219: Your Homepage Is Not Your Homepage Anymore with Rand Fishkin

Rand Fishkin, co-founder of SparkToro and one of the most influential voices in digital marketing history, shares research proving that AI brand recommendations are wildly inconsistent. You'd need to ask ChatGPT 1,500 times to get the same brand list in the same order twice. We discuss why AI ranking tools are selling metrics that don't exist, why Google is still 210x bigger than ChatGPT, and why brand building (not digital marketing tricks) is the real lever for AI visibility.

About the Guest

Rand Fishkin - Co-founder & CEO of SparkToro, co-founder of Moz and Alertmouse, author of "Lost and Founder"

Chapters

• 00:00 - Intro
• 01:04 - The best thing AI has brought online
• 02:51 - The AI inconsistency research: 1,500 prompts for the same list
• 06:17 - Why AI tracking tools give you a false sense of visibility
• 09:16 - Is AI search actually better than Google?
• 11:34 - Kung Pao Chicken vs Peanut Butter: prompt variability
• 16:26 - What AI tracking should actually measure
• 20:16 - Why the best brands weren't built on digital marketing
• 22:43 - AI as "Spicy Autocomplete": where the hype exceeds reality
• 25:17 - Why AI cannot be creative
• 28:58 - Google is still 200x bigger than ChatGPT
• 31:28 - Two ways to show up in AI: base models and RAG
• 34:34 - Brand mentions as the real lever for AI visibility
• 38:10 - Zero-click marketing and the death of traffic as a metric
• 41:47 - Why SEO careers are more important than ever
• 44:39 - What brands should actually do first in March 2026
• 48:10 - The one thing about AI that should be killed today
• 52:02 - Where to find Rand

Key Takeaways

1. AI rankings don't exist
2. Visibility percentage is the real metric
3. Brand mentions drive AI visibility 
4. The best brands weren't built on digital marketing
5. Your homepage is not your homepage anymore
6. SEO is more important than ever, but the metric changed

RESOURCES:

• Rand's AI Inconsistency Research: https://sparktoro.com/blog/new-resear...
• SparkToro (audience research): https://sparktoro.com
• Alertmouse (brand mention monitoring, free): https://alertmouse.com
• Connect with Rand on LinkedIn: https://linkedin.com/in/randfishkin

Episode URL: https://www.nohackspod.com/episode/219-your-homepage-is-not-your-homepage-anymore-with-rand-fishkin

No Hacks is a podcast about web performance, technical SEO, and the agentic web. Hosted by Slobodan "Sani" Manic.

Transcript

Rand Fishkin
===

[00:00:00] Rand Fishkin: Your homepage is not your homepage anymore. Your homepage is what Google says about you. It's what ChatGPT says about you. It's what Instagram says about you. It's what Reddit says about you. It's what YouTube says about you. 

[00:00:13] That was Rand Fishkin, co-founder of SparkToro co-founder of AlertMouse one of the sharpest voices in history of digital marketing. In this episode, we talked about his study published on the SparkToro website link in the description where he found that you need to ask an LLM to give you a list of brands 1500 times before you get the same list ordered the same way.

[00:00:37] That is wild.

[00:00:39] We also talked about why AI tracking tools are not exactly what they promise to be. Also why the best brands of the last 30 years were not built on digital marketing . You have to listen to the episode. This is no hacks. Let's get into it.

[00:00:54] 

[00:01:04] Sani: Rand, welcome to no Hacks. It's a pleasure to have you on.

[00:01:08] Rand Fishkin: Oh, I'm honored. Sani. Thank you for having me.

[00:01:10] Sani: There's a lot of bad stuff happening on the internet but let's talk about the good stuff. What is the best thing that AI has brought online?

[00:01:20] Rand Fishkin: Uh, I think for some people who understand it and use it well, it's enhanced productivity. I, I talked to a number of engineers in particular who say. A lot of their job has been made more efficient, easier, more stress free, more bug free. Um, and I, you know, Casey, my co-founder at Spark Toro was always stressed out about the number of features and requests that I was bringing to him.

[00:01:52] Rand Fishkin: You know, oh, I think our users would really like this and this and this. And only recently he said, gimme all your feature requests. I can build them now,

[00:02:00] Sani: it has changed in that sense for sure. Uh, I was a, I was a web developer, then I went into CRO and now that this is available, I want to do web development again. I want to build stuff because. You can go from idea to, I'm not fully polished products there. There's a lot of work, but you can move fast.

[00:02:18] Rand Fishkin: idea. Idea to prototype is much, much

[00:02:21] Sani: Incredible. Yes. It's the best time to to be the idea person and then realize that you don't have the audience and you don't have the delivery method. That's a

[00:02:29] Rand Fishkin: I mean, marketing, marketing has not been made easier by ai, if anything, the inverse because the, the bar is so much higher to get

[00:02:38] Sani: There's a reason vibe marketing is not a thing, and vibe coding is a thing and, and vibe Marketing will never be a thing unless you consider the cold email where you buy a list and then AI writes, no, it's bad. Just don't, don't, don't do

[00:02:51] Rand Fishkin: I don't.

[00:02:51] Sani: Don't do that. If you're listening to this podcast,

[00:02:54] Rand Fishkin: I really, I don't see, I don't see vibe marketing ever becoming a

[00:02:58] Sani: agreed. The research you did at Spark Toro. It, it's fascinating.

[00:03:03] Rand Fishkin: Oh, thank you.

[00:03:05] Sani: How consistent is AI when they're recommending products?

[00:03:10] Rand Fishkin: Um, AI is inconsistent across the board on everything because it's a probabilistic output model and, and so. The problem here, the fundamental problems, honey, that, um, I think listeners would appreciate and and need to understand is that if you ask chat, GPT or Perplexity or Claude or, or Google's AI overview or Google's AI mode, if you ask those systems and services a question, a prompt that would lead to a brand recommendation, a product recommendation, a service recommendation, or a list of those things. You are going, you are not getting an answer when you ask, you are getting one of thousands or potentially millions of, of answers, and every time you ask it's gonna be different. And every different person who asks is gonna get a different list, a different number of items in the list, a different order of list items, and a different set of product or brand or service recommendations.

