Do Backlinks Still Matter in AI Search?

00:00

Welcome to episode one of the AI search report. Today we'll answer: Do backlinks still matter in AI search? And does Google treat AI generated content as spam? You'll also learn why GEO is just overhyped SEO and how to rank on YouTube search. I'm your host, Nathan Gotch, and this episode is brought to you by Rankability, an AI SEO platform that helps you rank everywhere customers are searching. So let's get right into it.

So one of the biggest questions that everyone has right now is whether backlinks still matter for AI search. So we're talking specifically about AI chat platforms like ChatGPT or Perplexity or Google Gemini or even Google's AI mode as well. Okay, so do backlinks, which are one of the most critical ranking factors for traditional search results, which are the blue links that we're used to—do they actually influence these AI platforms? Well, SEMrush just did a little study and they found that the authority score, which is their third party metric, by the way, this is not a score that Google uses per se or that ChatGPT uses.

01:02

It's just their own proprietary metric. They did find a correlation between having a high authority score and being a citation in the AI chat platforms. Now, this is pretty obvious that this would occur. It's good to have data to prove this, but if you've been using these tools for a while, you know that typically when you look at the citations, they are usually from very authoritative websites. They are usually very trusted entities, and there's a reason for that because we know that when ChatGPT uses retrieval, same with Perplexity and all the other ones, they tend to just go to traditional search and each platform uses a different search engine. For ChatGPT, we know that it uses Bing. There's some evidence that it might use Google as well. We know that Perplexity also uses Bing, potentially maybe even Google as well. And we know that Claude, for example, uses the Brave search engine.

01:57

So the main takeaway here is that once again proving that if you focus on the traditional ranking variables in traditional search, you can also increase your odds of appearing in the AI generated responses. Now, let's rewind on this real quickly. So what you want to be thinking about is obviously your own website, your first party opportunity, which is growing the authority of your website. So then you can attempt to influence, at least give yourself a couple of points inside of this process. But really what you want to be thinking about is how do we seed our brand on as many sources of retrieval as possible? So the number one thing you want to be thinking about is how do we get our brand on all the top ranking results in traditional search for our most important keywords? So it's going to be those commercial based keywords because all ChatGPT does when it uses retrieval, it's basically just acting like a searcher.

02:52

It creates queries, then it goes and searches those queries in traditional search engines and then uses those top ranking results as a source of retrieval. Now, it doesn't use everything in those top ranking results. It will decide and pick and choose, but in general, that is what is going on. So this is more of like a proxy more than that ChatGPT necessarily uses Authority Score. That's not true. ChatGPT just goes to traditional search, which uses more traditional ranking variables. Be careful to say that ChatGPT uses backlinks as a ranking factor. That's not really true. It's really just a proxy because in traditional search that is what's used to rank there and it's going there to get its results.

03:50

So just to be fair though, this is not anything new. I literally did a study on this in November of 2024. You can go and watch it. I'll drop it in the description below. But I talked about this in a much smaller study than what SEMrush did. I actually use SEMrush's metric in this study, and I found that the median authority score was 38. So meaning that the citations that appear in ChatGPT come from very authoritative websites. Once again, makes a lot of sense because to rank well in traditional search, having an authoritative website tends to be very, very effective.

So the other one too is, and this is one I've talked about recently a lot, which is you don't have to worry about nofollow versus dofollow. You don't even have to worry if the link is clearly sponsored or not, or the brand mention is sponsored. We've actually found that you can still influence the AI generated responses just through a number one unlinked brand mention.

04:45

But also even if the link is nofollow, even if the link is sponsored, as long as your brand is appearing in one of those top ranking results or one of those top sources of retrieval, you still can influence the AI generated responses. So that's a very unique characteristic of AI SEO, which is there's a lot of similarities among everything we're doing, but this is one area where clearly it does not matter, and that was not actually the case for traditional SEO. So that's one variable that's definitely unique in this new paradigm that we're in.

