI mode and everything. And, uh, since you are known to be like the technical SEO of technical SEO, uh, I want to go deeper with you. So, uh, in this episode, I want to cover things that we didn't, uh, talk about with Ryan. And, uh, since you had a chance to listen to the episode and make some notes on things that we didn't discuss, I'm basically giving microphone to you so that we can go into some additional stuff that people who are in the industry really want to know about, uh, ranking in AI search.
All right. Yeah, sounds good. Let me share my screen.
Yeah. So, you did, uh, say, "Tell me what we got wrong."
Yeah. Yes. So, I got a couple things. Uh, not too bad, though. Uh, one was that y'all talked about, uh, that these AI bots, they don't render JavaScript, but it's not entirely true. You've got Gemini from Google, you've got Copilot from Bing, they're out there, they're rendering. Uh, so they got to have an advantage over these other systems in that regard. They are going to see that kind of content, and anyone that's powered by one of them, uh, like if for the RAG models, the retrieval augmented generation and stuff, uh, they're also going to get that data, that content.
Okay, let me interrupt you right there because, uh, I'm not as technical, uh, as you are. So let's talk just a little bit about JavaScript. Like, how common is it? So, for example, if I'm using some website builder, I don't know, Wix, Squarespace, uh, or some other kind of, uh, system where I don't code my website myself. How common is it for these kind of, uh, CMS, content management systems, or website builders to use JavaScript?
Yeah, I think every website uses JavaScript in some way. Uh, but it's mostly going to impact it if you're not showing the content by default. And that's going to be more common in systems like React, Next.js, Vue.js, that kind of thing, like the JavaScript frameworks. Uh, on the majority of sites, I'm going to say 90% plus, it's probably not a problem.
I wonder, like, did we do any studies at Hrefs? Because, uh, we run our own web crawler, and we executed JavaScript. Did we do any studies like how much, what percentage of content on the pages is kind of hidden within JavaScript?
Uh, I don't think so because that's going to be like a difficult, uh, it's not just that it hides the content, but it can also like reorder, change the thing. So like doing a comparison of that is extremely resource intensive. I, I think if I asked our devs, they would be like, "No."
Okay, let's move on then.
All right. Um, there was also a conversation you and Ryan had where you kind of thought that, um, what is going to work best is to have different content. And I think you need different content, and you also need the same content as the current websites have. And the reason for that is you...
Sorry, I will interrupt you for a second. You mean we were discussing like, in order to to make kind of AI interested in you, because AI is pulling from a few different sources, and I think what we were discussing is that if you're saying the same content that 10 other sources that AI is pulling from, the likelihood that they would use your citation is smaller. So I think one of us was suggesting that you, you need to provide some, like, it is an opportunity to provide some unique content because then the AI can look like, "Oh, these nine pages are saying the same, but this page is saying different, so let me cite this page as well for this different insight." Right? And your take is that you need to say the same thing that other sites said and also like add to that. And it's because there's going to be different systems to be in that first consideration set, you're going to need similar content, like, basically, you're going to have to be a good answer to the the query. And then to differentiate yourself, uh, you will actually have to say potentially something different, add something of value to that. And whatever you add of value probably has to be repeated on other sites too for them to actually choose you. But there's, there's a bunch of systems like this, uh, you know, like we've built some of this, and there are some that are just kind of like they have to look at things that are almost contradictory because it's different processes. Like another one might be, um, uh, do I choose a strong site or do I choose a topical site? And the answer is, you probably want some balance of both. Like, you just don't want Wikipedia for everything because they're a really strong site. You also want like a niche relevant site like Hrefs, that's also pretty strong. Um, so like there's, there's all kinds of balances like that. Or, uh, do I want complete content or do I want content that's shorter and specifically answers what the user asked? And the answer is, like, both are probably good answers for different things, or, or even the same thing. So like these systems, they have to make decisions like that where it's not just, you know, you can't just say like, I want to create one long 9,000-word page about this because that might be good in some cases, and in another case, it might be like, well, I can create five pages out of this that cover different topics, and really, you can go either way. And when they're choosing the answer, they can go either way.
