Jon Clark and Joe DeVita welcome back Rand Fishkin, the founder of SparkToro, to discuss his journey from Moz to SparkToro, the evolution of the SEO industry, and the impact of AI on marketing. Rand shares insights on the importance of data analysis, the challenges of e-commerce, and the future of search in a world dominated by Google.
https://page2pod.com - In this landmark 100th episode of the Page 2 Podcast, Rand Fishkin returns to reflect on life after Moz, the liberation of leaving venture capital behind, and his journey with SparkToro. He opens up about the recent Google API leaks, the real impact of AI and LLMs in marketing, and how the digital landscape is evolving toward a zero-click future. With transparency, humor, and loads of insight, Rand dives deep into audience research, entrepreneurship, and what still excites (and frustrates) him in search marketing.
Whether you're a digital marketer, entrepreneur, or SEO nerd, this episode is packed with actionable advice and real talk from one of the industry's most authentic voices.
🔍 In This Episode:
🎯 Rand explains the emotional toll and freedom of leaving Moz and VC pressures
🔍 Behind the scenes of SparkToro’s clickstream-powered insights
📁 The real story behind the Google search API leak and its validation
🔄 Why the SEO influencer space feels stale—and what might change
🧠 Rand's take on AI: useful tools vs. overhyped tech
📊 How SparkToro integrates LLMs to improve audience targeting
🧪 The importance of testing and data-driven marketing experiments
🧑🍳 A surprising story about ultrasonic knives and finding product-market fit
🛒 Smart Q4 e-commerce tips for brands both big and small
🌐 Why your website still matters—but so does zero-click marketing
🔮 Rand's predictions for SparkToro, AlertMouse, and the future of search
From philosophical shifts in business to the power of podcasts in training AI models, this episode is a masterclass in modern marketing.
📚 Resources Mentioned
• Rand Fishkin on LinkedIn: https://www.linkedin.com/in/randfishkin/
• SparkToro: https://sparktoro.com
• SnackBar Studio: Rand's game dev studio - https://snackbarstudio.com
• AlertMouse: Rand's media monitoring tool - https://alertmouse.com/
• Seattle Ultrasonics: Knife startup mentioned - https://seattleultrasonics.com/
• Google Leak: https://sparktoro.com/blog/an-anonymous-source-shared-thousands-of-leaked-google-search-api-documents-with-me-everyone-in-seo-should-see-them/
• IMEC Lab: https://sparktoro.com/blog/imec-lab/
Jon Clark (00:00)
Rand Fishkin co-founded Moz and helped shape modern digital marketing. But today, he's running two very different companies: SparkToro, an audience research tool, and Snackbar Studio, a boutique video game shop. In both, he's trying to prove that you don't need venture capital or growth-at-all-costs to build something that matters.
This episode covers a lot of ground — from the psychological toll of tech’s growth obsession to the quiet power of audience-first marketing in a world driven by AI and zero-click results.
Rand explains why SparkToro avoids the SEO label, even though search marketers love it, and how data from clickstream providers is rewriting how we understand influence. He also unpacks the tension in AI hype cycles, the problem with tracking LLM visibility, and why most brands are failing to future-proof their digital presence.
This was an honest, thought-provoking conversation from start to finish. Let’s get into it.
Jon Clark (00:58)
Welcome to another exciting episode, milestone episode — episode 100 of the Page2Podcast! I’m your host, Jon Clark, and as always, joined by my partner in crime at Moving Traffic Media, Joe DeVita. And today, we're very excited to welcome a guest who sort of started it all way back in 2018 on episode one — Rand Fishkin. Welcome to the show.
Rand Fishkin (01:19)
Thanks for having me back, guys. Great to be here.
Jon Clark (01:21)
So, I’ll admit, I was a little nervous for this episode. I was telling Joe I had to do some jumping jacks before we hopped on the call, just to get out some nervous energy.
But I thought maybe — Joe and I have a long list of questions, and, you know, we won’t get to them all. But maybe to start, we'll go back to 2018, sort of follow up on maybe that conversation.
So, that was right around the time there was a Moz transition, the launch of SparkToro. And, I know, with the evolution of moving from Moz to SparkToro, there were probably a lot of questions around what was negative in that experience. But I’m curious — now that you’ve had some time to look back on that period of time — what was the best part about sort of moving on and sort of starting something new?
Rand Fishkin (02:03)
I did not realize, Jon, how incredibly freeing it is to change your mindset — from one of addiction to growth and financial success at the cost of everything else, to a world where you prioritize what you actually care about. That is awesome.
And, you know, the world of venture capital is not built for that. It’s built for the opposite of that. It doesn’t create good incentives, it doesn’t create healthy lives — it promotes a lot of misery, unfortunately, and a lot of concentration of wealth and wealth inequality.
Which I’m not actually a fan of either, you know, even if I’ve been a beneficiary of it. It kind of sucks. It’s weird, you know? It’s weird to be like, “Yeah, I have millions of dollars,” oh, man.
Jon Clark (02:38)
Okay.
Rand Fishkin (02:45)
“God, there are more homeless people than ever in my city.” You know, like that.
It feels gross. I think you have to disconnect yourself from the emotional and psychological reality of other people and become an unempathetic dick just in order to survive, you know, psychologically. And that appears to be what’s happening with, you know, most wealthy tech folks.
So anyway, freeing myself from that world... I just didn’t know how good it could be.
Jon Clark (03:08)
Yeah. When you’re in it, right? It’s impossible to know until something forces you outside of it, right?
Rand Fishkin (03:11)
Yeah, totally. Totally.
I think, it’s like people who get out of problematic relationships. Like, I’ve talked to folks, you know, of course we’re at that age now. I think you guys are around my age, mid-forties, maybe late thirties, right? But like around that age, you know, people who’ve been in relationships for a long time, like have breakups or divorces, those kinds of things. And you talk to your friends. I think, especially women, probably even more than men. And they’re like, “Oh my gosh", you know, "it’s just so freeing. I feel like I can be myself again. I can explore who I am, I can get what I want out of life. I can prioritize the things I care about.”