05:41

Okay, so another new headline that popped up here was from Liz Reid. She said basically that Google is diligent about low value content, AI slop, even if generated by AI. And she said, not all AI generated content is spam, but the focus is on servicing human voice high value content. So just wanted to bring this one up because I think a lot of people are fearful to use AI content, but at the end of the day, what really matters is that the content, number one, fulfills the intent of the keyword phrase. So whatever that topic is, your page properly fulfills the intent. That's the number one thing. Number two, it's relevant to that topic and you're actually helping the person who's landing on that page to achieve whatever goal they had when they initially made that search. So you're satisfying intent, you're helping them achieve their goals. And then number three, it's actually high quality. It's actually giving value to that person. Now, value is subjective, okay? Value could just be that you're giving them enough information to make a conversion decision that could be considered value. It can also be entertainment value, it could be education value.

06:31

So value is highly subjective, and we're working with algorithms here, and they're not particularly good at being subjective. That's not what they're good at. So when we talk about quality, quality is just that the person who's consuming that asset gets their goal, they achieve their goal by consuming that asset. If you do that, then that's what I would consider quality, and it's highly, highly subjective, but in general, don't just generate AI content and slap it on your site. That's just a bad idea. It's always been a bad idea, and just don't do that. If you're going to use AI to generate the content, just backload all of the additional editorial work once that asset is created. So what we typically do is we'll generate it with AI and then all that time that we save by using AI, now we backload it in the editorial process to make it much better.

07:25

Okay? Just do that and you'll be good to go. I also want to show this one from Tim Soulo, and he was just saying that the era of Evergreen SEO content is over, and Tim Soulo is the CMO of Ahrefs. But I want to show you this one because this, I actually do agree with this quite a bit. I have kind of a different name for it, but I agree with the sentiment here, which is I'm a big fan now these days of what I'm calling minimum viable content. And there used to be a time, and in fact I battled it out with Tim with Ahrefs many, many times for various keywords that were very evergreen in nature. And really the strategy many years ago used to be you would create a really, really epic piece of content. We're talking 7,000 words, 10,000 words, like something really huge, and that works really well for a long time.

08:18

And you could literally build this asset, spend 30, 40 hours on it, publish it, and it would rank for sometimes years before it finally started to fall off. But that's just a terrible waste of time now. That would be a really bad idea because essentially all you would be doing is spending 40 hours to train the LLMs, right? So whatever new model they're going to be training, you're just giving it the food that it needs, so then it can eliminate you. So I don't recommend that. What I would do instead is really kind of what Tim is talking about here, which is focusing on speed, speed of iteration. So for us, a lot of the time now when we publish any asset on a website right now, funny enough, I'm basically working on just one SEO campaign, which is Rankability. But when we publish an asset, we're not publishing an A plus asset anymore, we're intentionally publishing a B level asset.

09:16

And the reason is because I want it to get crawled and indexed and I want to get it into the market and I want to see what happens and see where it lands, see where it starts to stabilize or plateau, and then from there, we're going to start to iterate on that asset until it eventually gets to, if it gets to the goal before it needs to be an A plus, then we'll stop. But if we have to keep iterating on it until it gets to where it needs to be, and there's only so much you can do to any given page, and once you've hit that A plus level, and if it's still not performing, then the next obvious move is to consider the other signals that would influence that page's performance, right? But I'm a big fan of this strategy, which is more thinking about speed, kind of treating your content like a startup, and I'm a big fan of this.

10:00

Okay, another one I wanted to bring up here is you might be seeing these statistics flying around. They're like, AI search converts way better than SEO. And I'm a little skeptical of these claims, especially when there's no citation and no proof of it, just kind of like numbers being thrown around. And I think I've heard before that, I don't know, it's ironically, but it's like some percentage of statistics are just made up, and I think that's the case a lot in marketing. I think a lot of statistics are just made up and then they just get spread all over the place. But it's really, really hard to determine the conversion rate of AI search because number one, even in Google Analytics 4, you could only extract a very small amount of that referral traffic. And this is under the assumption that a UTM parameter was actually added to that link and that someone was clicking on a citation inside of the AI search platform.