Mhm. So, uh, again, to to clarify this a little bit, we're talking about the content of one page. So, uh, you're not saying that you should have a page that's that is saying the same thing, and you should then have another page that is saying different thing. Like, if you want to get cited, you need to have a page that kind of is covering consensus but is also adding to it on the same page, kind of to deserve the citation.
Or like you can do all sorts of ways. You can have different pages. You can have on the same page. Yeah, it's, you potentially could have something different on a different page and that be pulled in, whereas one page has something similar. There's no like right or wrong for this, is what I'm trying to point out. Um, the, the systems are designed in a way that sometimes they'll prioritize one thing, sometimes they'll prioritize another. During their, uh, their versions of core updates, they may change that weighting. So what works today may not be what works next month.
Again, to clarify, I think it was it was me who was saying that, uh, kind of saying something unique and adding information gain can can deserve your citation. This is pure theory. This is like this is not backed by any, uh, experiments or research studies or anything. But I think it would be good going forward to try to analyze citations and see how different they are, like what, what number of citations have like consensus, and what number of citations are kind of bringing information gain. But I don't think anyone in our industry did, uh, like research of that kind.
Right, not in our industry, but in the information retrieval industry, they they have done that. And this is like one of the things when they're building the systems that they are looking for. Again, they're looking for the similarity and they're looking for, uh, that information gain. So, it's just depends on how they're weighted and how much of each you have.
I'm trying to, I'm trying to look for an actionable takeaway for, uh, listeners. And it's almost like, just create good content, uh, because you cannot kind of engineer your way there almost. You don't know what would work, like the consensus content that is like well put together and on an authoritative domain, or some like crazy information gain, some crazy new ideas on the same subject that is not in consensus with everything else. You don't know what, what will eventually get picked up by these systems, right?
Yeah. Everyone is going to have the same discussions, I think, where do I create one page, do I create five pages, do I create 10 pages, a whole hub of this stuff? And kind of like it was with SEO, there's not really a right answer. Like, you could try and split it out into multiple pages, see if that has an effect, but you might be waiting like six, eight months, uh, potentially for to see the changes in this case. Uh, so I think it's just, don't overanalyze this. Just pick one, go with it.
Yeah. Well, create good content is perfectly, uh, acceptable advice for me. That's what we're doing at Hrefs, and it's kind of been working for us for the past decade.
Yep. So, let's move on.
All right. I think, uh, what I, one of the things I didn't hear is how the LLMs actually work. And I want to cover this a little bit because it leads to something else that I'm going to talk about.
Mhm. Um, they're basically, in the simplest form, next word predictors. So if you've got the, uh, CMO of Hrefs is, it's going to try and like fill in Tim, obviously. Uh, so it, but it's going word by word. And these are based on patterns, uh, and probabilities of what the next word is going to be. And those probabilities come from generally the web content is what they're looking at. Uh, so you've kind of got randomness in this, the the different probabilities, you've got patterns in the training data that they fit in when they train the models, uh, which could come from PDFs, books, um, websites, social media, any number of different sources, depends on what they put in the the data. But we don't, we don't know. But we, we kind of have that data too. Like we crawl the web, we see all this stuff, uh, the YouTube transcripts, like all that is good data for training data. So if you have a bunch of training data that says the CMO of Hrefs is Tim Solo, then more than likely, like that's what it's going to write, uh, if it's trying to figure out who the CMO is. Uh, then you've got the, the retrieved pages, which is like a secondary process. So you've got the, the trained LLM data, and then you've got the data where it goes out and fetches a bunch of relevant pages. And those come with other probabilities. So the text on those pages say different things, and the words that are used basically are are what are you, it's using to try and write sentence after sentence, paragraph after paragraph. And then they just built in more randomness that they called temperature, which is generally like 0.8 in most of these systems. Um, and that's just to, to make sure it doesn't generate kind of the same thing over and over again. So that's why like if you ask it, you know, five times the same question, it's going to give you different answers. It's you've got probabilities on top of probabilities on top of probabilities.
