0:00 Okay, I've got Jeff Oxford who I did a 0:02 video with I think about a year ago on 0:04 how to rank products in Chat GPT. Super 0:08 insightful and that video actually did 0:10 really well. I think you actually got 0:11 clients from that video. And anyways, 0:14 Jeff has been like heads down learning 0:17 everything about you know AEO these days 0:20 and how to rank within AI search. And so 0:24 I asked Jeff to if you'd be willing to 0:26 hop on do a video and share you know, 0:29 what's working right now. I he's they 0:31 they have lots of clients at his agency. 0:34 He's got 0:35 is visibility labs, right? And then 180 0:38 marketing works with 0:40 lots of different businesses, a lot of 0:42 e-com businesses especially at the 180 0:44 marketing agency. But if there's one 0:47 person on planet Earth who's nerding out 0:50 deep in the weeds of AI 0:54 search, I know it's Jeff and Jeff sure 0:57 enough has an insane amount of 1:01 slide decks already on [laughter] 1:03 on this topic. So Jeff, thanks for 1:06 coming back on and I'm excited to hear 1:09 sort of your findings and what's working 1:11 with ranking companies right now on the 1:13 LLMs from what you guys are seeing. 1:16 Yeah, thanks Austin. Always great 1:18 chatting with you and sharing the latest 1:20 and greatest of what's going on with AI 1:22 search. It's it's definitely a bit of a 1:24 wild wild west right now. Lots of myths 1:27 and contradictory information in there. 1:28 So I just thought I'd present you guys 1:31 like here's where things are right now. 1:33 Here's what's working, here's what's not 1:35 working, here's where the trends are. So 1:37 I'm going to go ahead and share my 1:38 screen. Like Austin said, I do love my 1:41 uh 1:41 my presentations. 1:43 So let's go full screen and boom. We are 1:46 going to talk about how to win in AI 1:50 search. So few things that we'll be 1:52 going over. First I'm going to give a 1:53 quick primer on how these large language 1:55 models actually work. If you've heard 1:57 the term LLM, 1:58 it's basically Chat GPT, Perplexity, 2:01 Claude. These are large language models. 2:03 We're going to talk about what they are, 2:05 how they work. I want to talk a little 2:07 bit about is SEO and AI SEO the same 2:09 thing cuz there's kind of a debate going 2:11 on on if you're doing SEO, you're 2:13 already doing AI SEO. So we're going to 2:14 explore whether that's true. 2:16 Should you focus on SEO or AI SEO? Where 2:19 it makes sense, where it might not make 2:20 sense. And then I want to show a case 2:23 study giving you 2:24 blueprint essentially on how you can win 2:26 in AI search. And I'll we'll have some 2:27 actual results and make it really 2:29 tangible for you guys. So let's just 2:30 jump right in. How do large language 2:33 models work? So we have you know, we'll 2:35 use Chat GPT as an example, but this 2:37 applies to pretty much all major large 2:39 language models. They essentially going 2:42 to digest all this content and words 2:44 from different sources. So they're going 2:45 to take every a bunch of articles on 2:48 news websites. They're going to take 2:50 Wikipedia. They're going to look at all 2:51 the Reddit threads, even published 2:53 books. They'll look at YouTube 2:54 transcripts and even the rest of the 2:57 web. And they're going to take all this 2:59 content. We're talking maybe 1 to 2 3:01 trillion words of content and they're 3:04 going to put it into one massive text 3:06 file. It's one one massive document and 3:08 they're going to run a whole bunch of 3:10 computing power against that one 3:11 document to look for statistics and 3:14 patterns on you know, which words are 3:16 most likely to come next across this 1 3:18 to 2 trillion word text file. And that's 3:21 why we have all these massive data 3:22 centers that are sprouting up to do all 3:24 the computations required to look for 3:26 patterns between words. 3:28 So by doing this and by doing all this 3:31 analysis on this massive data set of 3:34 words, 3:35 we can get patterns of which words are 3:37 most likely to come next. So for 3:38 example, let's say you have the sentence 3:41 the boy went to the dot dot dot. 3:44 Statistically speaking from all the 3:46 information that the large language 3:48 models were trained on, you know, the 3:49 most likely word to come next could be 3:51 playground. But it could also be school 3:54 or it could be park or hopefully not 3:56 hospital. But this is just kind of a 3:57 statistical probability of what word is 4:00 likely to come next based off all the 4:02 words that the the AI model has 4:05 consumed. So I'll give you kind of a 4:07 more 4:08 business use case for this. So if I was 4:10 to ask Chat GPT what word is most likely 4:13 to come next? The best brand for 4:16 electric mountain bikes is dot dot dot. 4:19 And we can we can see what word is 4:21 likely to come next. It's filling in the 