[00:04:15] Rand Fishkin: And if you don't fundamentally understand that, here, let me, let me say it as clearly as the study made it, in order to get two lists of brands that are the same in an answer, on average, you would need to ask Claude or ChatGPT fifteen hundred times, 1,500 times before you get two answers with the same list of brands in the same order.

[00:04:42] Sani: It is not likely to happen, is what you're saying there. Um, why do people trust it blind? Why do people go to therapy? For therapy to ChatGPT. Why don't they ask again to see if, if it's going to stay the same? It,

[00:04:55] Rand Fishkin: Uh, you, you know what's crazy about this study? After I did it, I could no longer ask AI tools any question one time and believe the answer that I got back. Like, personally, personally, after doing this study, for example, I was, I was shopping for some, um, kind of getting into like the world of classic menswear and I'm like, oh, I need some shoes, some nice leather shoes, you know, that are going to Oxford's or derbies or something.

[00:05:21] Rand Fishkin: And so I, I asked, um, I think it was Google's AI mode, and I got a list of brands and then I was like, wait a minute, what am I doing? I need to ask again and again and again. And so I just asked 10, 10, 12 times. And then once I had the full assembled list, that's what I sent to myself to go check out each of those brands that got mentioned, which, which was, you know, it was like 40 ish brands that were mentioned ACO across 12 answers.

[00:05:47] Rand Fishkin: Sometimes as few as three were mentioned by Google. You know, this is one of those, like, if one of those brands was trying to understand their presence in the AI ecosystem, what, what are they, what are they gonna do, man?

[00:06:01] Sani: easy. Also, the number, different number of brands being mentioned in every response is very fascinating. Which, which? Doesn't, it makes sense when you know how they work, but as a user, as someone looking for information, it doesn't make sense at all. And, and it, it's just charitable.

[00:06:17] Rand Fishkin: of, it's kind of wild. I think even worse is many of the AI tools, you know, your profounds or whatever. The, they, they try and track your presence in these AI tools and give you your rankings and your presence numbers. And they intentionally ask the ais, give me five or give me 10 recommendations for this, which the AI will then do, it will then give you a consistent list, but nobody in the, in the data set.

[00:06:45] Rand Fishkin: When you, when you look at how people actually use AI prompts, nobody asks for a number like that.

[00:06:52] Sani: we'll

[00:06:52] Rand Fishkin: You are getting a really false sense from these tools of your brand presence in these places.

[00:06:59] Sani: queries. I think you called them that in, in one of the articles. Kung Powell versus peanut butter. We'll get to that. That, that, I love that, that, that, that's just amazing. Also, the fact that you used, uh, chef Knives and I have a collection of Japanese chef knives. I, I was hooked in from, from, from the beginning.

[00:07:15] Sani: I, I completely disagree with the top recommendation. It just pulled from Wirecutter, the Mac. It, it, it's not bad. Yeah. I

[00:07:25] Rand Fishkin: I really look if you care about chefs, knives, if you care about any product, any service. You need to understand how those products and services were chosen and why they're recommending those products in that order. What are the criteria that are being considered? And frankly, most of the editorial lists on the internet for product recommendations at, in your example of Cutter knives, it was really just, did the reviewer like it?

[00:07:55] Rand Fishkin: Did they like the way it felt in their hand? Which is different than your hand. Did they like the way it chopped vegetables, which they probably chopped vegetables different than you. What do they use A pinch grip or a handle grip? Are they a trained chef? Have they been to culinary school? You don't know.

[00:08:09] Rand Fishkin: You don't know any of these things. Uh, this is why the only thing that in, in my opinion is, um, reasonable in these worlds if you really care about getting the right product recognition, is it find the list with true criteria. For example, you know, Scott Hyman Dinger did the Quantified Knife Project where he has a robot arm.

[00:08:28] Rand Fishkin: Put, you know, piercing the skin of a tomato a thousand times with different knives, and then telling you which one pierces the skin with the least force. There you go. That's your, that's your, there's data and criteria behind a knife recommendation. Ais don't do it. And I wonder, Sani, I truly wonder if five years, six years, 10 years from now, people are gonna be like, ha, remember when we used random statistical lotteries of words to try and give us advice?

[00:08:53] Sani: Uh, uh, is it 10 years in the future or is it a year or six months from now? 'cause I, I think the reckoning will happen if it's 10 years from now. We are in trouble if, if people trust this for 10 years before we realize it, that this is not the way, how is this better than Google search? Is this better than Google search in your opinion?

[00:09:16] Rand Fishkin: I personally, I think it's worse, but. I will say so many people like it and have embraced it and feel, feel that it's useful. And you can see in the usage statistics that you know, approximately, um, 30 40% of Americans at least have adopted AI as a a primary, not the primary. Google is still, their primary Google's usage has not gone down even in the era of rising ai. Um. They've adopted it to the degree that I don't think you're going to easily pull it out of their work and, and personal ecosystems, but I hope, I hope that there are moments of reckoning that each person experiences individually or some big pop culture or media event or that kind of thing that causes people to reevaluate their relationship with AI and how much they should trust the answers.

[00:10:19] Sani: Do you think ads might make people trust AI less once if they become a thing, or is that just distraction?

[00:10:28] Rand Fishkin: I suspect, no, actually I, one of the odd things, I'm sure you've seen this too with, um, media that you produce, right? So like, it's, it's strange to say, but podcasts and YouTube channels that have sponsored advertising. That have sponsors are often perceived by listeners or viewers as being, uh, more trustworthy.

[00:10:53] Sani: that's a good point.

[00:10:55] Rand Fishkin: And I think that's because they, their past associations are, oh, big important things have ads in them, so this must be a big important thing. Things with no ads in them, it's free. Who's behind it? You know, what's going on there? It's really interesting, even though ads are fundamentally something we say we hate. They can cause us to trust a platform or a source of media

[00:11:20] Sani: That's a really good point. It is more legitimate if it has ads because TV has ads, radio has ads. YouTube has ads. I don't like

[00:11:29] Rand Fishkin: think we grew up with it. We grew up with it, so we're used to it. So now it's part of the pattern.