In in that regard, I, I typically, uh, advise people to look up, uh, an article by Stephen Walrram. How does Chat GPT work, or like, what does Chat GPT do? Something like this. Uh, it's an amazing article. It gets very technical very fast. But if you consume even, like, if you read even like 10 to 15% of it, you'll, you'll get a good understanding of how, uh, uh, this system works. And, yeah, that that's exactly what you just described here.
Yeah, I'm hoping we'll actually show this in Brand Radar coming up. Um, like we can see the probabilities of the, not necessarily each individual component, but like we have all the stuff we crawled on the web, and we could kind of build that, you know, these are the things that are commonly said, these are the probability that the next words are this or this. It's going to make this statement about, say, the Hrefs keyword index size or something. Uh, all that is kind of coming from the content that's on the web.
So the whole reason, one, one quick addition to this, uh, the temperature and probabilities is also why, uh, those LLMs hallucinate. Because once the kind of randomness in predicting the next word gets a little like off, it starts generating like complete nonsense. Typically, that's, uh, that's going to come from the training data. Like the RAG systems, I think, are less likely to hallucinate, and usually they're kind of, they're kind of grounded in the content that's that's retrieved. Uh, not to say that they can't hallucinate, but they're less likely to than if you're just like, if it's doing a web search, it's less likely to hallucinate than if it's just coming from the training data.
Then it's interesting. Is this actually validated? Like, was this studied? So if they, if they're pulling information, uh, from the web and then giving you an answer according to this information, they're less likely to hallucinate, right?
Yep. And you'll see that even with, like, the URLs. I know you all talked about redirecting pages that were hallucinated on the website, uh, if it's done, if it comes from like the web search part of Chat GPT, it's less likely to be a hallucinated URL, uh, than it is if it's just generated without the web search.
Okay. And the whole reason we're talking about that is the whole, the purpose of GO or AEO, or whatever we're calling this, still SEO for a lot of folks, is that we just want to control what's being said. Um, we only have a certain amount of control, though. Like, we can go and create a bunch of pages, a bunch of content. Um, but if you have real issues, like, if if everyone suddenly is like, "Nvidia's graphics cards suck," and they're saying that on social media and Reddit and everything, YouTube, uh, there's only so much you can do to combat that. Now, if you're a small business, you can probably control this pretty well. Like, there's not a lot out there about you or your competitors. If you're a larger company, you know, you, your best bet of controlling it is put a lot of budget behind it. Um, and even then, it might be drowned out because it's the entire internet that's talking, how they're talking about you, and that is what matters. So there's, there's really only so much that any business is going to be able to do. Larger companies, you know, unless they're, they're throwing tens, hundreds of millions of dollars, in a lot of cases, they're not going to be able to do anything. Uh, not enough to move the needle anyway.
So, I know I'm the tech guy, but I'm going to take this to business. Your top tactics are actually business-related. Just be better. Build a better product, better service, uh, better business. Address any issues or complaints your customers have, because if they're out there just constantly complaining, well, now I got to run a campaign to, you know, to try and counteract that. And there might be tens of thousands of them, and there's one of me. So what am I going to do? Uh, I'm going to be a drop in the bucket. It's going to be noise in the void. Like the internet is still the one controlling the narrative. So, there's only so much that SEOs are going to be able to do in a situation like that. Or like we had, um, you know, a few years ago, we had the credit system, and we had a lot of complaints about that. They just kept coming, social, Reddit, everywhere. And what did we do to fix that? Like, we basically got rid of the credit system. No matter how much we tried to explain it or say like, this was a good thing, that it actually, like, helped keep the price down for the majority of users, it didn't matter. We just still had the complaints. So until we changed that, which was a business decision, like, there, we would have to this day probably still be having the complaints, right?
Yeah. And I think like SEOs aren't necessarily involved in these decisions. So this is going to get contentious in a lot of organizations, uh, where you're going to have to do something that, you know, the execs are going to be yelling, "Hey, can you just change that? Make it go away." It's kind of like reputation management. You can do a certain amount, uh, but if, if your reputation is bad enough where they just are constantly hitting you, um, you can only make so much noise, depending on your budget. I mean, if you've got enough budget, sure, maybe. Uh, but now we're talking, like, everyone seems to be relying on AI more and like letting people go, and there's more work to do than ever if you actually want to control that narrative, if, if you want to have it say what you want.
