The State of GEO w/ Profound's Josh Blyskal

**Ross Hudgens** (0:11)
Josh, you are our AI search wizard.
You came on once, we talked about AEO, GEO. It went very well, so I figured, let's juice it up with the in-person version of this. So welcome back.

**Josh Blyskal** (0:26)
Thanks for having me, man. Absolutely. What a pleasure.

**Ross Hudgens** (0:28)
So you put out some recent research that I thought was a good place to start. There's been some studies flowing out the Internet of Web from various people. Seen it from Kevin Indig, who we were speaking glowingly of right before this, other people as well. How often should people be tracking these things for accuracy?
What's your take on that?

**Josh Blyskal** (0:48)
I'll put it simply and say that if you're trying to track a wide breadth of prompts, the relative value of repetition is much lower. If you're going wide, you don't need as much repetition. If you're trying to track a small number of prompts, repetition will help you. It really depends on the prompt portfolio. We did some research. We found that the difference between prompting once a day and 10x a day, there's only 0.25 percent difference in visibility. It's tiny. That's not an actionable gap.
We did a second test and found that the bigger change, that will actually change your visibility score, day-to-day or moment-to-moment is the prompt portfolio you're tracking. If I change my prompts, my visibility score is going to change way more if I expand the prompt set or decrease it.
Think about it as that. Intellectually, honestly, Profound can run repetition and we do for some customers, but there's tons of people out there that are paying for more repetition and more executions than they need if they have, say, 500, 1,000, 1,500 unique prompts that they're tracking. Because at the end of the day, pragmatically, marketers are trying to understand a wide subset of the market. You can do that in multiple ways. It's a methodology question, really.

**Ross Hudgens** (2:00)
Yeah, no, that's great. I assume you guys can charge on a model of the more you run it, the more expensive it is for people, right?

**Josh Blyskal** (2:08)
But we're trying to go to the highest signal with least number of executions so that you can do more with those executions that you do have. Say you can only run 100 executions a month, I'd much rather you get a wider sample of your market than having 10 little samples 10 times over.

**Ross Hudgens** (2:27)
Right.

**Josh Blyskal** (2:28)
It's important. People will do different things though. Everyone has a right to their own opinion.

**Ross Hudgens** (2:32)
Yeah, makes sense. I think of in classical SEO, there'd be people with maybe just have one head term that's just like the bulk of their world. Yeah. That would be for that person, it sounds like run that. I guess, what do you think the right frequency for that person would, let's call it classical SEO had one page with 90% of the search volume?

**Josh Blyskal** (2:56)
Wow.

**Ross Hudgens** (2:58)
What would your gut be on that?

**Josh Blyskal** (3:00)
My gut there is that you should run it between five to ten times if you're literally tracking. A day. Yeah, a day. If you're tracking literally one head term, one prompt, let's do it multiple times if it truly is like that for your business. But what I would say is if you're Nike and you're doing shoes, maybe there's actually many query fan outs or many prompts that you want to track. In that case, let's be intellectually honest about the different ways our audiences are actually searching.
If it's bigger than just what's the best shoe company, then you're probably actually looking at more prompts, even though you might think in traditional SEO.

**Ross Hudgens** (3:31)
Yeah.

**Josh Blyskal** (3:31)
Oh, that's fair. It's actually one head term. That world doesn't exist, Catch-22.
Exactly. That world doesn't exist. That might actually be 200 prompts, depending on how granular you want to look at it. You should go with the prompt volume first.
You see 90% of the search volume around this particular page, or you're seeing 90% of your referral traffic from ChatDBT on this one page. It's not enough anymore to say, oh, what's the keyword for that page, and then that's our AI search keyword. Not to be, this is not a plug, by the way, but with our prompt volume tool, we can see in the wider world, we have 1.5 billion prompt volumes, we have 1.5 billion conversations. We get the user's submission, but then we also get the model's response, and with the response, we see global citations. So we can see, if you put a URL into Profound, we could be like, your URL was cited, we can back into what prompts a URL is actually being cited for. So if you're like, hey, Josh, nike.com/shoesdashmen is our number one page by far, I could be like, listen, okay, there's actually like 20 main clusters that this is getting cited for, here are the topical clusters, here's the distribution. If you say, hey, it is actually adidas.com/shoesdashwomen that's cited, I could be like, oh, this is very different than anything I've ever seen. This is actually, there's only two or three prompts that are really uniquely showing up, two or three fan nodes. So depending on what that actually is, would dictate, I think, what I would recommend for you to track. But it's cool, but it does depend.