[00:11:34] Sani: a good point. Now, the kung POed chicken versus peanut butter. This, this is a brilliant analogy. And, and yes, I agree with that. So this is about prompt variability and most of the tools that I've seen that, that do visibility, tracking air, basically tracking, there's some synthetic queries.

[00:11:51] Sani: Synthetic prompts, and they keep asking and asking and like, like you said, give me five brands. Gimme 10 brands, gimme, gimme this. That's not how people talk. That's not how people ask their AI anything. How much of a problem is that?

[00:12:06] Rand Fishkin: Okay. Full honesty here. When I went into this experiment, what I thought was that it was complete. BS total baloney, right? Just that the whole AI tracking industry was a scam. That's really what I believed, and this experiment was designed to sort of prove that hypothesis. One of the core pieces of that hypothesis is that synthetic queries, which is essentially you ask an ai, Hey, a bunch of users are interested in men's leather brogues, men's leather, Oxfords. What are the brands that you would recommend and like, oh, give me the list of prompts that people would be likely to ask an ai that's a synthetic prompt. You've asked an ai what humans would tell the AI in their prompt. good at that? Do they get close to the same things that real human beings do?

[00:13:00] Rand Fishkin: Do they have enough training data to do that? The answer is weird. Because when we asked, we asked, um, a few hundred people to craft a, a b, two B prompt, and a B2C prompt. Like, Hey, you're looking for a, you know, a service brand to help your friend with a, um, like a marketing brand, advertising campaign for a, a local coffee shop.

[00:13:23] Rand Fishkin: And we asked them to come up with one for, Hey, you're buying headphones for a family member for the holidays. And so we asked them to make these two prompts, no two prompts were even close to the same, right? These 150 people who, who wrote up their prompts and submitted them to us. And the lexical similarity, which is, which is something you know, you can, this is great, a great use of ai.

[00:13:46] Rand Fishkin: You can ask an AI tool to do a lexical comparison in Excel, right? And they're really good at that. They can do that kind of math for you on the fly. The lexical comparison was so low. I wanna say the number was like 0.07 or something. Um, one would be they're saying exactly the same thing. Thi this number, uh, point 0.07 was something like the similarity between kung pow chicken and peanut butter.

[00:14:10] Rand Fishkin: Extremely low lexical similarity, maybe contains a few of the same core ingredients like peanuts. But other than that, fundamentally different. So the answer to Part A is no AI's bad at telling you what prompts people actually use. The weird part, Sani is the brands that are produced from the human created prompts and the AI created prompts show a lot of similarity.

[00:14:44] Rand Fishkin: It's as though, essentially, under the hood, the AI understands the intent of the headphone. Requests, right? The headphone for a family member request, no matter how you phrase it, even though the phrasing was totally different and people were asking about like iOS compatibility or they were asking about noise canceling headphones or good headphones for travel or whatever.

[00:15:05] Rand Fishkin: Even though those were all in there, the brands that were produced looked very similar to what, you know, the 10 queries that were run a hundred times each done by the synthetic AI that. Those two matched, which suggests, which suggests that AI tracking is not complete BS

[00:15:25] Sani: but as a share of model and not rankings, and of course.

[00:15:30] Rand Fishkin: rankings is right out rankings is fully out. Don't trust anything. With rankings you can roughly trust if the systems are designed right and you should make sure if you're using AI tracking, you should make sure that the systems are designed right. You should look at that research and then ask your AI tracking tool the questions that I put at the bottom there.

[00:15:48] Rand Fishkin: Um. If it's done right, you can measure things like how often is Crockett and Jones or Alan Edmonds or churches in the Men's Oxfords recommendation and how, how often are they missing

[00:16:02] Sani: And that's going to be fairly consistent. Not, not a hundred percent consistent, but you, you can get a good idea. That's brand visibility.

[00:16:11] Rand Fishkin: ask? Yeah, if you sort of ask the right number of prompts, the right number of times. With some variability. You, um, you can get a statistical number that's basically plus or minus 5% or plus or minus 1% if you go really hard so you can get there.

[00:16:26] Sani: With that, that, that's real, even though it's ai, but that entire industry, how much of it is snake oil based on the tools that are available today? In

[00:16:37] Rand Fishkin: from what I see that,

[00:16:38] Sani: in practice.

[00:16:40] Rand Fishkin: from what I see there are. few providers who seem to be paying attention to doing it right and making some attempts. I, I talked to some folks, um, this past week, they were basically showing me their system introduction from a friend and, and, um, I was like, I'm not in this industry.

[00:16:59] Rand Fishkin: I have no horse in this race. I just don't wanna see marketers get scammed. And they were like, well, here's how we do things. And I was like, oh. Well, yeah, that's exactly how our research shows. And they were like, yes. When we saw your research, we were very excited because we want our customers to be asking the right questions.

[00:17:13] Rand Fishkin: Some of the big names though, um, you know, I think many people have heard of a company called Profound, which is, which is backed by Andreessen Horowitz, which whatever. Yeah. I don't, I don't know how you feel about Epstein connected people, but you know, like whatever. And then like the.

[00:17:31] Sani: feel about them.

[00:17:33] Rand Fishkin: Yeah, I don't,

[00:17:34] Sani: try not to think about them, but,

[00:17:36] Rand Fishkin: right.

[00:17:36] Rand Fishkin: Like I don't, you know, who knows? Maybe the guys behind Profound are like awesome and great and they just happen to get backing from a bad VC or whatever. But I, I mean, I think that, um, the, uh, they, they seem, their systems seem to be doing some things that I think even proponents of them are like, oh, I wish they did it differently.

[00:17:59] Sani: That is interesting. So what are the questions that a business should be asking from this tool before they commit and before they, they spend money on it?

[00:18:07] Rand Fishkin: Yeah. Um, I think this is, this is the frustrating part. It is pretty, it's a little complex and mathy, I would, I would tend to tell folks. You should go read the research. You don't have to read the whole thing, but at the, at the bottom there's like these, these five or six questions that you should be asking your AI tool provider and, and your AI tracking tool provider.