You know, I don’t want to say Moz was the same as that kind of thing, but... it was a toxic relationship of a kind.
Jon Clark (03:51)
Yeah, very interesting.
I know you’ve mentioned that you received many, many supportive messages during that time. Did that help you get the strength or the will to go back into entrepreneurship? Because it’s not an easy life, right? It’s tough and lonely at times.
Rand Fishkin (04:07)
Yes, those messages were amazing, right? And I did receive thousands of them after leaving Moz — and have continued to, you know, in all the years since.
And now, for, you know, I get nice messages for stuff with SparkToro in order to the work that I do here, or Indie Founders with Peldi, or even SnackBar Studio.
But to be honest, I started SparkToro
Jon Clark (04:19)
Right.
Rand Fishkin (04:27)
the morning after my last day at Moz. So, you know, we’re talking about like a 12-hour transition period, right?
And I knew that I was going to do it again, mostly because I had to prove to myself that I could. I’m sure you guys have this, right? That fear of, “What if this thing that I did, that I was good at my life, what if that’s the only success I ever had? Or, like, what if I’m a one-hit wonder?”
Jon Clark (04:34)
Yeah.
Rand Fishkin (04:49)
And, you know, "There's can never do it again?".
That was what kept me at Moz for at least the two or three years before my departure. Just that fear, that fear that I’d never be able to do it again, that I’d be nobody without that. This one company I built, that my identity was tied to it.
It was dumb. Like, it was so dumb. I didn’t know how freeing it could be to try again — even to try and fail. You know, like, It’s okay to try and fail, and then try again. There’s no shame in going to get a job. Like, I could have gotten a job, right? There’s nothing wrong with that. Yeah, It's just so intense to have these feelings...
Jon Clark (05:18)
Right.
Rand Fishkin (05:25)
Because, you know, at least with me, and this is why, this is one of my top tips for every other entrepreneur is: go work somewhere else first. Like, go, get a few jobs. Try a few startups. See what that life is like.
Even if you hate it and it’s terrible and you’re like, “Oh my God, I never want to do the things these founders are doing,” that’s great — you know what not to do.
Like you have... I had nothing. I dropped out of college, I worked at two retail jobs, and then
Jon Clark (05:32)
Right.
Rand Fishkin (05:51)
I started the company that would become Moz. Everything I did was learning it for the first time. That is... I don’t recommend that.
Jon Clark (05:58)
Fair enough. I did think one of the things that was sort of interesting — sort of jumping ahead now — was that you sort of removed yourself from SEO, you know, directly, you know, sort of adjacent. And yet, when the Google leak came about, you were one of the first people who was sent the documents, which I thought was both interesting as it relates to SEO, but also, you know, kind of a validation of your influence in the industry, right? which I think is sort of extends from those days at Moz.
Can you go back to — and I might get the dates wrong — but I think it was May 5th, you initially got the documents of 24. And then there was some obviously making that public a couple of weeks later. What was, you know, when you received it, you know, there was a desire to remain anonymous, right? So there were red flags that would make it seem like maybe this isn't real. Like, how did you establish that it was and sort of actually decide to go public with that data?
Rand Fishkin (06:49)
Yeah, well, I'm lucky in that I have a big network of folks from my days at Moz, continuing on today. That network continues to grow. And so I was able to sort of ping a couple of ex-Google engineers that I knew and ask, “Hey, does this look like it was written in the Google codebase? Is this the right format? Are these the right kind of docs, the right tree structure?” All that kind of stuff. And they were like,
Jon Clark (06:53)
Yes.
Rand Fishkin (07:14)
"Either someone left Google and architected this intentionally — which would’ve been... God, you know, so much time it would have taken, or they're like: this is completely legitimate." Like that's the far simpler Occam's Razor answer. And then, you know, I think obviously since then it's been validated many times over by many other people, but yeah, that was quite helpful. I will say, one of the things that frustrates me,
Jon Clark (07:23)
Okay.
Rand Fishkin (07:37)
and you can tell me how you guys feel about this, and maybe I'm wrong, maybe this is misperception on my part, but my sense was that... or my sense is that the world of search marketing has gotten a little bit stale in terms of its influencers. Like the same people who you would see, you know, keynoting shows and on stages and like referenced in blog posts and sort of, you know, most followed on social media, LinkedIn or threads or whatever, 10 years ago. Those are the same people that are there today. You know, and for the 17 years that I was in the field, there was like new people popping into it every year or two. There'd be like, "Oh, this person's really on the rise", "Look what this amazing person figured out", like all these new entrants to the field. I was a new guy, right? Like when I started in 03, people were like, "Who is this newbie?"
Jon Clark (08:08)
All the same people.
Rand Fishkin (08:26)
You know, and, now it, just feels like that, like they're not that there hasn't been anyone new there. I'm sure there've been, you know, dozens, but the, like the big names that you hear in the field are, are big names that I still recognize. And that bothers me. That worries me. And I think what you're talking about where, where, you know, I was the person that Erfan, the guy who found the docs, you know, basically alerted me, and Mike King, and everybody else.
Jon Clark (08:47)
And.
Rand Fishkin (08:52)
He reached out to me because thought I was the person, you know, and I'm just worried that that hasn't changed. I don't know.
Jon Clark (08:56)
Right.
Joe DeVita (09:00)
I think there was a long period of time where SEO didn't change so rapidly. it's hard for a new person to break in and say, “I’ve got this incredible insight.” For the last five, maybe eight years, it's all been very tactical. And now it seems like it's becoming strategic again. Now it feels very exciting. But I think you're right. Before 2024, it was just very slow and steady, incremental change, same voices, you know, and trust. But it seems like way different now.
Jon Clark (09:29)
I was going to say the same thing. It seems like a lot, it's hard to establish yourself as an expert because a lot of the core foundational principles, I don't think have changed that dramatically, until maybe the last year or so — with a lot of these leaks in AI, and LLMs, and maybe some new things that are applicable there. So I think the hope is that with more of these new platforms to optimize around, like search everywhere optimization, as you've called it, like maybe there are some opportunities for folks to sort of step in and become some of those newer faces and roles. I think, yeah, it's funny.
