← Voices of Search // A Search Engine Optimization (SEO) & Content Marketing Podcast
Voices of Search // A Search Engine Optimization (SEO) & Content Marketing Podcast
SEO, GenAI/LLMs as search, lack of performance data for SEO professionals
00:25:50
Loading summary
Transcript35 lines
- [00:00]
Tyson Stockton
The Voices of Search Podcast is a proud member of the I Hear Everything Podcast network. Looking to launch or scale your podcast, I Hear Everything delivers podcast production, growth and monetization solutions that transform your words into profit. Ready to give your brand a voice then visit iheareverything.com welcome to the Voices of Search Podcast. A member of the I Hear Everything Podcast network, ready to expedite your company's organic growth efforts. Sit back, relax and get ready for your daily dose of search engine optimization wisdom. Here's today's host of the Voices of Search podcast, Tyson Stockton.
- [00:43]
Search is no longer just about Google or once again, it's no longer just about Google. Generative AI tools like ChatGPT are redefining how people find information and or work as SEOs is changing. According to Momentix 2025 State of SEO Google vs ChatGPT report, ChatGPT users are now 2.3 times more likely to click through to external websites than Google users. And while Google still dominates our industry, user behavior is shifting at a very quick pace and we need to take notice of this into how we're competing in search. This is the Voice of Search podcast. My name's Tyson and joining us today is Tyler Einberger, who is co founder and head of growth at Momentiq. Tyler's here to talk to us about generative AI, LLMs and a bit more about the study that him and his company produced. So with that, welcome to the podcast, Tyler.
- [01:44]
Tyler Einberger
Thank you, Tyson. Really appreciate you inviting me here. Super excited to talk about what's happening in search and LLMs.
- [01:52]
Tyson Stockton
Yeah, and I think if I remember correctly, I came across one of your posts on LinkedIn. I was interested in the study, so I gave you a shout out to be like, hey, love to have you on the podcast. As we were talking, I found out you were also talking to Jordan, one of the other co founders and hosts. So I'm happy to have you on here and really diving in a bit deeper into the study that you've put together. With that, let's set the stage for the listeners or viewers as far as what is the study, what's the why behind it and kind of, you know, why did you see to embark on it?
- [02:31]
Tyler Einberger
Yeah, so the report, I like to call it a report just because I didn't put in like really hard. I mean I put in a lot of time but not like a ton of hard, different inputs really. I was using similar web clickstream data and I did do some cleansing of some different tables, but for the most part a lot of the data is straight up just from similar web. So it's a, it's a report. Yeah, I guess it could be like a little bit of a study. But yeah, I think the, the main piece that came out of it was, you know, we, we've been hearing from different people online or from our bosses or clients bosses that like hey, SEO is losing its luster. It's not an attractive channel to invest in currently. And I think there's a lot of doomsday narrative that's out there like Google search is dying or ChatGPT is taking over. So I really wanted to get to the bottom of that question. But then from a broader perspective, you know, I started this research because you know, if we're really heading towards a zero click future and if that's true, I think we need to shift our influence, our areas of influence from owned surfaces to you know, earned or maybe discovery surfaces. So I wanted to map out all of the places on the web where I think three things are true. And the first thing is where are places that marketers can market and do marketing. And then second factor would be what places where that's true. Our audience is also okay with being marketed to or being influenced. And then the third requirement would be like are these places measurable meaning from the outside looking in and available through clickstream data? They have to be browser based.
- [04:40]
Tyson Stockton
I appreciate that kind of almost taking a step back because it's like as practitioners we get so focused into just more of performance within one medium and performance within that one medium isn't the best illustration of what overall impact or performance is. So I think that piece of you know, taking a step back really is helpful. And then really like that's like one of the benefits of qliksync data in my mind. Like I forget how many years back is I got access to a bunch of like jump shot data and I had so much fun just looking. And it was like at one point too, it was just like curiosity things that I was like, okay, well like I wonder how many people went from this to this to this. But I do think like that's helpful to kind of broaden the lens versus just looking at like rank position or something like that. Maybe if you could elaborate a little bit more though on like the clickstream data. So I know you said similar web data was used for this. Like what was kind of like your methodology or approach for this report?