I, I absolutely love what you're saying, uh, and this is of course very much in line with our own kind of marketing ideology and what, what message we're giving to to, uh, our audience and to the marketing people in general. Yeah, marketing starts from building a better product. It's, it's the, the best way to to sell, to become popular, to build a brand, is to create something meaningful. I do agree with that. I also love how you mentioned that, uh, SEOs don't often have a say in all those kind of business and product decisions. Uh, and they are just like, "Go fix it." But like, to fix it, you need to fix something in your business. You need to fix something in a product. And this is why I also try to promote the narrative that SEO is more holistic than many people think. It is more integrated to kind of larger business and product structures, uh, than many people think. It's not just about like researching keywords and building links. Yeah, you need to understand like what market you're in, what product you're building, what people are saying about you, and then it kind of it informs your SEO strategy.
I completely agree with you about this. But again, people want specific SEO tactics. So, yeah, building a great product is great, uh, building a great business is great, but SEO still exists, uh, as a discipline, where you can do certain things even if your product is not ideal, even if your business is not ideal. Uh, so are we going to discuss those things on your next slide? I'm not sure where you are going with this.
Yeah, there's, uh, there's going to be more about business integration, uh, coming up. But you talked about the tactics, and that's where I'm going right now.
Yeah. Is that on your site? You and Ryan talked about a lot of stuff. He had the whole bluff flame framework, say the right things, like all that. Absolutely great. I just want to add a few more things. Don't use marketing jargon. It's not going to come up in these these, uh, AI searches, AI assistants.
Site your sources. This is something where Google was always like wishy-washy on this, like, does it help? Does it not? You know, if you link out, if you cite your thing, they, they were always like, maybe it helps, maybe it doesn't. And, uh, so a lot of SEOs don't really do that. We got out of the habit of it because it's like, we don't want to leak out. But for these systems, that does matter. It is something that they look at, and again, that comes from the information retrieval side. It is something when they're building these systems that they're considering. Uh, and there's the whole concept of, uh, what is it? The Matthew effect. So like, if you're already a popular company, popular brand, popular product, like, you get mentioned more, you get cited more, and that just leads to like more fame. It's the, the rich get richer, basically, is the Matthew effect.
Say what your thing does. There, there's so many sites out there where I'm looking at a page and I'm like, I don't know what this company does. I don't know what they're selling me. I don't know what their product does. Uh, you have to be very specific now. Like, if you've got a certain feature, we, we're having this issue right now, I think, with the, uh, the new social media monitor. All the AI tools are telling us like, "Oh, Hrefs doesn't have a social media monitor." Well, we do. Uh, I think we finally created a page for that, but like, we kind of have to spread the word, get it going, uh, before it'll actually show up that we we have that, we have that feature capability.
And then like more pages. We SEOs have typically been like, we're going to build one page because that's what's going to rank in Google. Well, even Google will show like two pages from the same site now. Um, so you can have more content that kind of overlaps than in the past. But there's other types of content too that, you know, businesses always are like, "Oh, I can't mention my competitor. I can't show, uh, I can't show a comparison table. I can't show their features or talk about their product." And this is usually a hold-up within the legal departments. You almost have to do that now. Like, figure out a way, if even if you're maybe not naming them directly, or use some similar name, that you need this content because it is highly cited in these systems. And I'm already seeing a lot of, uh, kind of manipulation of this. Um, like I'm seeing people actually use networks of sites and say wrong stuff about their competitors. There's going to be all kinds of really interesting lawsuits, I think, coming up from this, because, yeah, they're, they're literally like almost defaming their their competitors, and that is not allowed. But they're doing it, and they're, they're putting that content on a bunch of sites. They're paying for that content to be on other sites, you know, if they don't own them. And that is really something that, like, you don't want. If it's, if these systems start returning bad information because you don't have anything that that counters it, that's going to be a real problem for your business.