We're always looking for guests for the podcast and a lot of them sort of default to a lot of those same names, right? And then, and I've even reached out to folks who work in house and maybe aren't on the speaking circuit and things like that. And then their company shut them down sometimes, like they're not allowed to come out and, or they have to be, all the questions have to be reviewed and everything around AI gets struck. And then it's like, okay, well, what's, what's the conversation? So I think.
Rand Fishkin (10:11)
Yeah, right?
Joe DeVita (10:29)
Not interesting. Yeah.
Rand Fishkin (10:31)
Yeah, what's the point? Yeah, that corporate thing is honestly, I think they shoot themselves in the foot PR wise. And what's so interesting to me is the corporate obsession with limited communication absolutely wrecks their opportunity to be in AI answers and LLMs and stuff. Of course, the corpus for LLMs,
Jon Clark (10:31)
Yeah, so I think it's difficult sometimes.
Rand Fishkin (10:55)
even the retrieval of augmented generation stuff, all of that relies on: "Hey, is this person or this brand commonly associated with like these words and phrases all over the web?" And I can tell you guys right now, even, even this podcast, like the transcript of this podcast will get picked up by, you know, Google and Gemini by, by OpenAI, ChatGPT, like, and all the other ones, right.
And it will become part of the reason that people associate, you know, Rand Fishkin or SparkToro with audience research or audience research software, audience research tools or whatever, right? It's like, cause podcasts and YouTube appearances and speaking at conferences and events and like, you know, putting, getting press, all of that stuff is what gives you the ability to be in those answers. And so they're just...
They're all so obsessed with it. Like they're so obsessed with it. And yet they won't do the thing, like the thing that lets you be in that is just, it's wild. The irony is crazy.
Jon Clark (11:43)
Exactly. Thank
Joe DeVita (11:50)
I've always, I've always wished there'd be like a 24 hour news channel of just marketing and it could go on all the time. It's like global, at night in the middle of the night. It's like what's happening in Japan and other places, but marketers are so secretive. Like if it's not their corporate PR saying, "Oh no, you can't say that." Just people just hold secrets. It's like, you know, it's like you're never going to get like Procter & Gamble to say: "Well, this was a test we ran and it was hugely successful".
They don't want their competitors to you know. I feel like that's why it's mostly talking heads and agency people that can speak in generalities who get the exposure.
Rand Fishkin (12:29)
I mean, this is why I love third party data, right? Because if Procter & Gamble is successful, Datos, and Comscore, and Nielsen, and Similarweb are going to see it right in the, like in the clickstream data panel that, you know, from all the devices, they're going to be like: "Yeah, look at this, this many, you know, an increase of this percentage of people made it to the checkout page on like the Amazon URL for, you know, Dove soap" or whatever, you know, whatever product they're selling.
You can't hide from clickstream data. And so you can see the impact of that stuff. I love watching that after like a big outdoor campaign or Superbowl ad or, you know, whatever. And being like: "Oh, yeah, everybody's talking about how this ad is so successful." And you go and look at the numbers, like: "No, that didn't work." You did not experience, you know, sales lift over the following three months.
Jon Clark (13:05)
Maybe that's a good segue into SparkToro. So, SparkToro, correct me if I'm wrong, runs on clickstream data. Like that's how you put the graphics together. I'll be honest, I just recently started playing around with the tool. Maybe just for the audience and folks who haven't used it, can you break down sort of how SparkToro collects that data and then sort of translate that into graphical form?
Rand Fishkin (13:24)
Sure. Yeah. So we actually buy our data from companies like the ones I mentioned. mostly, I think exclusively right now, we're buying from Datos, which has a... I think I saw their panels now, 20 million, 20 million devices worldwide. So, you know, obviously there are whatever, 5 billion internet users, right? So they're getting a small percent of the world's internet users. I think the panel is 10 million ish in the US right? So a good
Jon Clark (13:54)
Very good.
Joe DeVita (13:55)
Representative.
Rand Fishkin (14:04)
percentage share and what they do is they, you know, get people to opt in, you know, through programs, right? It would be like: "Hey, you know, you're gonna, whatever... download this piece of software onto your phone or you're gonna use this app on your desktop" and like: "do you consent to doing this" and sometimes there's like, you know, an incentive around that, where they're like: "We'll, whatever... pay for your internet for six months or something, and like, you give us your anonymized clickstream data", which is basically every URL that you hit in a browser. So that obviously has some frustrating exclusions. Like we can't see if somebody visits a page in the Instagram app on their mobile phone. But we can see if they visit that page in their desktop browser or in their mobile browser, you know, in Chrome or whatever. And when I say we, Datos sees that they anonymize, they aggregate.
Jon Clark (14:43)
And. Thank
Rand Fishkin (14:53)
They build up like sort of their demographically, you know, constructed panels so that there's not bias and, you know, like, we have whatever, a bunch of people from wealthy counties in the Northeast. So we're going to like lower that in the panel and we're going to raise up. Like we, need more sort of, you know, rural panel stuff from the South or the Midwest or whatever. So they'll, balance all that stuff out and then they essentially sell us numbers.
Like we, we, what we buy from them is, data about, you know, how many people visit certain websites cross domain visitation. Like, people who do X also do Y. Right. So if you, whatever, if you visit the Page2Podcast and like, you know, listen to this podcast, what other websites do you visit in like in higher proportion than average than, the rest of Americans?
And so then when you go and you ask SparkToro, like: "Hey, my audience is the people who listen to the Page2Podcasts and they're super interested in digital marketing and SEO and whatever", SparkToro will then go build an audience based on that and then give you data about what they read, watch, visit, listen to, subscribe to. It's these YouTube channels, it's these other podcasts, it's these websites that they visit. They're using these AI tools more than average. They're using these search tools less than average, they're using these social media networks, like all that kind of stuff. We don't do what we just talked about in the previous question, which is we don't say: "Here's how many people buy Pantene Pro-V conditioner on Amazon". But Datos does sell that kind of data to lots of people in financial markets and e-commerce players and brands,
Jon Clark (16:07)
Yes. Okay.