[00:18:28] Rand Fishkin: And those are essentially things like, uh, where do you get your prompts and like, how do you. Determine that those are similar to real human prompts. Where do you get your real human prompts to compare against what's, what's the source? You know, are you getting clickstream data? Who are you buying it from?

[00:18:43] Rand Fishkin: Whatever. And then, um, how many times are you running each prompt to sort of, you know, get to sta statistical significance? Have you published some research showing the statistical significance of your prompts and like, you know, with a strong stats approach and all that kind of stuff, how does that compare to, I link to, um. Neil, who's the, the PhD stats candidate who works with gum shoe. He, he published a great post showing essentially like, here's how many times you need to run a prompt to get to, you know, whatever, um, confidence intervals, uh, that are, that are high enough. Uh, so you should, you should be asking those kinds of questions.

[00:19:22] Rand Fishkin: And then I, I would, I would also be asking what are, what's the metrics that you report back? Are you, are you pretending that rank. Is a thing and it's real. Are you, are you using a number in your prompts? Are you saying Gimme five, gimme 10, gimme 15? 'cause nobody does that. So are, you know, measuring that variability is really important.

[00:19:44] Rand Fishkin: And then presence, you know, essentially, um, percent of visibility is the number that that's real, right? That an AI tracking tool should be giving you. So it's not like Google rank tracking. It's more like, um, it's more like when brands in the 20th century used to survey consumers and they would say like, Hey, have you heard of Nike shoes?

[00:20:08] Rand Fishkin: Have you heard of Adidas shoes? Have you heard of Reebok shoes? Right? And consumers would be like, oh, I've heard, you know, 80% of consumers have heard of Reebok, and 90% have heard of Nike. You know, that kind of

[00:20:16] Sani: That's a full circle moment just with ai how come Mad Men did branding better than we did in the 21st century with all the tools we have available? Like the, it is just mind boggling that it took generate not ai, LLMs and generative ai.

[00:20:32] Sani: Let's not call that AI because it, it's a subsets not the same thing. It's not a full match. Why did it take that to go back to doing things properly?

[00:20:42] Rand Fishkin: I mean, I. First off, I would argue we are not doing things properly, properly yet. Right. But I'd also say, I don't think, yeah, I don't think, I don't think that world, that world of sort of branding, brand advertising, positioning, I don't think it ever went away. I think those folks, right? The, the ogilvy's and the, um, you know, the brand builders of this world.

[00:21:11] Rand Fishkin: I think they had a shockingly successful run over the last 30, 40 years. It just, it's just been disparate from the digital marketing world. Like digital marketing had so much growth, basically growing from zero to. Hundreds of billions of dollars. What? You know, most of that is advertising with Meta and Google.

[00:21:33] Rand Fishkin: Granted. But that world grew so much, so fast, brought so many people like you and I, you know, digital marketers into the fray that the brand world, which was still building great brands and doing great brand marketing, just is not, it's not in our awareness

[00:21:52] Sani: that's an excellent response. There's so much, let's call it noise. Over the signal that you can't even see the, the real branding, like the things you mentioned. So not the way I looked at it before, but I, I think I agree with that. Take a hundred percent. Let's talk.

[00:22:06] Rand Fishkin: I am worried Sani, like I'd be worried that if you were to analyze the companies and brands that had the most success over the last 30 years, I'd be worried that very few of them would be built on digital marketing. A lot of them would be built on classic brand marketing, brand building. Brand advertising.

[00:22:27] Sani: If you exclude tech companies, I, I think it's a hundred percent almost. Yeah. I mean,

[00:22:33] Rand Fishkin: Yep.

[00:22:34] Sani: interesting, interesting, interesting

[00:22:36] Rand Fishkin: Yeah. Show me the consumer brand that won their market by ranking number one in Google. I don't know that they exist.

[00:22:43] Sani: Ah, that's a good point. That's a, that's such a good point. But let's talk about AI as a. Hype machine. How does the hype exceed reality? Is, is any of this real, you call it? I'm quoting you. I like this. That's why I'm quoting you Spicy Auto complete. My, my phrase is Grammarly on steroids since it came out and, and kind of what it is.

[00:23:07] Rand Fishkin: Yep.

[00:23:07] Sani: for some tasks. Absolutely, and I think it's, it's very useful for a lot of things that people don't really care about and not as useful and good at things that people are pushed into caring about. And we have a problem there. What's your take? Is it, it's not bad. It's not all bad. You can do stuff with ai.

[00:23:25] Rand Fishkin: No, no.

[00:23:26] Sani: that.

[00:23:27] Rand Fishkin: Yeah. I mean, look it, um, I have had hundreds of thousands of rows of data inside Excel that I needed categorized and classified in linguistic ways that before an AI plugin would've been. I mean, I basically would've paid someone thousands of dollars, tens of thousands of dollars, right? To, to like do it manually or build a system to do it. And now it's like 10 cents. You know, like I, it's, that's really nice. Like, that really does save me time. It allows me to produce research that previously would've been inaccessible to someone like me. Wonderful. Um. I think it can be great for surfacing insights. Like, Hey, here's this analytics data, here's this traffic data.

[00:24:10] Rand Fishkin: Are there any interesting trends or correlations that you found that look problematic or promising? And it can surface things, right? Th things for you to investigate. It's great for writing code. Like it re it writes code as well. As a, a good engineer would, um, if many people are saying it's getting toward what a great engineer could write.

[00:24:32] Rand Fishkin: Granted, you still have to connect all these systems. You, you know, you can't be like, Hey, just can you just connect stripe to my bank account and then dah, dah, dah. And you're like, whoa, whoa, whoa. And you're not gonna confirm that. And you don't care if the money goes to the wrong place. Like you gotta confirm everything, right?

[00:24:47] Rand Fishkin: So it, it's become like a little worker. Be, but, but I think the unfortunate part, and the part that is to my mind fundamentally wrong is when human beings think that. AI has taste and a point of view and, and real world perception or experience. It has none of those things, and therefore it cannot create anything creative like by its nature.

[00:25:17] Rand Fishkin: It can't be creative. It can only produce patterns that it's seen before. That's fundamentally how it works. There's nothing new and creative that it can make. It can put words into a potentially new order. And then, you know, I, I don't know. Some people can be fooled into thinking that's new, but it's not new.