- [05:53]
Tyler Einberger
Yeah, absolutely. Love to dig into that. So we looked at us web users only and the reason why I wanted to focus on users was there are plenty of reports that existed already from similar web clickstream data from other clickstreams that reported on visits. But I wanted to know like audience, like if we're, if we're taking an audience first approach, you know, thinking zero click thinking, you know, the people that we're trying to connect with through our channel. That's why I looked at users. So I looked at them from February 2024, which is as far back as I could get Data for through March 2025. I've updated it since to include April and soon May. But the Data comes from SimilarWeb's opt in clickstream panel which would be like millions of devices and then also supplemented with ISP provided data. We deduplicated some of the audience for for certain views, especially when we're saying hey, this many people or this many clicks per person happen, like that's based on a deduplicated audience. Which means, you know, assuming that, you know, I'm a user and I use Google at work on my desktop computer, but then I also use it on my mobile phone, that would be a duplicated user unless we were to duplicate it. So I think similar web is doing some good stuff there to try to give us more representative, I won't say accurate because I don't think we're in an era of accurate data, but representative data of what's actually happening. And then we took that data. When we're looking at clicks out from Google, clicks out from ChatGPT, I wanted to scrub out self referrals. And for Google there weren't many self referrals except for clicks within platform. But for ChatGPT, because of the URL structure and how clickstream works, I had to scrub out self referrals which was, I don't know, like over 20% of the clicks that happen within ChatGPT.
- [08:01]
Tyson Stockton
That's interesting. Can you elaborate on the self referrals a little bit for the listeners?
- [08:06]
Tyler Einberger
Yeah. So self referral would mean. Well, clickstream is based on like a key piece of ClickStream would be URL, anonymized URLs. So for ChatGPT for example, a lot of the clicks that happen within platform are out to an open AI domain, whether it's to log in, whether it's to switch products to maybe operator, whether it's passing through some sort of. You're logged in, but you still have to go through an off. So those are self referrals that we scrubbed out with Google. We didn't really need to because it was based on domain wide or subdomain wide www.google.com which is 99% probably plus just Google search fair.
- [08:55]
Tyson Stockton
Okay, so diving into the report, I feel like one of the pieces most people want to know is what is. And we said in the intro too of like, you know, ChatGPT users versus Google users who's clicking through. So as marketers we're interested in who's actually reaching our websites. Like what's kind of from the high level as far as volume and also like percentage. Like what how would you find as far as like referrals from each of these sources?
- [09:29]
Tyler Einberger
Something pretty surprising. I think the one thing that wasn't really obvious before finding this was hey, there's a lot more people clicking out from ChatGpt than we thought. And I think that's a big thing. So if we're talking about numbers, what the data is telling us is that ChatGPT is 2.0. The April or the March number was 2.3 times better at sending out people to other websites than Google.
- [10:03]
Tyson Stockton
Wow.
- [10:04]
Tyler Einberger
Which was extremely surprising because Google's main revenue comes from sending people to websites.
- [10:10]
Tyson Stockton
And within this is this counting both organic and paid?
- [10:15]
Tyler Einberger
Yeah, yeah. This is not. Yeah, this is agnostic for sure.
- [10:19]
Tyson Stockton
Interesting.
- [10:20]
Tyler Einberger
So that was surprising. And I think there's important things to consider when we're talking about the accuracy of this data. You know what similar web will say is typical sampling swing is plus minus 30%. If I look at similar web data for a website that I have access to their analytics, it's more like plus minus 10 to 15. So it's a little bit tighter than they allow for. But I treat these numbers directionally instead of absolute truths. But I think the, I mean with clickstream data, not just similar web, there's a really big blind spot and this is going to be website traffic that comes from apps. So I think that's an important thing to consider because if we're just considering website traffic that comes from the open web or a browser based measurable, then we're leaving out a huge chunk of the picture. Especially with ChatGPT on mobile.
- [11:26]
Tyson Stockton
Interesting. So it correct me I'm wrong. So from that within this we would not see if I open up ChatGPT app on my phone and then I'm going and visiting or similarly would it also not capture if I'm going through like the Gemini app?
- [11:44]
Tyler Einberger
Yeah. So Gemini, I can't speak to Gemini because I haven't really looked at how. Well actually I'm not sure if Gemini gives you Links to other websites. I'm not sure about that. But with ChatGPT, yeah, you're sort of right because the intent is actually there within chatgpt to pass along UTM source. It's like the worst. Not the worst, but it's a pretty bad utmost structure that they've implemented. But you know, some, some broad testing on my end and I don't have anything formal yet, but it's like 1 in 10 URLs that are clickable within ChatGPT contain the UTM. So we're talking about 90% of clicks out when you're in, in the app experience are coming through to your analytics as direct because that refer referring source isn't passed through when you're in the app.