10:59

And the way people interact on these AI platforms is that the citations are basically buried and hidden. So it's really hard to even click on a citation to get to a website. So in that regard, the level of friction to actually get a click on one of these platforms is very high. So if someone does click and they do get through to your website, it should technically have high conversion because of how much friction they had to go through to get there. But what's much more likely from a conversion perspective, which is nearly impossible to measure now at this point, is someone goes through an AI chat dialogue and they go through the full, I call it a compressed sales journey inside of one chat, and by the end of it, they're looking for a brand to work with. Now they're ready to work with a plumber, they're ready to buy a product, whatever it may be.

11:49

ChatGPT doesn't really send people to websites. So what it'll do is say, okay, these are the top brands you should work with. And what's more likely to happen is that someone just going to go to Google and do a branded search. That's probably much more likely. And once that occurs, once there's this cross-platform situation, attribution just goes out the window. And unfortunately when you look at Google Analytics, whatever attribution you're seeing is not typically reality because basically, I'll just give you a small example of this right now on YouTube, we publish a YouTube video, people watch seven of our videos, and then one day they decide I'm going to go to Google and check out their brand, see what they're all about. They go to our website and they start a free trial in Rankability. Okay, now who's going to get attribution for that?

12:40

Well, Google. Google will get attribution, Google search will get attribution. YouTube will not get that attribution, even though that was the original source of that person finding our brand. And this happens all the time, and plus there's so many touch points that people are going through. Someone might find you on ChatGPT, and then end up on Google, then watch a YouTube video, then find you on X and then go to LinkedIn, and by the end of it, what is the original origin of that lead? You just don't know. Okay? So whenever I see any statistics talking about hard attribution percentages, I get very skeptical just so any marketer knows that attribution is really, really challenging. But also even the way that the AI platforms are designed, they are similar to social media where they don't want you to leave. They do not want you to leave the platform.

13:32

So they're intentionally designed this way to get you to not leave. So more than likely, if you're going to be seeing conversions, you should be tracking your branded search volume. That's a better metric to be tracking and seeing that that's growing over time because at least you can somewhat deduce and you can track obviously with a lot of new tools, even the one we're building, you can track your brand visibility inside of the AI chat platforms. But in general, branded search is going to be a big one that you're going to want to track, and you can just track that in Google Search Console.

14:38

Okay, now I want to talk about this one today. So this is an article from Search Engine Land about how to plan for GEO in 2026 and evolve your search strategy. So I don't normally do this, but sometimes when you see an article that's just kind of, I hate to say this, but spreading misinformation about what we do here, I think it's important to talk about. Okay, so I'm not going to tear up this whole article, but I want to go through a couple key points that are very important that could lead someone in the wrong direction. So number one, which is, okay, this is a direct quote, by the way, a generative engine optimization GEO isn't just another tactic, it's a new lens, a new mindset that's reshaping how we think about search content and customer discovery. Okay? I'm fairly certain this article was definitely AI written just based on how it sounds, but this is a very important point. Might seem very generic and why would I care about this? But the truth is I would be careful pretending like GEO is a completely separate discipline from SEO. Yes, there are certain qualities of influencing the AI generated responses to include your brand. Yes, there are certain characteristics and variables that are unique to that, but it's not entirely unique. It's not like we're going to a completely different paradigm. I would argue that you could continue doing 100% of what you've been doing for traditional SEO, and that will get you about 90% of the way there across these platforms.

15:42

The other 10% is some of the newer stuff like I mentioned earlier, which is getting unlinked brand mentions and trying to get your brand seated on as many of these sources of retrieval. But even then, that's always been important to do even from a link building perspective. You should have always been thinking about that. So it's just more that it's not necessarily that links are always the most important thing. And then also reviews are big. A lot of third party signals are kind of the newer signals, but in general, the standard stuff that we should be doing, which is ranking in traditional search, ranking many placements in traditional search, building topic authority, getting links, promoting your website. This isn't really that new of stuff, so that's why it's getting a little silly where we're trying to pretend like it's a completely new discipline. It really is not.