I really like the, the bullet, the bullet point, "Say what the thing does." Well, it kind of, it is connected to "No marketing jargon." Just explain it plainly so that people understand, so that people have to guess, won't have to guess. I believe there is a, is this a book or like, is this some copyrighting or marketing framework? "Don't Make Me Think." I think there was a book with this title, "Don't Make Me Think." And I really like this concept. And, uh, as marketers and copywriters, we can often get carried away by trying to like be smart, create some kind of sentence or a headline with, like, an interesting twist or like some catchy thing, completely forgetting that we need to explain our thing to people in the simplest terms.
Yeah. I, I used to do this exercise with the, the content team, uh, at my previous company, and I would literally take like headings, H1s, from a bunch of different web pages, and I would be like, "Tell me what this page is about." And they were just stuff that was so generic that you would have no idea. And it wasn't just the headings. Like, the content on the pages was was similar as well. It was, you could go to a dozen pages, and they all kind of sound like they do the same thing, and they're completely different products. So I don't know why anyone bought anything.
Yeah. I also think that marketers in each, uh, organization, at each brand, they have to remember that they are in a bubble where they are, they know like a lot, as much as is is possible to know about their product, and they're communicating with people who are also at the same kind of level of knowledge. But they're talking to people who might have discovered your product like a minute ago, and they know nothing. And if you try to be smart in explaining your product to them, if you try to be like, creative copywriter, that might be a bad strategy. So even our marketing department sometimes guilty of it. When I'm reading some drafts of our landing pages, I'm like, "Why don't we just say it as it is? Like, say it clearly? Why we try to hide, uh, the truth behind some metaphors or something? Why do we need a metaphor? Like, let's say it plainly." So, yeah, that advice resonates with me a lot. And, yeah, like back to AI search, when AI, this statistical model, is generating an answer because it uses the words that you used previously to generate its answer. When you're using those metaphors, this jargon again, you can just take it sideways instead of just using plain words, like, "Here's what it actually is," and then it won't be able to get it wrong, basically. So this is what you're saying, correct?
Yeah, exactly. And you brought up metaphors, which goes right to my next point, which is internationalization. Uh, when we first looked at this, um, I was trying to figure out why the traffic to these pages was so different from traditional search compared to the AI search. So we actually dug in, and, um, I'll show you some data in a second, but one of the main reasons is that the, the international corpus when they're training this data is actually like much smaller. So even if someone is searching in another language, like Japanese, Korean, Spanish, uh, an English page may actually be the one that's cited. Um, and again, I will show you some data on that on the next slide, but if you don't translate your content at all, then you definitely won't be cited. So you're still better off to try and have your content translated. You don't want to use metaphors on your pages if you're doing that, because they don't translate well. Um, but yeah, it's, here's, here's the data. Is, uh, what we had, Korean, over almost half the time, a Korean page was not cited by Chat GPT when people are searching in Korean. That is crazy high. But all the languages are kind of like that, where you're seeing 20, 30% of the pages that are cited. Even though someone's searching in Japanese, 36% of the time, it's like an English page that's cited, not the Japanese one. So if you don't have your content in the languages at all, you definitely don't have a chance. But even if you do, it still may be the wrong pages. It still may be, you know, the English language pages that actually get cited instead.
I think this is a kind of proof that these systems probably need some kind of like Hreflang system. I don't think they're taking anything like that into account that would let them, you know, see the proper page for the thing that's cited, because they, they may be actually like pulling something that you have a Korean page or a Spanish page, and you have an English page, and they may cite the English page anyway. So, this is more a problem of like the systems for now, not necessarily something that you did wrong.
So translations, uh, mostly apply to, uh, big international brands, not so much to, uh, strictly US businesses or like English-speaking businesses. So I don't want to, uh, dwell too much on this, but just give me the advice. Uh, it is quite easy right now to use, like, AI translations without even having, uh, native speakers review that. Would that be enough? Like, what's your, like, what's the minimum viable way to to do the translations?
Yeah, I would just run it through AI. I run it, run it through a process there. But even in the US, what like 40-something percent of the people speak Spanish now? I guess it's not something I would ignore here. At least have the two versions.
Makes sense. Yeah. Um, and then when you're optimizing your pages, something that's different. We've been looking at the mobile designs and mobile content and everything for like a decade now. Everything has been mobile first, but these systems are not being used on mobile first. They're, they're desktop first again. So, you're getting people to your site, um, maybe you're focusing on mobile conversions, but then if the people are coming from these AI search systems, the AI assistants, uh, it's very, they're more likely to be on desktop, 65% desktop.