[00:25:38] Rand Fishkin: And the other thing it cannot do is it cannot do creative endeavors in ways that are acceptable, acceptable for publication. Can it give you ideas? Absolutely. Can I give you something to review that you then polish and sometimes you're surprised 'cause it's 90% good and sometimes you're less surprised 'cause it's 10% good and you gotta start all over.

[00:26:00] Rand Fishkin: Sure. But um, I think the other issue, I'm sure you feel the same way, Sani, is everything creative that it does or everything creative that it, it can do or is asked to do? Um, it's built on something that built on technology that is fundamentally unethical and should have been illegal.

[00:26:24] Sani: I, I could not agree more. Uh, I, I, I have my takes on, on big tech that I've shared a lot of times on this podcast that they're, none of them are good. Uh, positive. Maybe they're good. Maybe I believe they're good, but they're not. I think the world is better without big tech. That, that, that, that's, that's my take on big tech.

[00:26:42] Sani: Without tech billionaires, without all the damage they're doing, I think we would be in a better place. Uh, I mean, look at the, the, is it deposition trial for Zuckerberg? That's happening now about social media addiction and I was talking to a friend, uh, over the weekend. My, my daughter's 11. His daughter is 11.

[00:27:00] Sani: They, they, they go to school together. How lucky we are that we, it's going to get banned just before they jump on the train because Europe is starting to ban social media for under 16, and I'm so glad that she wasn't 11 five years ago. 'cause life would be significantly. And those are the formative years.

[00:27:20] Sani: Did they, did they ruin life? Anyway, big tech, we can talk about big tech and, and the damage and hopefully humanity will for a long time so it doesn't happen again.

[00:27:29] Rand Fishkin: I know there's people who sort of believe hopefully that AI slop will make social networks and social media so unusable that no one wants to go to them. That essentially like they're filled, it's filled with so much junk, you can no longer believe anything on there. You know, there's whatever, here's an engaging video.

[00:27:52] Rand Fishkin: Oh, was that, is that real? Did that happen? No, it didn't. I can't even tell. And so people turn off, uh, social media in a way, in the same way that they sort of lost trust in other institutions in the, you know, last 40, 50 years. I don't know. That seems like an optimistic case to me,

[00:28:10] Sani: that's not the way to do it. If you're solving a problem, you don't set your house on fire because the problem is inside the house. Like it's just a, if you're hoping AI slop

[00:28:20] Rand Fishkin: Oh, no, no, I, this is the American way.

[00:28:22] Sani: Okay. I will not argue with that. No, but it's wrong. It's just slop will save us. Is is the weirdest thing anyone can think, and I hope no one seriously thinks that, but even without, with all the negative things that a lot of people do know.

[00:28:39] Sani: If you, if you read Karen, how's book, if you follow people like Gary Marcus and, and all of the, it's add tron on Better Offline Podcast, all those people, you know that AI is not all flowers. It's not all roses. There's a lot of negative one. A lot of people still believe it's the best thing ever, and it certainly gets a lot of hype around it.

[00:28:58] Rand Fishkin: Yeah.

[00:28:58] Sani: And yet Google is still 200 times bigger than chat GPT and, and not going away and chat. GPT, uh, is declining. Uh, this is a like recent metric that the number of users, also the number of queries in RTPT, they're not all searched. This is not killing Google Band. This is just a a, a chat buddy, like a virtual friend for a lot of people.

[00:29:21] Sani: Do you think we'll get the narrative back? The, the narrative will kind of write itself and people will realize that this is not killing Google. This is not replacing anything. This is maybe an addition to the field.

[00:29:34] Rand Fishkin: I think so. I, I mean, I recall, um, several cycles like this before right there. O one was mobile is gonna kill desktop. Right? And what happened was mobile did become almost two thirds of internet use. Yet desktop activity basically stayed the same for the last 15 years. I mean, it's wild to remember, but in 2011, I just looked up the statistic.

[00:30:00] Rand Fishkin: In 2011, only 7% of web traffic was mobile devices. That's not that long ago. Like that's crazy. 93% desktop. Now we're in a, you know, now we're in a 66% mobile world. Um, and the the next one was social media. Facebook and Instagram and Twitter. They were, there are so many articles. You can go back to 20 12, 13, 14, 15, so many articles written about how they were gonna kill search.

[00:30:28] Rand Fishkin: They're replacing Google TikTok. For a while when TikTok was rising, people were like, oh, they're gonna kill Google. Nothing ever kills Google. Like it never, ever happens. They, it stays big. This is the same thing, and I believe in a few years, we'll look back. People, everyone will be like, oh yeah, I never believed that.

[00:30:45] Rand Fishkin: Even though they're the one who wrote the article, you know?

[00:30:47] Sani: and let's not ignore Metaverse in because that was supposed to replace everything. The $70 billion went somewhere that was supposed to replace life and Google and every with no lag and, and. Such a weird, such a,

[00:31:02] Rand Fishkin: Insane. So insane. I mean, to me it was a lot like the crypto world or the world of blockchain. Um, NFTs, right? NFTs were gonna be. This massive asset class and oh my God, what a,

[00:31:17] Sani: So if, if this LLM visibility is not about rankings, what is it? How do you optimize your presence with AI as a part of the

[00:31:28] Rand Fishkin: Yeah. Well, so, um, a lot of folks smarter than I am on this stuff have pointed out. Um, and I'm sure you probably have some of you on your show, right? That there's essentially. Two modes of appearing, two ways that you can appear inside AI responses. One is through the base models themselves. So this is essentially, your brand appears next to the words from the prompt and from the query fan out frequently.

[00:31:55] Rand Fishkin: You know, if, if you wanna be, um, ranking highly for, you know, uh, uh, leather Oxford Brogues men's shoes. Make sure that whenever people talk about those topics on the web, your brand is mentioned on that page. Hopefully close to those other words. Um, and in context, you know, if the New York Times is doing a new piece on what are the best leather shoes, you know, Oxford's for men, your brand better be on that list, right?

[00:32:24] Rand Fishkin: If, um, a bunch of bloggers in the shoe world are talking about it, if people on the style forum, you know, website are talking about it, if there's subreddits in men's fashion that are talking about. You need to be in those conversations. If you're not mentioned in those places and your competitors are, they will show up when the model produces results.