Rand Fishkin (16:30)
sort of SimilarWeb, sort of ComScore, sort of Nielsen, all these folks.
Jon Clark (16:34)
Got it. There's been a whole bunch of news around Google removing the ability to expand results to 100. My initial thought was, I wonder if SparkTower is being influenced by that. And then I realized, no, they're collecting user data. So they're sort of insulated by that impact.
Rand Fishkin (16:41)
I saw that, yeah.
Well, so he, there's one thing conceivably, which is the pages per visit to Google might go up, right? Cause like you can't, if your default was I like to search with a hundred results and now you can't, now you have to only get 10 results. You want to click to page two, like the podcast... Hey, the Page2Podcast just became twice as relevant. To go to page two, page three, page four, right? Like you're clicking, and so those clicks will mean that Google
Jon Clark (17:00)
Sure. Ha ha
Joe DeVita (17:11)
Ha ha ha
Jon Clark (17:12)
Perfect.
Rand Fishkin (17:18)
gets more browser time than it did historically. We're looking at visitation data, like visits to a site. We don't care how many pages are browsed. But other tools and data providers might show interesting things around that. So it does have a change.
Jon Clark (17:33)
Got it. You've been a little apprehensive of AI, I guess is a fair way to say it. And I think yourself and Brittany Mueller also sort of talk about, there may be sort of this negative halo effect that could be coming with this over investment. You guys integrated some AI features into SparkToro. Do you see?
I guess I'd love to get your thoughts on where you see AI as useful versus maybe that potential overinvestment. I guess there could be a thin line, but also a wide gap there too.
Rand Fishkin (18:02)
Yeah, it's really interesting because I think, what I am is I'm skeptical of the hype and the hype cycle more than the technology itself. For example, AI is built on and it's basically just an advancement of machine learning, which I was a huge proponent of and loved and evangelized for years when I was at Moz. We used machine learning for all sorts of things
Jon Clark (18:07)
Hmm.
Got it.
Rand Fishkin (18:24)
trying to reverse engineer which elements in Google were influencing the algorithm more, how we built domain authority and page authority. Like, we use machine learning for all this stuff. And I was like: "Man, more marketers need to you know about machine learning, they should be embracing it and using it, more technologists should. So I'm not against it. It has its uses. It's really great for some stuff. The problem is people are extremely confused about how AI systems work. I don't want to...
Jon Clark (18:24)
Bye. Okay.
Rand Fishkin (18:49)
People who are listening
Jon Clark (19:10)
It really is.
Rand Fishkin (18:50)
to this podcast probably totally get it, but a ton of people think it's just magic, you know? Or they think that they're thinking machines. They assume that this is actually, whatever, the movies and they created that Hollywood artificial intelligence and nothing could be further from the truth. It's just spicy auto complete. And sometimes you really wanna know which words commonly come after other words in documents,
Joe DeVita (19:09)
Ha ha ha!
Rand Fishkin (19:16)
you know, on and off the web. Sometimes that is incredibly useful. It can produce things and give you ideas and let you brainstorm stuff and, you know, expand on concepts. Awesome. It can, it can help you write code, it can help you debug code, it can sometimes suggest medical treatments that might be, that you might've overlooked. Like, it can do all sorts of things, but it is not a thinking machine when people assume it is.
Just, it really, really bothers me. And of course, the AI tools themselves have this huge incentive to make themselves, it seems like more than they are. And the press and media are often to blame for, you know, repeating their marketing messaging uncritically. So anyway, that, that's sort of my long winded rant about it, but do I think it can be useful? Absolutely. For example, you know, we were talking about, like, how SparkToro works, and if you were to type in, right, one of the things you can do now to SparkToro is just describe your audience in natural language, the same way you might describe your audience to me in a phone call.
And you can just type that right into SparkToro, and then we use AI to say: "Hey, take this prompt, essentially right, this description of an audience, and turn that domains and other and demographic attributes that are most likely apply to this audience and expand that" and then we turn that into the thing we send to our algorithm to get clickstream data and visitation data and all that kind of stuff, and it works awesomely well, because it turns out stereotyping a description of an audience. Is something AI is really good at like it's just really good at that. There's a great use case.
Previously if you wanted to do that, you would have had to describe it to like me, or Amanda, or Casey, and then have us go: "Okay, we get it. Now we're going to manually input what we, you know, sort of the statistical, you know, behaviors and then get the output". Now AI can do that for us. It's great.
Jon Clark (21:10)
And that's the real, sorry, go ahead. I was gonna say, and that's the real unlock of LLMs, right, is the speed to the result. Yeah. Go ahead, Joe.
Joe DeVita (21:10)
Can a young company, go ahead, Jon. I was just gonna ask...
Rand Fishkin (21:12)
Sorry, go ahead.
Joe DeVita (21:20)
Can a young company use SparkToro if they don't really you know who their audience is yet? Can you help them build the persona for where to start? What's the minimum they need, I guess?
Rand Fishkin (21:30)
Yeah. The answer is technically yes. Like you, you can do that. I just worry when I recommend people do that. Actually, I recommend against people doing that. I strongly suggest that you go figure out who your core customers are before you come to SparkToro, because otherwise you might make assumptions. You might say: "Oh, well, the same people who visit like these websites or who have these characteristics...", I don't know, you're selling a, you know, my buddy Scott, just started an ultrasonic knife company. He's an inventor, culinary inventor. He did like the first sous vide machine, home sous vide machine, and a bunch of other cool stuff. He was at Modernist Cuisine for a while. And he basically invented this technology that passes an ultrasonic current or wave through the knife blade of a chef's knife, a home chef's knife. And so it, you know.