Wait, wait. Uh, you mean that people mostly use, uh, AI systems like, uh, Chat GPT, Perplexity, or even Google's AI mode on desktop over mobile?
Right? I don't know that we can properly manage, uh, properly measure if it's coming from Google. But in general, when we looked at this, we look, use the, the LLM segment. So that's Chat GPT and Perplexity and all those. Um, and yeah, they, I would imagine GPT app, sorry, I would imagine the Chat GPT app on mobile phones is pretty popular. But can you properly track how many people are using it if they're sending the user agent? I don't think I tested that one. Uh, I did test a bunch of user agents, uh, to see if they were passing earlier this year, back in like May or something. Uh, so there is a study on the blog on that, but I don't know particularly on the mobile app there. But I would assume so. They actually like fixed a bunch of stuff after I published that article, like by June, like their numbers doubled basically because they fixed a lot of the ones that I pointed out were wrong.
Uh, so, and what's the actionable takeaway here? That check your site on desktop. It's something like I haven't really done that in years. I, I normally like default and check the mobile site now. Uh, now you really got to be looking at both, cuz yeah, it matters. Probably should have been looking at both the whole time, but yeah, when, when mobile was like 75% plus, it's like, well, I'll just look at mobile.
Real companies do things that are not just optimizing their website. They're, they're in more places. They're on social, they're on YouTube videos. Uh, so one of the big things Ryan talked about was that I had mentioned was that a lot of this now is not your site, it's other people's sites. And that means like where your audience is. One of the easy wins or things you can do is just like repurposing content. So if you've got a blog, we do this often where we will have a blog, with it also as like a presentation, and then social media posts go out on, you know, X, LinkedIn, etc. And then there may even be a video about it. And all that is great. Like, you're, you're syndicating your content to multiple channels that all can get picked up, have the same kind of messaging, have the same, like, in our case, we have a bunch of data studies and statistics and that kind of thing. And all that helps it get picked up to say, you know, we, uh, I think some of our top pages are simply, uh, statistics where we had seen the 34.5% drop in clicks from AI overviews, or that like 12.1% of our conversions were from AI search. All that stuff really got picked up, gets returned, and people end up on those pages for or because of those statistics.
Um, so as many of these as you can get it on, I mean, obviously don't spam this. There's, there's tons of stuff from 15 years ago where it was like, let me put all this on plague sites and blah, blah, blah. Don't do that. And don't, don't be spamming Reddit and all that, but actually be doing these things. Go out and present, create the YouTube videos, share your your stuff on social. Like, we're really good. I think, uh, usually one post, sometimes is like two, three different social posts, and might even make it into like a roundup of a bigger of more posts and stuff. Uh, yeah, all that stuff.
Yeah, one quick note about repurposing. I think what a lot of people misunderstand, they think that repurposing goes from one kind of big piece of content into smaller ones. So, for example, we're recording this podcast interview, and then we'll cut it, cut it into snippets, then we'll take, uh, small takeaways and share on social, then we'll, we'll write an article about it, etc. But it's actually completely the other way around. It starts from small things like tweets. You have an idea, you share it on LinkedIn, you share it on X. Then if that idea has legs, you might actually like do a study or write a bigger post about it on the blog. Then once you have a few of these ideas that are related, you can do a talk at a conference. You can do a webinar. And then once you have a webinar, you can go and do it, repurpose it as a podcast interview. And this is exactly what happened with with Ryan's, uh, previous episode that we did with Ryan. It started from a few random tweets and shares that we were doing. Then we did some research studies that we published on Hrefs' blog. Then Ryan took a bunch of these studies and created a webinar that we ran for, uh, our audience. And then I saw the slides of his webinar, and I was like, "That sounds like a podcast interview that we can record. Let's talk about this on a different format." This is how repurposing works. People think that you just hop on a call, you talk about random stuff for an hour, and then suddenly you have a lot of great blog articles and, and posts and everything. I think it, I see, like for us, it works reverse. So again, uh, with this podcast interview that we're recording, all the insights that you're sharing come from different articles, from different things that you share. And so you literally share the screenshot of your tweet. So it's actually all the other way around. Repurposing works, works the other way.