- [12:40]
Tyson Stockton
That's, I mean like when I first went through the study too, I was like, okay, the over 2% or over two times caught my attention. I was like, that's interesting. We know it's been rising at like an alarming rate, so it's already on people's radars. But then I'm just pulling up kind of like the report again over here, and if I'm looking at like the April numbers that you shared with me prior, it's like, yeah, Google's still like significantly larger where it was like, you know, 25, I think it was 173 million. And then in chat GBT, I'm looking at like, let's round up like 70 million, but without the app component, like that's going to make it a lot closer because I know just for myself, like I never access ChatGPT through the browser on my phone, it's always through the app. And if you look at like total amount of desktop versus mobile, I would assume, you know, it's not going to be probably over 50%, but I would assume maybe, and this is a complete guess, but I would assume probably around 40%, maybe 30% of ChatGPT's activity is happening on an app base, whether it's app within the operating system. Because I also use, you know, chat GBT within the app quite a bit more than I would in the browser at this point. So I think like, regardless, if anyone was doubting, it's clearly a player that we need to, you know, prepare, align to. Clearly there's a bunch of Google announcements recently which is kind of like following that similar direction. But like, what else within this report too kind of like really surprised you or like caught your attention?
- [14:31]
Tyler Einberger
Well, I think, you know, first to call out like the discrepancy in like app based activity for ChatGPT. Like it's pretty obvious in the data. Like according to similar web clickstream, 82% of ChatGPT users are on desktop. There's no way that's correct. And that's where like the key gap is like really the key gap that we are talking about with mobile app being dark in clickstream data. Like there's no way that 18% of ChatGPT usage is on mobile. Like that's gotta be false. So yeah, to close that loop. But I think that you know, there's the users click out to other websites 2.3 times more than Google does or Google sends. I think there's like Google's audience is still at least six times bigger within browsers than ChatGPT. Again, we're not talking about visits, we're talking about people. I think another thing that's not obvious in the report but still comes out is that Google's US audience is pretty close to the maximum that it can be given U.S. census data on adults in the United States. So it's pretty clear that they have to play defense and to avoid erosion, they have to make sure that people use Google as much as possible. So that could explain some of the AI moves that we're seeing that seem like they're kind of off the cuff or rushed with, with AI overviews, with AI mode launch. Yeah, it's kind of wild. I think another thing that was pretty obvious and I think a lot of people assume this was that Google is kind of forcing people to stay within or nudging users to stay within Google search to answer queries. And that's evident by you know, on average per session. Now this is now looking at traffic, not users per session or per visit to Google Search. Users are now clicking on average over 10 times across devices. Over 10 times per visit.
- [16:48]
Tyson Stockton
Dang.
- [16:49]
Tyler Einberger
Which is kind of wild. And I think two years ago that was under eight.
- [16:54]
Tyson Stockton
Wow, that is a great call out. And it's like kind of similarly that was one of the pieces that jumped out to me is great. Google has this leg up from like a volume again perspective. But it was in the report too. You were showing that it was like thing was like 970% increase of like clicks going out. Which you know, some of that could be user behavior, some of it could be growth to the problem. But like do you have a comparison of like the increase in clicks out versus like user growth? Because okay, clearly ChatGPT is moving more users along. They're also acquiring a lot more users so is it like, is the growth being more fueled from behavior or users, new users?
- [17:50]
Tyler Einberger
Well, I think the third piece that we could add in there is product features. So the most obvious thing that I see is that since the launch of SearchGPT, I think it was originally called, since the launch of that in, I think the public launch was in October of last year, we've seen growth in clicks obviously out to other websites at a higher rate. Yeah, the multi hundred percent increase year over year, I think that's not really indicative or it's not even tied to user growth, more indicative of feature launch. But ChatGPT is growing at a really high rate. I think it's over 80% in users year over year. And then choose your source for just visits growth, which is not included in here. But visits growth is off the charts too. So people are coming back within a month or a reporting period, often to use ChatGPT multiple times in a reporting period.
- [18:51]
Tyson Stockton
Interesting. And okay, so selfishly speaking, have you thought for future iterations of this report? So it's not obviously anything that's already been captured, but have you thought of or have you played around with adding in perplexity? Claude, as I went through it too, I was then wondering, okay, yeah, these are the two primary players, but how about those fringe kind of on the peripheral players as well. I know and also from that like David's a huge clot like proponent. So.
- [19:27]
Tyler Einberger
Yeah, yeah, so I do have this data.
- [19:30]
Tyson Stockton
Oh nice.