[00:32:42] Rand Fishkin: The second one, the second way to show up, is through what's called retrieval, augmented Generation Rag, which essentially is just, it's the AI doing web searches the same way you and I would search Google, and in fact, you know, chat, GPT has probably produced them. Ton more Google searches because we know that they just, you know, they send a lot of their queries right over to Google.

[00:33:06] Rand Fishkin: So when you wanna show up in those results, the answer is rank highly in Google, right? Definitely the first page, hopefully the top few results. Um, and often you'll be pulled from if you're not, if it doesn't have to be your website, weirdly enough. As an example, um, you know, if if there's a, a popular post on men's fashion in the subreddit and that ranks highly, and your brand is mentioned in those results, good news for you, right?

[00:33:38] Rand Fishkin: Like that is gonna be pulled in into the search result and then shown in Google's AI mode or Chachi PTs answer, or Claudes or whatever. So there's the, these two ways. The, the challenge here is. It's always changing Sani, like it is, it is changing so, so fast. It's like the early days of Google where they're, you know, they're trying to put the hammer on various spam techniques.

[00:34:02] Rand Fishkin: Like, remember for the last year until probably like two or three months ago, you could just make a bunch of shitty blog articles, you know, with, with that were like top men's shoes and then put your brand at the top of 'em. And you would just show up at the top of chat GBT. And that's working less now.

[00:34:21] Rand Fishkin: And I'm sure in six months it won't work at all, but you know.

[00:34:23] Sani: Early days of Google is really the best possible comparison, which is why lately I've tried to get as many of the SEO OGs. 'cause they have figured this out once. No, I'm not joking at all.

[00:34:33] Rand Fishkin: Yeah. Yeah.

[00:34:34] Sani: Figure this out once how machines interpret things, and maybe we go to the people who already did it once and, and, and try to find the answers. So another thing you wrote about is that now what you are online and what your brand is, is basically what AI decides.

[00:34:50] Sani: It's

[00:34:53] Rand Fishkin: Absolutely.

[00:34:54] Sani: Simple brand mentions. Is there a direct correlation between simple brand mentions and AI visibility?

[00:35:01] Rand Fishkin: Yeah. Yeah, absolutely. Um, the. The more complex answer is, it depends on where the mention is, uh, who the mention is from, and whether the mention is relevant. And it also depends on when the AI model was created and how frequently they're updated and, um, the curation process for those things. So it, it's a complex answer, but absolutely.

[00:35:29] Rand Fishkin: I, I think what, you know, what's kind of nice about this world? What's a little bit nice about AI visibility versus SEO visibility in the early days, SEO visibility, you know, especially up until like 20 12, 13, it was so much about these technical things that if you didn't, if you weren't closely connected to the SEO field, you would never even think about like,

[00:35:54] Sani: I

[00:35:55] Rand Fishkin: you know, the biggest example is why do I care who links to me and who doesn't?

[00:35:59] Rand Fishkin: If my brand, if I'm producing a good product and people are talking about my brand on the web and they're saying good things, and I get good reviews, and whenever the press publishes articles, they say good things and you know, editorial co coverage of us is good. People are making videos about us. Why do I need to care whether that mention contains a link or not? And you're right. And then the people who said, oh, I got a link. Yeah, but the link doesn't have the anchor text for the keywords of your, what the hell are you talking about? Like, that's such crap. That is such a, a, a stupid ranking signal. And yet it, it really influenced how Google showed things Nowadays with ai, yes, there's some gamification and there's some like shady things you can do, but you know what?

[00:36:44] Rand Fishkin: If you ask 10 people, Hey, what are some, you know, fancy men's shoe brands that you've heard of that make Oxfords and Bros and derbies people would name like, I don't know, five or six brands. And so will ai, they will also name those brands because what tends to get reflected in AI is what people talk about and, and what gets sort of promoted and what's in the, the mainstream conversations.

[00:37:12] Rand Fishkin: That's kind of nice.

[00:37:13] Sani: that's great.

[00:37:14] Rand Fishkin: That's nice because the same way you build your brand offline is the way you now build your brand online.

[00:37:21] Sani: So there's something good about it. There's something positive

[00:37:23] Rand Fishkin: There's something. Yeah.

[00:37:24] Sani: Also, the, the name

[00:37:24] Rand Fishkin: It's better than last time.

[00:37:27] Sani: the name of the podcast is No Hacks, and I started five years ago. This week was 50th anniversary because all those things you said, you mentioned they're hacks that worked until they didn't, and. We are now moving into the no hacks era.

[00:37:41] Sani: Sure, there's always going to be something, but it gets exposed much faster and it backfires in a more spectacular way. Like all those Reddit bombers that wanted to get into chat GPT by bumping Reddit, where are you now? It, and they knew it, wouldn't it? It wasn't the right thing to do. It's just today this seems to work, even though it's stupid.

[00:38:05] Sani: Let's do it as much as we can. I like all of that. Let's,

[00:38:09] Rand Fishkin: Yeah.

[00:38:10] Sani: one thing that you've talked about, let's talk about the framework for this new world and how brands can exist and attempt to thrive in this because nobody knows what fully works. One thing you've been talking, one thing you've been talking about for a long time is the, the zero click and, and you've been, you've been on that before.

[00:38:28] Sani: It kind of officially be before the crocodile mouth in Google Search console before all that. Is this the next step of that zero click and traffic as a energy metric.

[00:38:39] Rand Fishkin: Yeah, ab absolutely. I think, you know, Google took a lot of pride in, was that 2024? 2023 saying like, Hey, despite these, you know, AI overviews and the new answers that we're putting in, and all, all these instant answers, we are sending more traffic to the open web than we ever have before. And last year they shut up about that because, because right, that crocodile graph, impressions go up, clicks go down, and, and Google is absolutely, um, answering more and more searches without sending a click out.

[00:39:15] Rand Fishkin: And many people, you know, I mentioned 30 to 40% of Americans are turning to AI tools and, and Google's AI mode rather than a list of potential search results to click on. so what, what that means is when people search for your category, your topic, the thing that you do, or you specifically, their first impression of your brand is not coming from your website.