It becomes insanely sharp, even if the blade itself is not very sharp. And so, you know, it's this crazy cool technology. He's miniaturized it from like industrial settings into like, into home settings. He just launched it last week. If he had come to SparkToro and said: "Hey, my audience is, professional chefs in restaurant kitchens who do, you know, an incredible amount of prep and they tend to run their own restaurants or they work for restaurant conglomerate groups, generally higher priced restaurants", right? And then he could... you could search for that audience in SparkToro, and SparkToro could be like, oh, boom! Here are all the outlets you want to hit: this is where you should do your PR, these are the YouTubers to invite, these are like that influencers, you know, here's the podcast you need to get on YouTube... I would strongly recommend to get that, because if Scott had done that, he would have frustratingly found that professional chefs are not the first early adopters.
Jon Clark (22:50)
Hmm.
Rand Fishkin (23:09)
Like, professional chefs have been training, have been using their knives and their structure for generations. Some of them will think it's cool. Some of them have, right? And there have been a few who bought it. But, like, the early adopters for Scott's tech has turned out to be passionate home chefs, knife collectors, people who are obsessed with the best tests, like the knife sharpness tests, and like who love having that kind of stuff, some like home bartenders.
It's a lot of folks kind of like me, actually, more so than it is professional chefs in kitchens. Maybe that'll change in the future, but the problem would be like, assuming you know who your audience is before you actually validate, right? And so what Scott did that was really smart is he went, nobody likes this process, but he went and interviewed. He was like: "Hey, can I come by and like show you this?"
And did that over and over again with like hundreds of people, probably a thousand plus people, and then he saw who was crazy into it, who was like: "You need to email me the day this is out", "I want to buy it now", "can I invest?" like, finding those people told him who his audience and customers were. Then he ran those people through SparkToro and that's how he's like, you know, getting out. So now, you know, the people who are doing publications is like Oh, it's like the GeekWire in Seattle, and The Spoon, and CNET, and Eater, and like, you know, it's not whatever, you know, restaurantchefs.com or what have you.
Jon Clark (24:32)
There was an Anthony Bourdain episode where he was talking with a chef and he showed him his knife and it was almost, you know, it then whittled down from here down to like this because he just sharpened it and that was the only knife he used. Yeah, it makes sense, Chefs like what they like, right?
Rand Fishkin (24:46)
Yeah, I mean, right? Like
I like what I like too. It would be very hard to convince me to whatever, switch away from using OBS to some new software, or switch away from Wistia, or switch away from ProfiWell, or, you know, whatever it is. You've got your... WordPress. I know I should get off WordPress, okay. Like, I know, it kind of sucks. I know, Mullenweg is kind of a troll now and I should get off WordPress, but I'm so comfortable with it. I've write... I don't even... I haven't even explored.
Jon Clark (25:02)
Ha ha ha.
Where would you? Where would you go if you got off WordPress?
That's the thing.
Joe DeVita (25:13)
I'm going to try to,
Rand Fishkin (25:13)
Joe, no, go, yeah, sorry.
Joe DeVita (25:15)
I have a good, like these entrenched customer consumer behaviors that we all kind of, everyone has them, right? So maybe I can bring us back to use of AI a little bit here because I think Jon and I were listening to the webinar that you did with the Datos team about, you know, the first half search and how, you know, there is an increasing number of people who are using LLMs. There's even some anecdotes you'll hear about people who just don't use Google anymore because they can only, they get everything they need from Perplexity or whatever. But it's small. They're probably outliers. Google's still very strong. What's it going to take for an LLM, what's it, what's it going to take to change that consumer behavior from relying on Google for answering all their questions to start relying on a ChatGPT or Perplexity?
Any thoughts on that?
Rand Fishkin (26:04)
Let's see. I think if you go back in time and get the judge in the DOJ Google case to, ⁓ no, no, this year, the one that was just announced like a month ago and, get it, get the judge to actual remedies to prevent Google's monopoly domination of search. You might, you might make a dent. Otherwise I don't, I don't see it happening. I think it's possible
Joe DeVita (26:11)
2020.
okay.
Rand Fishkin (26:28)
that a ChatGPT or looks increasingly like perplexity will not be at their growth rate is slowed considerably. They're having a tough time, you know, breaking out maybe Gemini. Gemini is growing nicely, but sorry, not Gemini, Claude, right? Gemini is Google. But the problem is that Google owns everybody's home pages, owns everybody's default behaviors is now allowed to continue, you know, paying Apple to be the default, paying Firefox to be the default. They, they have two decades of user behavior, which as we found out in the DOJ case is the number one ranking factor now, right? That Google essentially looks at what do people search for where, how, in what order, what do they click on? What do they click the back button on?
What do they stay on? Like that's their bread and butter. And nobody else has that. ChatGPT doesn't have it. They can't buy it. There's no way to get it. Google's been told that they'll have to do limited sharing of some data with rivals. I strongly suspect that will not make a big difference. And then I also think that Google is, because they still send traffic out, granted, an ever decreasing amount.
But Google sends, gosh, 2000 times the amount of traffic out that ChatGPT does. It's just really hard for like ChatGPT to say: "Hey, you should let, every website, you should let us crawl you and, like, stop blocking us". Nope. Right? Like websites have basically said: "Screw you, pay me. Maybe I'll let you crawl me". But Google is allowed to crawl everything for free. Right? In fact, Google...
It costs you money to have Google crawl your site, right? Like you're paying the bandwidth on that. I just don't, I don't see how someone is going to overtake them. I think maximum we're looking at, you know, in the next five years, maybe, maybe Google loses 5 % market share. That's at the, at the outside limit of what I can imagine right now, unless something crazy different happens with tech, or some new, you know, maybe a new administration comes in and a new case is launched against Google. You know what? I'm not a fan of the current presidential administration of the United States, which, probably surprises no one, but I will say they're very vindictive. So if Google pisses them off, it is not impossible to imagine a future where like, Trump is like: "Screw Google", "I hate the Google", like, "no Google for anybody", right? And then, and then I don't know, like at the ISP level, the US government does, like China did for a long time and just redirects all Google traffic to, I don't know, Grok, you know, like whoever, whoever his buddies are.
Jon Clark (28:41)
Oh my God. So you had a really good post on LinkedIn about how you could potentially start to get a sense of what ChatGPT results look like, where let's say you had 100,000 people across the world all do the same search, right? And what those variances are? Like, how important do you think that type of data is? In the light of
Rand Fishkin (29:07)
Yes.