- [19:31]
Tyler Einberger
I have not released it yet. Yeah, so this is like I mentioned before, like this is part of like a bigger project and like the biggest thing that stood out to me that I could deliver and provide value to the community was this specific thing. Now we do have other data too and I think other people have called out the fact that Bing Search is really, really struggling right now and that correlates with ChatGPT growth as well as copilot release on a subdomain. So in that data set, that representative group that represents the three, the three things that we're looking out for for, you know, a marketable web, I call it influenceable web right now as my working title for the group of websites. But all the web accessible LLMs are in there, including you know, crossovers like Perplexity, which would be search plus LLM. Perplexity is a lot smaller than we think it is, at least for me, a lot smaller than what I would assume a company valuated at the amount that they're valued at would be in users.
- [20:42]
Tyson Stockton
Interesting. Anything that you can share further about the larger project. Like, I mean, I feel like at this point there's probably a fair amount of listeners that are, you know, intrigued, wanting a little more, like maybe just like setting expectations of when people could expect to see more kind of these other components that you referenced.
- [21:05]
Tyler Einberger
Yeah, absolutely. So I think the, well, one, and I can't remember if I mentioned this before, but the biggest thing for me is that the influenceable web is eroding or decreasing in users by 3 to 4% per year over the past few years. So that's a big takeaway. It doesn't mean that audiences are going away on these places. It means that measurable audiences are going away. I think it means that people are. I mean, the most obvious answer would be that people are just moving to app experience, right? So they're not going to come through in this type of data. But the usage of LLMs is huge. The adoption of anything outside of ChatGPT is super small. Gemini is hard to measure because you could say that, you know, AI overviews are part of, you know, Gemini or AI mode. That's going to come through as happening in the Google ecosystem instead of the Gemini subdomain. But it's a top three LLM and then Claude, which is my favorite, to be honest, in use, personal use is a lot smaller. I think it drops to fourth or fifth place in the list, deepseek, which is one that got a ton of attention after they went viral earlier this year. They've been decreasing at a heavy, heavy rate since that viral peak. So that's something to also consider. I'm curious what happens when they release something notable again, if that usage from people spikes. But then also I think a bigger trend that we're seeing is that overlapped users totally exist. So if you're using ChatGPT, you're really likely to be using other LLMs too. And then if we apply that to Google search, I think a really interesting thing is that, yes, people who use ChatGPT have close to 100% overlap with Google Search. So there's not really any exclusive users of ChatGPT. But if we flip that around and look at Google's shared users with ChatGPT, that number is increasing at probably a scary rate for Google.
- [23:23]
Tyson Stockton
What is it about Claude that you like the best? Because I mentioned to David Bell, one of the other co founders of Pre Visible, he's been banging that drum for a minute and I've been a little more hesitant to move over. What is it about that system that.
- [23:41]
Tyler Einberger
You'Re enjoying more Yeah, I think it's important to call it system because I like the company too. But if we focus on the system, it feels more human to me. So when I'm interacting with Claude, it feels less like a yes man and more like a helpful person. That's the primary thing. And then if you're say you use it to like draft an email or something, I find that the output is. I have to edit it less from Claude's models. And I started really liking Claude around 3.5, which they're now at Opus 4, I believe, or Sonnet 4. So around 3.5, things started to get really good for me and I felt more of a. And I know we're not supposed to humanize LLMs and bots, but it felt more human and it felt like I enjoyed interacting with it more. I don't like the interface more, I don't like the usage limits more, but the output and quality seems to be higher.
- [24:42]
Tyson Stockton
Quality, interesting. Well, I mean, I played on it a little bit, but yeah, maybe I should kick the tires a little bit more. Well, that's gonna wrap up this episode of the podcast. Thanks again to Tyler Einberger for joining us. And I feel like I could have continued this conversation much longer, but I've kept Tyler a little bit too long. If you'd like to get in touch with Tyler, you can find a link to his LinkedIn profile in the show notes or go on over and check out his company's website@momentic marketing.com Also, be sure to check out the kind of local meetup group in Milwaukee, Milwaukee's Digital Marketing club. We're going to be doing some stuff with them in the near future, so if you are in the area, be sure to check them out. And if you haven't subscribed yet and would like a daily stream of SEO and content marketing knowledge in your podcast feed, hit that subscribe button in your podcast app or on YouTube and we'll be back in your feed in the next following day. And that is all for today. Thanks for stopping by the Voice of Search and we will see you in the next episode.
- [26:01]
Tyler Einberger
Sam.