[00:39:43] Rand Fishkin: Your homepage is not your homepage anymore. Your homepage is what Google says about you. It's what ChatGPT says about you. It's what Instagram says about you. It's what Reddit says about you. It's what YouTube says about you. It's a different

[00:39:59] Sani: is more honest. And another former guest in this podcast, Jonah Alderson, who you probably know as well, the technical SEO guy, he said, this is, this is basically upstream engineering. If you were optimizing the website, time to go upstream and and

[00:40:13] Rand Fishkin: Yep, yep, yep.

[00:40:14] Sani: if you're just wanting to, yeah.

[00:40:17] Rand Fishkin: yeah, if you think back to the 20th century, no, no. You know. No shoe brand thought that their store was the first and only experience, right? It was, Hey, people have this brand perception and that comes from television and radio and word of mouth. And it comes from, uh, their, their driving experiences and highway billboards and their perception when they go to the mall and see the front of the shop. That that is also now true on the internet. Right. It's no longer I search for the shoe brand. I go click on their website or I search for, you know, whatever top Oxfords, and then your, your site is ranking number one. That looks suspicious, like, now that looks shady. I wanted to see a list of shoe brands.

[00:41:02] Rand Fishkin: Why is your one website ranking number one? That's sketchy, man. Like you, you must have, you know, fooled Google.

[00:41:11] Sani: We are not calling SEO as a profession. Sketchy. Let's be clear about that. The entire industry, it's not. Absolutely not, but something's

[00:41:19] Rand Fishkin: It's, it's so much less sketchy than it was like when you and I, when you and I got into it, it was so sketchy. Now it's a very normal

[00:41:27] Sani: stuffing used to work, content cloaking used to work. All those stupid things. Used to absolutely work until they stopped working. So with SEO, with everything changing, what, what career options do SEO people have? Like how do they, how can they continue doing the work that they've been doing for 20, 30 years?

[00:41:47] Rand Fishkin: Yeah, I mean, I think their work is more important than it's ever been. More important than it's ever been because remember that alligator chart that we're talking about, right? Where essentially. What, what, what we're saying is impressions are going up, meaning more people are seeing your brand, your website, you know your stuff in their results, but clicks are declining simultaneously, and so this means that the only people who can influence the perception that your brand receives across the web.

[00:42:21] Rand Fishkin: Are those who can influence what's in those results. Influencing the results themselves is now more important than it ever was. It's just that the way you measure it is no longer did we get a click to our website. That's the part that's changed. And so, you know, for, for anybody who's an SEO, I think, I think you have fantastic job security.

[00:42:42] Rand Fishkin: So long as you can explain that concept to your boss, your team, your client, right at your agency. But you're gonna have to change how you measure success, and you're gonna have to change your goal. Your goal is now get the brand more sales, not not traffic sales, right? Like your job is help that brand improve their online perception.

[00:43:06] Rand Fishkin: Get those impressions, get that reach, influence people's opinions, and then, and then make the

[00:43:12] Sani: Uh, so does that mean that SEO and CRO cannot exist without one another in this new world? Are they two separate industries or is, is this, if we're talking about sales and onsite conversion optimization, that becomes a part of. This entire discovery process.

[00:43:31] Rand Fishkin: Yeah, that's quite interesting. You know, I haven't thought about a ton about this, but I would, I would guess that a lot of the work that CRO used to do. That happened in the discovery and consideration phase, which was very website centric, has now gone upstream and it's happening in places that you don't own and control.

[00:43:54] Rand Fishkin: And so essentially the CROs job has changed as much, if not more than the SEOs where their job is now. How do we change perception and visibility and experience of people who are. Really experiencing all of their discovery and all of their investigation and consideration on Instagram, on Reddit, on YouTube, in ChatGPT prompt answers in Google AI mode.

[00:44:24] Sani: It, it's, it's more complicated for sure. If I'm a brand and I want to work on my AI visibility, not rankings, share a model, what's the first thing I need to do? What's the, the thing for me to do in March, 2026? Okay.

[00:44:39] Rand Fishkin: Um, figure out if your audience actually uses AI tools to get brand recommend, right? Like I. I am seriously worried that the hype train, the AI hype train means that, know, every brand thinks they have to go do AI stuff. And I, I was talking to, um, the marketing team who runs a, a resort, it's like a wellness resort in Arizona yesterday, and one of the conversations was, Hey, what percent of your audience really uses AI rather than.

[00:45:15] Rand Fishkin: You know, a Google Maps hotel search or, um, asking their friends or going on Instagram or going on TikTok or asking on Reddit, like, you know, our chubby travel or big travel or whatever, it's probably not that big, so let's just make sure, I'm not saying don't invest in it, but maybe don't invest 50% of your

[00:45:38] Sani: Oh, absolutely.

[00:45:39] Rand Fishkin: Right. That's. That'd be really crazy to go through. So the first thing, first thing to do is make sure that your audience is actually using these AI tools to do the things that, that you expect. Um, the, the second thing that I would say after that, once you confirm that and you're sort of like, okay, this is how much budget and time and effort we should put towards this.

[00:46:01] Rand Fishkin: The second thing is figuring out how you need to position yourself to win in those spaces. Right. We, hopefully you already know that your ideal customer and your ideal customer perception match a set of words and descriptions that, that when, when those click people buy from you, like when you say, Hey, our wellness resort is the absolute perfect place for mindful women travelers between 35 and 55.

[00:46:36] Rand Fishkin: Great. Fantastic. Amazing. Knowing that you know how you need to describe yourselves and the words and phrases you need to be around and you know which Forbes travel list to be on and which one to ignore and you know, which, which PR outlets to have your, your PR person pitch and which ones to ignore. And you know, what, when to participate in a Reddit conversation and when to ignore it and which YouTube videos and which influencers to bring to your resort.

[00:47:01] Rand Fishkin: Like all, you know, all those things, but you can waste a lot of time and energy if you don't have that dial in.