4:22 blank. So we can see Specialized or Trek 4:25 or Giant. There's a lot of these notable 4:27 brands because 4:29 Chat GPT in this case took in all this 4:32 information from articles, Wikipedia, 4:35 Reddit. And statistically speaking, 4:37 Specialized was the most was the brand 4:39 most likely to come next from that 4:41 training data. And sure enough, if I ask 4:44 Chat GPT what are the best electric 4:47 mountain bikes, we see Specialized 4:49 number one. We see Trek number two. And 4:51 we see some of those other brands 4:52 mentioned. Now it's not going to be a 4:54 one-to-one because these large language 4:56 models are probabilistic, meaning you're 4:58 never going to get the same results 4:59 back-to-back. But there's a lot of 5:02 correlation here where getting showing 5:04 up in the training data 5:06 can impact how you know, what types of 5:08 outputs Chat GPT's going to happen. So 5:10 what does all this mean? Well, 5:12 essentially what it means is the more 5:15 mentions of your brand you can get in 5:17 the large language model training data, 5:19 that's you know, Wikipedia, Reddit, 5:21 articles, YouTube transcripts, etc. with 5:24 positive sentiment, the more often Chat 5:27 GPT, AI overviews, Perplexity is going 5:30 to recommend your brand versus your 5:32 competitor. So this is the foundation of 5:34 how these large language models work. 5:39 So with that, now that we have kind of a 5:41 primer on how LLMs work and how AI AI 5:45 search works, I want to talk a little 5:46 bit about SEO versus AI SEO. Cuz I hear 5:49 a a lot of things out there. Some people 5:51 say it's completely different, some 5:52 people say the same. Honestly, a lot of 5:54 the talk about SEO and AI SEO is the 5:57 same thing is coming from SEO companies 5:59 who maybe don't want to put in the time 6:01 and effort required to adapt to this new 6:03 trend. Takes a lot of time to learn a 6:05 new skill set and kind of figure out 6:07 what's true versus what's not true and 6:09 what actually works. So I'll give you 6:10 kind of a quick breakdown. 6:12 If we look at SEO versus AI SEO, what's 6:15 the goal? Well, with SEO, it's going to 6:17 be ranked higher to get more traffic and 6:19 revenue within Google and other search 6:21 engines. And that's going to come from 6:22 those 10 blue links that we've loved for 6:25 so many years. With AI SEO, it's more 6:27 about getting recommended by AI 6:29 assistants. So it's almost agnostic to 6:31 your website. It's just getting your 6:32 brand recommended and that's going to be 6:34 in Chat GPT, AI overviews, Perplexity, 6:36 other AI search engines. 6:38 With content, if you're doing SEO, 6:41 you're going to focus on keywords that 6:42 have high search volume and topics that 6:44 are getting searched by your customers 6:46 that can rank in Google and then drive 6:48 traffic. With AI SEO, it's more about 6:51 focusing on topics that Chat GPT and 6:53 other AI assistants are citing for the 6:56 prompts and queries you want to target. 6:58 When it comes to outreach, SEO it's all 7:00 about backlinks. We want to improve the 7:02 site authority. With AI SEO, it's about 7:04 brand mentions like we talked about 7:06 earlier. Want to increase the repetition 7:08 of your brand with a positive sentiment 7:10 in the AI training data and in the cited 7:12 sources. 7:13 With prospecting, we're looking for 7:15 websites that have high domain rating 7:16 and high traffic. With AI SEO, we want 7:19 to focus on websites that are are are 7:22 already getting cited often in these 7:24 large language models. 7:25 >> And you find that out just through what 7:28 they're sourcing. So on Chat GPT and 7:30 Claude and whatnot for searches that 7:33 someone would want to rank for, you're 7:34 just looking at all right, where is it 7:36 actually pulling from the most? That's 7:38 exactly it. If you want to take a manual 7:40 approach, let's say you want let's say 7:42 you sell protein powder. You could ask 7:44 Chat GPT what's the best protein powder 7:46 brand and look at the you know, go to 7:48 the the bottom, you click the sources, 7:50 you can see all the sources it's pulling 7:51 from and you could be like okay, like 7:53 these websites, these articles are 7:55 getting pulled into Chat GPT. I want to 7:57 make sure my brand is mentioned in these 8:00 places. And there's tools that will kind 8:02 of aggregate this for you which I'll 8:04 I'll talk about a bit later where it 8:05 scales the process so you don't have to 8:06 manually do the searches. But in theory, 8:09 yes, that's exactly how it work. 8:10 >> Okay. 8:13 So optimization with SEO, it's all about 8:15 keywords. Want to include our target 8:16 keywords in the content with related 8:18 keywords. With AI SEO, it's more about 8:20 just giving AI assistants enough 8:23 information about your brand and your 8:25 products and your service so so they 8:27 feel confident recommending you to 8:29 users. So it's FAQs, use cases, 8:31 specifications, things like that. And 8:33 with KPIs and reporting, SEO it's all 8:36 going to be about search revenue, 8:37 organic search traffic, keyword 8:39 rankings. On the AI side of things, 8:42 really post-purchase surveys asking 8:44 people how they found about you how they 8:46 found out about you is one of the best 8:47 places for attribution just cuz there's 8:49 not it's a it's a little less tangible 8:51 than traditional marketing channels. 