Yeah. You know, and it's, it's all iterative. Uh, so the more you like, you cover the same thing, the more you realize like what people actually want to know about it, and you can kind of refine your messaging and that kind of thing. So, yeah, I, I love that, uh, that comment. Um, specifically on some of these video, YouTube is the second largest search engine. There's a lot of focus on all these AI assistants that are in total driving less than 1% of your traffic. YouTube might be, uh, you know, a lot more. I mean, I don't know how much it's going to drive traffic to the website, but there's a lot of eyes on it. I know for us, like, we see it in our signups. Like, Samo's videos are awesome, and people come and they sign up for Hrefs. And this is exactly why we're doing this right now.
Yeah. Yeah. Your, your podcast and everything. It's, it's an incredible channel that I think as people need to diversify, to kind of hedge their bets on where their traffic is coming from, this would be my first choice. Like, go and do more video. There's your action item. If you're not doing it, go do more video right now.
Yeah. I feel that, uh, a lot of people get fixated on, uh, AI assistance like Chat GPT and Perplexity and optimizing for AI search because they are kind of looking three, five years ahead, and they are kind of projecting that in three, five years, that might be the dominant thing, how people search, it would be like all the traffic you get, blah, blah, blah. But again, if we focus on today, today, like you said, YouTube is much more popular than Chat GPT, and YouTube has a lot more business potential than Chat GPT. So, yeah, definitely you have to invest in AI search. You have to do the groundwork that would help you rank in Google, rank in Chat GPT, and everything. But if you're sleeping on, YouTube isn't going anywhere either. I don't see YouTube losing its relevance five years from now. I can only see it getting bigger and bigger because the, the new generation of people that is growing up right now, they are very video-oriented. Short-form video, long-form video. So, yeah, definitely if you're, if you're putting all your eggs in the basket of Chat GPT but not putting anything in YouTube, that's a big mistake.
I totally agree with you. Yeah, and it's not just YouTube. Like, you mentioned the shorts, and that's that's Instagram, that's TikTok. That's where people are. Tons of people are there. And yeah, it's, uh, you made a comment that people are looking three to five years ahead for the AI search. Well, they're, they're 10 to 15 years behind on YouTube already video formats. So get on that, uh, prioritize that, I would say. And it's the same social media, uh, be active, you know, to actually like have conversations. But anytime you're sharing something, it's more eyes. And we did this, you know, when we did blog articles or make a video, we share it on social because the idea is like, more people are going to see it. And that mean, uh, meant and traditionally more people might link to it. Well, now they might say the same thing you said, they might repeat, get a statistic, they might, uh, actually mention your, your article or your video. They might cite it and say like, "Hey, look, you know, Hrefs had this statistic on this post and blah, blah, blah." And all that stuff matters. It matters a lot for the these AI assistants. Again, it's, it's that like flywheel, where they're, the more popular you are, like the more likely you are to actually be cited.
Let me clarify something. Do we actually see that AI chatbots like Chat GPT are actually pulling citations from LinkedIn, from Twitter? Because they, they're not. So when you're saying be active on social media, you're talking about the secondary benefits. You, you're basically, you're promoting your ideas, you're promoting your business, you're promoting your brand, your products, your services. Uh, it gets more eyeballs, which increases the chances of someone talking about it on their website, and then this is something that, uh, AI chatbots would eventually pick up. So it's not kind of a direct benefit, it's kind of a secondary benefit. And it's the same with ads. Like, we ran that, uh, study where we, we didn't even, uh, have the blog post public, but we ran ads with a bunch of, what was it? SEO statistics. And people started linking to the SEO statistics post, and the only way they would have seen that was from those ads. And it's the same in social. Yeah. More eyes, more chances to be again mentioned, linked, for them to say that narrative that you want them to say, to mention that statistic that then the AI system might repeat, or to say that, you know, Tim is the CMO of Hrefs. Like, whatever you're trying to push. Like, that's a good way to get, it's just more eyes, more and potentially be mentioned more. Okay?