Jon Clark (29:13)
you know, maybe ChatGPT never goes over like 5 or 10 % of search volume?
Rand Fishkin (29:19)
Well, so I think the interesting part about it is different for every different company. So, even... let's imagine a future, like let's fast forward 10 years and ChatGPT is, I don't know, maybe twice the size of Bing. They're between 5 and 7 % market share of all searches, but people use them for a lot of things that aren't just search, right? So there's lots of like tasks and processes that people run through ChatGPT. I mean, they, you know, in fairness, they get 2.5 billion prompts a day now or something I think they announced.
Jon Clark (29:50)
Right. Yeah.
Rand Fishkin (29:51)
So like and, you know, the average session is like eight prompts along so, you know, if you were really going down to it, but I think the thing I could fathom is that companies, some companies in particular, maybe it's like B2B SaaS companies or HVAC suppliers or, I don't know, science fiction authors like they find that their audiences are really using ChatGPT. And so they care a tremendous amount about their brand being presented in the results, right? I think media hype is skewing too many people to care right now, but some people should probably care a lot, right? For some people it matters a huge amount.
And I was, I would also say, you know, ChatGPT can give you one output, right? They can say like: "Hey, here are 5, whatever, people we'd recommend to redo your walls at your house" and then you perform 20 Google searches off of that, because you're like: "Oh, I'm going to check out every one of these. I'm going to look for their website. Then I'm to Google for their reviews. And I'm going to Google for alternatives. Then I'm going to Google and see if they like do this particular kind of thing. If they service my neighborhood". I'm doing a whole bunch of searching. So like, ChaGPT can be the progenitor of a huge amount of search. You guys saw the, there was like that Semrush study. We could, we could try and link to it in the, in the show notes or whatever. But the Semrush guys use the Datos data set and they analyzed what people do after adopting ChatGPT. And you can see that their number of Google searches goes way up.
Like as soon as they start using ChatGPT, they Google way more times than they used to. And I think a big part of that is because you are getting answers that then lead you to more questions. And those questions are often better answered by, by Google. So I don't think it's like a world where people won't or shouldn't care about that. I think they will. I think the problem is like, is that actually trackable yet? You know, is it meaningful at all? For example, you know, a lot of these tools that supposedly track your LLM results. Do you guys use any of these?
Jon Clark (31:46)
Not yet. A very limited capacity. Usually it's on a client request basis, we don't trust the results yet or even you don't know how to really apply the results because we can never replicate it. and so I don't... Yeah.
Rand Fishkin (31:55)
Yeah, thank you.
Yes, this is how I feel exactly. like I'm actually buddy. But, like, I know a couple of guys at Gumshoe, I've had like people from Profound reach out to me. I know Semrush is offering something on this, offering something like... there's all these people, right? Who are investing in: "Hey, we're going to track your LLM results or whatever". Great, but running a prompt against an LLM, you know, once every week or once every month and then telling me where my brand came up or if it did. I just can't tell if that is meaningful at all or actionable or worthwhile. And people are putting so much... This is the thing that pisses me off. Jon, Joe, it boggles my mind how much energy people are putting into the tracking of it, how, much companies are obsessed with tracking of it versus how little they are willing to do anything to try and influence it.
Jon Clark (32:46)
All right.
Rand Fishkin (32:47)
What the hell is that about? Like, do you guys remember the early days of SEO, right? It was like, well, we can't really track, you know, our Google rankings or whatever, but we're gonna, we're really gonna try to get it at the top. And then later, after we try to get ranked to the top, then we're gonna see how well we're doing. Why is it this way with LLMs? I don't understand.
Jon Clark (33:05)
Yeah. We interviewed Marshall Simmonds a couple episodes ago. And when I was preparing for that podcast, I came across an email thread for the IMEC lab. You remember this? And I think the last test that we had run was trying to determine if Gmail actually crawled links in Gmail ads. And at the time the answer was no.
Rand Fishkin (33:07)
Ha ha ha. Yeah, I remember that.
Jon Clark (33:27)
I think we probably assume that's not the case anymore. But when you were sort of thinking through like how you would potentially build a cohort of people to actually get insights on how varying the prompts, or like how often they change in a certain time period. Do you ever like miss those like testing groups, you know, like creating those tests and getting the results?
You're analyzing so much data even today, right? Like that was what you've always done and done it very well. But I feel like that testing component has been missing from our industry for quite some time. It'd be amazing to evolve it to sort of some sort of LLM testing group.
Rand Fishkin (34:05)
Yeah.
I'm always surprised how few people do this because I love it. So I stopped doing the IMEC thing, I M-E-C, that like labs project, post-Moz, but I still do it quite a bit. We just ran something last year where we asked a big cohort of a few hundred people to click on all the social media links. So like we created a, you know, a tweet from a profile that that hadn't existed before, all this kind of stuff to see whether Twitter passed a referrer string, right? Whether it would tell you, hey, this visit came from Twitter, how consistent it was, all that kind of stuff. Twitter, by the way, does pass a referrer string reliably. Facebook does not. LinkedIn does not. So like, you know, it's super weird. I think in our tests, like 15 or 20 % of LinkedIn clicks came with no referrer string, but the other 75 or 80 did. Why?
Jon Clark (34:50)
Really?
Rand Fishkin (34:53)
Who knows? Question mark. What's LinkedIn doing? I can't figure it out. There's no pattern to it. wasn't like mobile does, but desktop doesn't, right? But I love doing tests like that. I think the LLM world desperately needs one. I want to run this panel like we're talking about. I'm actually talking to a buddy of mine who's a software engineer who can like help, yeah, help do the analysis of it. I think look for an email in your inbox because I'll probably ask you guys to contribute.
Jon Clark (35:17)
Amazing. Amazing.
Joe DeVita (35:19)
You,
Jon Clark (35:19)
We'd be happy to.
Joe DeVita (35:21)
I know you're, the NDA that you have signed has kind of pulling you a little bit out of the world of search marketing last handful of years and it's coming, you know, I think that it, expires soon. I think I heard you said maybe this year in a podcast.