8:53 >> I agree. I I I tell all of our clients 8:56 if especially e-com, right? If you're 8:58 not doing that, 8:59 >> [clears throat] 8:59 >> then you're missing out on a key piece 9:01 of attribution cuz you can't try 9:04 forget and this goes into like ROAs and 9:06 all that stuff. Like it helps you guide 9:08 in app, but that should not be your 9:10 north star as a business, right? And so 9:13 let's take all of these touch points and 9:15 actually the one which is the simplest 9:19 is just literally a post-purchase 9:21 survey. I Grow My Ads, we we ask where 9:24 you've heard from us 9:25 when you fill a form out for the free 9:27 audit. And it's the number one data 9:30 point I use for the entire agency in 9:31 regards to how I look at our own 9:33 attribution here at at the agency. 9:36 Yeah, I mean there's enough attribution 9:38 issues already with some of these 9:39 platforms and then you got some mobile 9:41 devices or iOS or there's just there's 9:43 limitations there. So yeah, couldn't 9:45 agree more with that. And you know, 9:47 looking at this, you might realize it's 9:50 a lot of the same activities. You're 9:52 still creating content. You're still 9:54 doing outreach. You're still optimum 9:56 some website optimization, content 9:58 optimization. It's just the goal post is 10:00 different. You're looking for different 10:02 things. So, again, a lot of the same 10:04 activity with just different goals and 10:07 metrics that you're going after. 10:10 So, should you focus on SEO or AI SEO? 10:14 Well, I have a bunch of data I want to 10:16 share that's going to help us answer 10:17 this. We're going to look at market 10:18 share, we're going to get trends, we're 10:19 going to see where you can get the best 10:21 ROI between these two marketing 10:23 channels. 10:24 So, first let's talk about just overall 10:26 Google search volume trends. Um I took a 10:29 sample of 2,000 transactional keywords 10:32 related to, you know, shopping. So, 10:33 these are going to be keywords like 10:35 protein powder, air fryer, microwave 10:37 oven. I mean, you you name it across 10:39 every single vertical just to get a a 10:41 large sample of what are people 10:43 searching for. I then pulled search 10:45 volume data for each of them. I 10:47 aggregated it together, and this is just 10:48 going to get kind of give us a sample of 10:50 what do trends look like over time. So, 10:53 as expected during COVID, you couldn't 10:55 go to the store to buy things. Things 10:57 spiked up quite a bit, and if this 10:59 pattern kind of falls where we would 11:00 expect, you know, in the holiday season, 11:02 there's a lot more search activity, 11:04 drops off with the new year, and we see 11:05 that pattern over time. 11:07 So, I wanted to zoom in on just kind of 11:09 the post-COVID uh trends, and this is 11:12 what that looks like. So, it's, you 11:14 know, people for a while thought ChatGPT 11:17 is going to be the Google killer. You 11:18 have ChatGPT or Perplexity. People are 11:20 going to stop using Google search. Uh 11:23 it's going to go down. Well, we're not 11:25 really seeing that in the data. In fact, 11:27 we're actually seeing the opposite. 11:28 Despite ChatGPT, despite these large 11:30 language models, we're seeing Google 11:32 search market share still continuing to 11:35 increase. In fact, in just the last 12 11:37 months, it went up about 3.3%. 11:40 So, the trends is there. Google people 11:42 are still using Google search. We 11:44 haven't seen that decline just yet. 11:46 And that that is from Google search 11:48 trend data or This is coming from 11:51 Ahrefs, which has a big Yeah, a bunch of 11:54 clickstream data. Yeah, okay. 11:56 Yeah. 11:58 Um let's dive into just ChatGPT. So, um 12:02 you know, I work with a lot of brands. I 12:04 We have lots of analytics data. I was 12:06 able to aggregate it together for about 12:08 94 e-commerce stores just to see how 12:10 much traffic ChatGPT is sending over 12:12 time. Um in April is when ChatGPT 12:16 started putting product carousels with 12:18 images in the results, and after that 12:21 happened, we saw noticeable improvements 12:22 in ChatGPT sending traffic. However, 12:25 that started to level off in November, 12:28 and actually went down December, which 12:29 is surprising cuz you would think 12:31 November and December is going to have 12:32 the highest uh search volume