16:29

There's just some new variables, but in general, don't treat them as a completely separate entity. And then also it is mentioned that ChatGPT alone is processing 72 billion messages a month. When you use big numbers like this, everything seems like okay, there must be so much demand and so much opportunity on this platform. But the reality is when you look at the real data, this is actually from ChatGPT, they released a paper and they said that they have 700 million weekly active users, which is a huge number. That's absolutely crazy. And they are the fifth, I think actually they just passed Instagram the other day. I think they're the fourth most popular website on the internet. But that doesn't tell you the whole story because when people use ChatGPT, they're not just going in there and searching for "best baseball cleats," okay? That's not the way that most people are using the platform.

17:25

And there's actual data to prove this. This is directly from ChatGPT. This is not some made up data. About half of the messages, 49% are what are considered asking intent. So that would mean treating ChatGPT as an advisor instead of just asking it to write you a letter. That would be a scenario within that 49%. There's a very high probability that there's probably queries or prompts that have commercial intent, but we can assume that just based on the amount of phrase queries that have commercial intent in Google, I think it's between 10 and 20% or something like that. So we could argue that from that 49%, we could say 20% of those have potentially commercial intent. So you can see how that starts to really chop down the amount of demand significantly. And then among the other types of intent, 40% are what's considered doing intent.

18:27

So that means writing a letter or creating some code or creating a free tool. Those can somewhat lead to what would be considered commercial intent, but they're so high up in the funnel, they're so far from a conversion and any type of even brand mention that there's nothing you could really do with those. You can't really do anything with that 40%. And then 11% is what's considered expressing meaning. It's not actually reflection or exploration or brainstorming or just totally useless stuff or I guess useless in the eye of the beholder, but stuff that isn't actually going to drive revenue. Okay? So we're talking about of the many, many searches or many, many interactions that occur on ChatGPT, only a very, very small percentage of those are actually going to have commercial intent. And those are the ones obviously that we want to focus on as a business.

19:23

But let's not overblow it and pretend that all 700 million weekly active users are just searching commercial queries in there. That's not the way the platform works, and that's not even how Google works either. And then another one here, which is, this is a quote by the way, in 2026, EEAT is going nowhere. It needs to be your strategic cornerstone. This one's tough because the truth is ChatGPT does not use EEAT. Let's just start with that. ChatGPT and Perplexity and Claude, which is Anthropic, they do not use EEAT. And there's a very simple reason why, because that is a criteria that was created by Google. So ChatGPT is not going to use a criteria that was created by Google. So that's number one. Number two, Google has stated many, many times that EEAT is not a ranking factor. It is a quality guideline that their quality raters follow their manual reviewers but is not actually baked into the algorithm.

20:30

Now, this one kind of consolidated concept is not baked into the algorithm according to them at least. But I do think there are some elements in the EEAT model that probably do matter as far as experience, which is actually a content related element, which is the first E. I have not found at least, and we're talking specifically in the AI search platforms at ChatGPT, I have not found that experience driven content performs any better than just regular content. I have not found that to be the case at all. And obviously Rankability, we're looking at thousands and thousands of pieces of content every single day. And I have not found that if you say that you tried the pizza versus if you just say these are the best pizzas, that it makes much of a difference. There's no clear quantifiable evidence for that to be the case.

21:23

So that E is basically eliminated. The next E is expertise. Once again, I have not found a correlation here between having like, oh, do they have an author bio or does it say that this is fact-checked? All those little variables don't seem to matter very much. In fact, even on Rankability, our own website, even on our blog, we don't even have an author bio box on our blog, and the author on our blog is literally Rankability. It's like our brand. It's not even like me, for example. It's just like Rankability, the brand. So once again, I don't think that these platforms use it. I think it's kind of made up. I don't think it matters, but I will say the thing that does matter is the A. So of all these little elements, the A in EEAT is the most important, which goes back to what we talked about earlier, which is authority, and that's using SEMrush's metric here, authority score.

27:30

[YouTube SEO section continues... transcript continues with discussion of YouTube ranking factors, relevance engineering, and video optimization strategies...]