Rand Fishkin (35:34)
Yeah, this, think this year, maybe June, my non-compete expired. Yeah. So I still have the NDA, which is in place for as long as I don't know, maybe forever around the transaction, you know, when Moz sold, but I have, I am technically able to accept money to do SEO things now. I just kind of don't want to.
Joe DeVita (35:44)
Well, so, all right, that's my question. Maybe that's your answer. But are there conversations that you've shied away from, maybe because of that NDA, you're kind of excited to jump back into when that expires?
Rand Fishkin (36:07)
Ooh.
I mean, one thing, you know what's nice is we don't have to be as obsessively worried about features we had to SparkToro. So like for, you for a while we were like, oh man, is there a way that this could help with SEO? If there is, we really should like not build it. Now I think we can more reasonably do things like there's this new...
It doesn't exist in the product except for me and Amanda and Casey, but Casey's like built this thing again using AI that basically takes the results from a SparkToro search, so an audience analysis, and then passes it to an LLM with a prompt to take further action. So if you were like: "Hey, I want to analyze the content on my website homepage and tell me whether I have missed"
You know, key aspects that would appeal to this audience for, like, landing pages. Maybe that's more CRO than SEO, like, conversion rate optimization, but those types of things are things that we want to enable in the future. We already you know that we saw like 25 % of our customers exporting. Like they just go to SparkToro, they export the CSV, then they upload it to ChatGPT or Claude. And then they do a bunch of stuff with it. They're like: "Hey, you know, help me make a better landing page for this product" or "help me like optimize my PR campaign" or "help me with messaging for my content marketing", you know, whatever it is, come up with a bunch of social media post ideas that would do well with this audience. Like, and so we basically were like, oh man, we should just help people directly with that. Right. If we can do some nice like prompt engineering and get consistently good results. We can just send it to the API of the LLMs, return the results, show them nicely in the SparkToro interface, and probably save people both time and energy, but also give them something more useful from the results.
So that thing is something I no longer have to fear doing. It's nice to be out from under those. This is one of the reasons, not to bring it back to politics, but like I was so excited when the Biden administration got rid of non-competes. Like they basically said, hey, those are illegal, that is interfering with, people's ability to make money and they're in commerce and all that kind of stuff. And then of course, the rule got rolled back and it just makes me sad. It's like, surely, surely there must be something good this administration is going to do. Like when's it coming?
Jon Clark (38:17)
Well, maybe to ask that a little bit differently, I mean, I think SparkToro definitely has SEO use cases. Does that open up the marketing of some of those features in the application?
Rand Fishkin (38:30)
Yeah, conceivably. I don't know how you guys feel about this, but my concern with marketing a product is that when you start to say: "Hey, this product does everything".
You really become... you lose your specialization and you lose your ability to do hyper-targeted high quality marketing. Like I think that's way, way more difficult, especially for a small company. Like SparkToro is three people, you know, we're under $2 million a year in revenue. Like we do not have huge budgets to be thrown around. We don't have a big marketing team. So we have to be extremely judicious with everything that we invest in.
And as result, I want to focus. I really want people to, when they say: "Hey, where does my audience hang out on the web?", "What are their sources of influence?", "Where can we go to reach them?", "What content should we create?", "What PR should we do?", "What influencer marketing", "What social media marketing?" I want them to think of SparkToro. I don't want them going, isn't that a kind of an SEO tool? And like, we're not really.
Jon Clark (39:09)
Yeah.
Rand Fishkin (39:29)
I don't want them thinking of us as a competitor to Semrush or Ahrefs or Moz. Like we're not. You can't substitute what we do for what they do or vice versa. I think that's actually a really good thing. I don't want to compete with them.
Jon Clark (39:40)
Yeah, for sure. They have a totally different marketing budget cohort there that you don't want to be competing with.
Rand Fishkin (39:45)
Yep. And there's, there's already so much confusion speaking of large language models. Like when people ask about top SEOs, I'm always on those lists. And so people think, even though it's been God, almost eight years, right? So people think, Rand Fishkin SEO. So SparkToro must be SEO. So let me try to use SparkToro to do SEO. That's a terrible idea. Even by saying SparkToro and SEO, I'm probably shooting myself in the foot.
Jon Clark (39:59)
BOOM
Rand Fishkin (40:11)
Cause this podcast is going to get picked up by all the large language models in front of the room. Anyway, just for ChatGPT's sake, SparkToro, not an SEO tool. Not an SEO tool. There you go.
Jon Clark (40:21)
Before we jump to rapid fire, I had one other question. We're coming up on Q4 very, very quickly. You had some great data from Datos, and there were couple slides around e-commerce, Amazon obviously the lion's share. Do you have any thoughts on how any of that data could be applied to someone who is planning or trying to finalize up their Q4 planning for e-commerce specifically?
Rand Fishkin (40:32)
Let's see. So when you look at the e-commerce platforms that are, that are doing quite well with American consumers or, or European ones, you know, you're going to see a lot of names you recognize: Amazon and eBay and, Shopify, and TEMU and all, you know, Walmart, etcetera. Where the way I think about this is if you are big, really, really big and you have good margins and you have a strong brand, it pays to probably put your product in the places where your audience is already going. Right? So if we're talking about a brand that's doing north of $50 million in e-commerce revenue, I would be thinking about your Amazon presence, your eBay presence, your Walmart presence, potentially, right? Depending on your brand
As being as important as your homepage and your product page and your, you know, checkout process, because you really need to think about, how consumers are finding you. I, you know, there's like a skincare brand that I used to buy that exclusively, was exclusively available on their website. I can't actually remember the name. Now it started to be available on Amazon cause they've gotten really big. Which makes sense. And I am ashamed to admit that because of the free shipping, and the faster delivery, and the fact that the prices are a little bit lower. I've been using Amazon, right? Like I just add it to my regular order. I'm, you know, I've got Prime like blah, blah, blah and I would, I would think about that.
Jon Clark (42:08)
Right.