And it's the same other sites. Um, so if you're in communities like Reddit, don't be spamming. Be actually active. Use your employees. Like, if you're not going to have, you know, a dedicated Reddit person, dedicated Reddit strategy, just have some people at your company that are probably already on Reddit to go and look and monitor this stuff, respond to it. Because the Reddit is, is being heavily leaned on for a lot of these. So if you want to control the narrative, like you have to be there. I think, uh, in some subreddits, potentially people are a little nicer to us because we're actually active there, and we've seen cases where like mods will take stuff down and that kind of thing. Um, so do that. Do digital PR. It's the same thing, like spread the word. In that case, like you can actually have, uh, content control. So you, whatever you're putting out, potentially is what will be on the other site. So you, you can control what is being said with those content syndication. The same, uh, influencer marketing, like leverage your network, leverage pay people. Uh, any, any affiliates, if, like, if you run an affiliate program, like this SEOs don't scale. And businesses are cutting back and saying like, "Oh, you have to do more with less," because they're like, "Oh, it's going to be AI everything and blah, blah, blah." But no, if you really want to rank in these AI systems, it is more work than ever. It is no longer just optimizing your website. It is now kind of controlling that narrative across all these different places. And that means it's not just you, like, it can't be. So any, anyone you can leverage, any friends, any affiliates, influencers, etc. But that can be other teams. So bigger companies, they might have a team that manages the affiliates. They probably have a PR team for digital PR. They have an influencer team. Uh, the ads team would probably be the one that handles content syndication. Social media team for all the community social media websites. It's, it's way more than SEO now that that is going to help you be successful. And like this is what big businesses have done for a long time anyway, but not necessarily integrated as well as they should. And that, that's something that now we have to figure out, like, if we really want to be successful and we really want to control that narrative in all these, uh, AI assistants, we actually have to integrate with all these people and work with them and work together to make that happen.
Let me share, uh, an interesting observation based on what you've just said. While our entire industry is bickering over, is it SEO? Is it SEO? Is it search engine optimization? Is it search everywhere optimization? What you're describing here is essentially just marketing. It's not even SEO, it's marketing. And again, this brings me back to the point that SEO is a lot more holistic than some people think. Yeah, there are industries where SEO is its own discipline. If you're building like an affiliate website, yeah, you're not really building a brand most of the time. You're just doing keywords, links, and all of that. But for a business, SEO is very holistic and very integrated. And this is exactly why, uh, Hrefs Evolve, our conference, uh, this, this podcast would probably go out after it, but this is exactly why we're not limiting ourselves to just SEO people. So, if you look at the list of our speakers, they're like marketing generalists. There are B2B marketers. There are people from PR. There are people from, like, all, all walks of marketing, because I genuinely believe that seeing the bigger picture and understanding everything that happens, like, across all of digital marketing is what helps you to succeed in search, is what helps you to get found, whether in Google or whether in AI search. And this is exactly why this podcast is not so focused on SEO only, because SEO is quite a narrow discipline if we're talking about SEO specifically, keywords, links, domain authority, that's about it. But if you start looking at it holistically and all the things that feed into SEO, then it suddenly becomes digital marketing. And again, this is why Hrefs, as a company and as a brand, is moving towards becoming a marketing platform rather than an SEO platform, because if you want to succeed in SEO, you, you don't need to just do SEO, you need to do all of digital marketing to be successful with SEO. Is, is this something you agree with?
Yeah. And I think earlier we also talked about it being product and business decisions on top of just marketing. So it's, it's everything. We're touching everything at this point, and everything matters. Uh, so the people that really want to do this well, like, it's just have to work with a lot of people to get it done.
Okay. What else do we have?
That's all I got.
Awesome. Well, Patrick, uh, I won't drag this longer than it should be because again, we're on YouTube, and we need to be mindful of people's times. Thanks a lot for coming in. Thanks a lot for closing the gaps, uh, in my interview with Ryan. So, uh, if anyone didn't watch my interview with Ryan, go watch it. Yeah, thank you, Patrick. Uh, I think we'll see you again on this channel, uh, quite soon.
Awesome. Well, thanks for having me, Tim. Take it easy.