[00:47:06] Sani: But dialing that in is a lot of work. And Mo, most brands, no, I'm not. I'm not saying don't do it. I'm

[00:47:12] Rand Fishkin: this is good. This is good for marketing employment. I'm telling you,

[00:47:15] Sani: is, but the reason most brands never did that is because paying for ads or just tricking Google to rank me, number one, is a shortcut. And it's easier to do and it's faster, easier to justify spending money on.

[00:47:29] Sani: I'm, again, I'm not saying any of this is good, but that there is a reason the entire digital, most of digital marketing industry is built on these principles. Let's, let's get the shortcut. Let's get something that works today because my boss. Expect something on Monday,

[00:47:45] Rand Fishkin: Right. Yeah.

[00:47:47] Sani: I guess we are, we're kind of fixing all that.

[00:47:49] Sani: This is a great setup

[00:47:50] Rand Fishkin: I mean,

[00:47:51] Sani: for the final question.

[00:47:52] Rand Fishkin: think this is Sani. I, I think this is on marketers to convince their bosses, right? Like, it's your job to convince your boss of how things work, and if you cannot. Please by, please go find a boss who will believe in the way marketing actually works.

[00:48:10] Sani: Great advice. Final question. We started with the best thing about AI on the internet being part of it. What is the, the, the one thing that you, if you could kill it today and remove it today, you just snap your fingers into it?

[00:48:24] Rand Fishkin: Um,

[00:48:26] Sani: We are not talking about brands. I'm not, I'm not asking

[00:48:28] Rand Fishkin: Yeah. I think,

[00:48:29] Sani: that.

[00:48:30] Rand Fishkin: I think for me it's. It's probably visual content. I think, um, I think one of two things should be true. Either AI should entirely stop creating visual content and, you know, policymakers should outlaw it should essentially say, Hey, this is, that is not a legal way to create stuff. If, if we find stuff that's made this way and it is published publicly, uh, the person responsible, you know, can be held.

[00:49:00] Rand Fishkin: Whatever, accountable for some misdemeanor or crime or that kind of thing. Um, and the only alternative that I can see is that every time, uh, an AI generates that image, whatever artists works, we're used to create, you know, to train that data, get a paycheck,

[00:49:23] Sani: Uh,

[00:49:23] Rand Fishkin: like, oh, you wanna, you wanna make. A video that looks like, you know, Disney's the Incredibles.

[00:49:29] Rand Fishkin: Well, guess what? Disney gets a paycheck. You know, like Disney takes a cut of that. And actually Disney is a good example because they might be the ones in a lawsuit to be able to force AI to actually pay them because they're,

[00:49:42] Sani: I was gonna say, I don't feel bad for Disney, but what they did to Giblin Studio with Sam, Sam Al's Twitter avatar is still, I think he hasn't changed it. It's still

[00:49:51] Rand Fishkin: I mean, it's just a giant, it's just a giant middle finger

[00:49:54] Sani: Yes. And, and

[00:49:56] Rand Fishkin: e everything, everything about copyright law and how artists should be treated and how creatives deserve money. And he, he's basically saying, because I am rich and powerful and a billionaire and whatever, like an Epstein style character. You know, like I, I'm not saying, I'm not talking about like the, the,

[00:50:19] Sani: Of course. Absolutely

[00:50:19] Rand Fishkin: like the sexual assault stuff, but the he is, he is in that world and that kind of character and those, those gestures, right?

[00:50:30] Rand Fishkin: Making his avatar that is essentially a public constant f you to anyone who believes that artists should be treated fairly, should be paid for their work, that they own their work that, um. Creativity is belongs to a human.

[00:50:47] Sani: yes. And that is a, first of

[00:50:49] Rand Fishkin: that, that's the thing I hate

[00:50:50] Sani: first of all, you, you just killed mornings. For a lot of seniors on Facebook, they will not be able to scroll and see all those visuals that were AI generated. They will be sad now. Sorry. All the cat videos, all, all the cats playing pianos and all that stuff, you just killed all

[00:51:04] Rand Fishkin: I mean, I guess it's, it's better than them seeing Trump is Jesus memes. So.

[00:51:10] Sani: Hockey player is the latest one that I've seen this morning and, and it's worth checking out. I'll just say that. Rand, thank you again. This has been an absolute pleasure 

[00:51:21] Rand Fishkin: pleasure for me as well. Uh, you're a wonderful interviewer and I really enjoyed our conversation. If you, if you make it to Seattle, Washington. Drop me

[00:51:28] Sani: When you get the sonics back, I will, I'll tell you that 'cause that was my team in the nineties. I love them. Sean kept Gary Peyton era of Sonics is just,

[00:51:39] Rand Fishkin: Benoit Benjamin. Oh my God.

[00:51:40] Rand Fishkin: Ray Allen. I mean, we had Ray Allen. Uh,

[00:51:43] Sani: it was. The team of the nineties as much as the bulls were, I think the sonics were that even more.

[00:51:48] Sani: I hope you get them back there. There's some rumblings that it might happen, so we'll

[00:51:52] Rand Fishkin: Yeah. Yeah. I think that all the sports leagues have been pointing at the sonics for, you know, three decades and saying like, if you don't fund a stadium voters, we'll take your team.

[00:52:02] Sani: go. You, you deserve to have them back. What is the best way for people to connect with you and follow you?

[00:52:08] Rand Fishkin: sure. Yeah. Well, so if you are, uh, interested in understanding. What, what your audience does on the internet, their behaviors and demographics, and whether they use AI tools, for example, uh, spark toro.com is a great place to go. If you think Google alerts sucks and you're wondering why it's even still a product and you wish that something would help you track your brand.

[00:52:28] Rand Fishkin: Alert mouse.com is free. Uh, passion project with a couple of co-founders and if you love. Uh, cooking video games where you forage and fight monsters and bring back your Ponta and make your spaghetti ra. Uh, you can check out snack bar studio.com where we are making our, uh, our first video game that launches next year.

[00:52:49] Rand Fishkin: I.

[00:52:50] Sani: That sounds absolutely amazing. I cannot wait for the game. Ryan, thank you so much for everyone listening. I will talk to you next week. 

[00:52:56] Oh

[00:53:24] no. Hatch,

[00:53:43] no hatch.

[00:53:48] Go back.