Rand Fishkin (42:10)
But if you're smaller, if you're a smaller brand and your niche, Amazon kind of sucks for you. Like in a lot of ways, they take a big cut, they present alternative products and competitors up against you, they undercut your brand's perceived value and exclusivity. There is a superpower in being like: "oh you can't find this on Amazon. You have to go to their website". That really means something. Was that... There was that big olive oil brand that launched, I think Graza or something like that.
They did huge numbers, like absolutely massive success story in consumer goods space. And their strategy was were available in retail stores that often, that rarely sell food. Like, that was their initial thing. They went to boutiques and like clothing stores and jewelry shops. And they're like: "Oh, what's this olive oil?" "It's very special. You can only find it here in a few other boutiques". Now you can get it on Amazon. Like it's huge. It's big.
Jon Clark (42:43)
Thank
Rand Fishkin (42:56)
You can find it at Whole Foods, whatever. But that exclusivity when you're small makes a big difference. And I would think about that. You know the answer to every marketing question, Jon, it depends.
Jon Clark (43:03)
Yeah. Thanks.
Joe DeVita (43:08)
I know we only have a couple more minutes left and you can treat this next question as rapid fire if you want to, but it's kind of a prediction question. The importance of the website in the next five years. Do you think it changes, you think... a lot of commerce still happens on your own website, but a lot of it moves to Amazon, Walmart and other places. Does the importance of a company's website become less important in the next five years?
Rand Fishkin (43:32)
Relatively speaking, yes, but from an absolute value, no. By which I mean, do not under any circumstances, start investing less time and energy and importance in your own website, but also make sure you are investing lots of new time and energy into your presence on Google's homepage, and in their AI overviews, and in what ChatGPT says about you, and in, you know, what... how you show up on, if you're in e-commerce, right? Amazon and eBay and all these other places. Like you have to now curate all those other things because we are moving to a zero click world and a zero click web. So you got to do zero click marketing.
Joe DeVita (44:07)
We need you for another hour. We need you to cancel your next meeting, Rand. We've got too many questions.
Jon Clark (44:07)
There you go.
Rand Fishkin (44:12)
I know. All right, well, I'll come back on the podcast. We'll do episode two, zero click Boogaloo.
Jon Clark (44:13)
All right, let's...
Yes. All right, a couple really rapid fires. What is the most helpful word to you know when arguing in Italian?
Rand Fishkin (44:26)
Wow.
Basta!
Joe DeVita (44:27)
That was mine too.
Jon Clark (44:27)
Awesome.
Rand Fishkin (44:28)
Basta, I mean, means stop.
It's like, it's like stop right there. Stop you right there. Basta, basta.
Jon Clark (44:33)
Yeah, love that.
Rand Fishkin (44:34)
Or if you need to slow someone down, weirdly, you know how the way to make someone so upset in an English argument is to be like, whoa, whoa, don't get excited. Like calm down. That doesn't work in English, but weirdly in Italian Piano... piano, piano. Actually works nicely. It's sort of like take it soft and slow. I've noticed at least my wife's family, they can bring it down a volume notch, which is nice.
Jon Clark (44:43)
Yeah, yeah.
Calming word too by the way.
Rand Fishkin (44:58)
Yeah, piano, piano.
Jon Clark (45:00)
We filled out the agency survey. Any idea when those results are coming out?
Rand Fishkin (45:04)
Thank you. I think we're still collecting until the end of September. Probably they will launch the end of October is my guess after the analysis.
Jon Clark (45:11)
Okay, perfect. And then last one, five years from now, do think you'll be more known for SparkToro, SnackBar Studio, or something you haven't even started yet?
Rand Fishkin (45:20)
If I had to put money down, I would say it'll probably still be SparkToro, but that is, hope only because I think SnackBar Studio is going to be much more associated with Geraldine, and Nicolas and Francesco and Miriam, the other founders of SnackBar Studio who are actually making the game. I'm just the CEO. Like all I do is pay the bills and like recruit and interview and...
you know, sort of say, how's the schedule going, frenzy? I can't, I can't make video games, but they can. AlertMouse is interesting too.
Jon Clark (45:50)
Right, yeah.
Rand Fishkin (45:57)
Like AlertMouse has gotten a shocking amount of traction for a private beta product. Like you can't even use it, right? It doesn't exist yet, or it doesn't exist publicly, but like we have been shocked at how many people have somehow heard about it here or there. And then like email and say: "Oh, I really want this!". I think it's just cause Google alerts sucks so much.
Jon Clark (46:09)
Yeah, it's awful.
Rand Fishkin (46:10)
Like it's so, it's so bad. And people kind of are like, I don't really want to pay a huge amount for like Talkwalker or mention. Surely there must be something that's like cheap and good to track my mentions on the web. Scott used it for Seattle Ultrasonics, it's been pretty damn great for that, like tracking the launch. Maybe, I don't know. Hard to say.
Jon Clark (46:27)
Yeah, awesome.
All right, this has been amazing. Before we let you go, let everyone you know where they can find you.
Rand Fishkin (46:34)
Sure. yeah, if all of this sort of marketing, zero click marketing, you know, data around AI tools and search and all that is of interest, I'm on the SparkToro blog regularly publishing that stuff, which is sparktoro.com/blog. I'm most active in social and like the videos that I do stats that I share on LinkedIn and on Bluesky, Rand Fishkin on both of those. And, if you would like to assume the role of a chef in a magical version of 1960s Italy, who fights arcane boars and then brings their pancetta and guanciale back to your restaurant and makes your spaghetti alla carbonara, then you should check out snackbarstudio.com. Actually, I have a little, look at that. There you go, there's little characters from it.
Jon Clark (47:14)
Cool. Yeah.
Rand Fishkin (47:17)
You guys, thank you so much for having me. This has been a delight. I look forward to our next episode and I wish you well in the crazy, crazy world of marketing and SEO.
Jon Clark (47:25)
I couldn't have wrapped it up better myself. If you enjoyed the show, please like, rate, and review. We'll see you next time. Thanks so much, Rand. Bye bye.
Rand Fishkin (47:32)
Take care.