and highest 12:34 traffic. This is when most people are 12:35 doing their Christmas shopping. Well, it 12:38 it definitely plateaued a little bit at 12:40 the end of 2025 going into 2026. And 12:43 it's not just my data showing this. 12:46 There was a a massive study of millions 12:49 upon millions of of users, and it showed 12:51 the exact same thing. This this green 12:53 line here 12:55 is how many people are using ChatGPT. It 12:57 peaked in September and has been 12:59 steadily declining since September of 13:01 2025. So, it looks like it's going down, 13:04 while at the same time we're seeing 13:06 usage of Gemini steadily going up. So, 13:09 according to a lot of different data 13:11 sources now, the peak of ChatGPT may be 13:14 behind us, and they might be on the 13:17 decline right now. 13:20 And I was curious like I want to see 13:22 what overall AI search adoption looks 13:25 like. So, I literally took this, I fed 13:27 it into Claude, and I wanted to see the 13:29 total. So, this blue line is now kind of 13:32 combining both together, and it does 13:34 look like there's quite a bit of more 13:37 and more people using these AI search 13:39 tools, whether it's ChatGPT or Gemini. 13:41 Uh so, it has been increasing quite a 13:43 bit. A lot more people are using these 13:45 AI search tools than before. Sure. 13:49 Yeah, I even look at my own usage since 13:53 September uh where you showed ChatGPT 13:55 peaked. I don't even use ChatGPT 13:57 anymore. Um I'm like 14:00 fully on Claude pretty much. 14:03 Uh 14:04 >> [laughter] 14:04 >> and we will uh Luke will use uh Codex 14:09 for stuff. 14:11 Um so, he uses different 14:13 uh LLMs for different uses. Um but 14:17 for the most part, I have found I'm just 14:20 on ChatGPT or sorry, just on Claude, and 14:24 then I'm now using uh Cohere a little 14:28 bit, and then, you know, uh Claude Code 14:31 for the nerdy stuff that we're when 14:32 we're working on like automations and 14:34 stuff for workflows. 14:36 But, I have I don't think I can't 14:38 remember the I even 14:39 I switched out I had ChatGPT on my 14:42 phone, and I switched that out for 14:43 Claude, and I didn't do it for I know 14:45 there was like a political thing that 14:46 happened. Um and so, they had this mass 14:50 amount of users that deleted because of 14:52 that. That's not why. I just naturally 14:54 gravitated over to Claude um throughout 14:57 time, which I thought was interesting. 14:59 I've always thought the outputs are 15:01 better, and it hallucinates less. Like I 15:03 use Claude. I think most people in the 15:05 business community are probably using 15:07 Claude now. Um for the everyday consumer 15:10 that's not technical, like my mom's not 15:12 using Claude. She's probably going to 15:13 use ChatGPT or Gemini. So, across like 15:15 that consumer segment, yeah, ChatGPT, 15:18 it's it's surprising to see, but they're 15:21 kind of on the the downturn now. Well, 15:23 right. I 15:24 The whole thing's again that that, you 15:26 know, you can go down rabbit holes of 15:28 the the industry, and they want to IPO 15:31 and all these things, right? And is this 15:33 a bubble still and and and whatnot. But, 15:36 uh it's interesting to see then like 15:38 search usage there, the adoption of 15:41 users using these tools. We know it's 15:43 going to increase, but it's interesting 15:45 to see ChatGPT's obviously not, you 15:47 know, like dominating and continuing to 15:50 to exponentially grow. 15:52 Yep, agreed. 15:54 The other thing we we can talk about is 15:56 AI mode. Like Google's been getting more 15:58 aggressive with AI mode. It's now in the 16:00 Chrome search bar. It's now in the 16:02 Google homepage. Even with all that, 16:04 it's still not even kind of scratching 16:07 the surface of sending traffic. So, 16:10 it only sends less It still only sends 16:12 less than 1% of all traffic on the web. 16:15 You know, you sure it has been steadily 16:17 growing, but it's just not growing as 16:20 fast as you think it would for literally 16:22 being on the google.com homepage Now, 16:24 it's other places. So, this this 16:27 this graph is showing traffic a click 16:30 that it generates to your a website, 16:32 correct? Yes, across all web traffic of 16:35 you know, the massive sample of millions 16:37 upon millions of users 16:39 based on their usage, like how much 16:42 traffic is AI mode 16:44 But, isn't that kind of then the the 16:46 argument where it's like, well, AI mode 16:48 is just going to kill traffic for any 16:51 type of search that it can literally 16:53 just give an answer to you because of 16:55 now being 16:57 using its LLM technology for that? 17:00 Yeah, that I mean, I'm when I was 17:02 looking at this, it looks like monthly 17:03 visits to AI mode's web pages. I'm not 17:06 100% sure if that's this the main AI AI 17:08 mode chat interface of users actually 17:10 using that or how much traffic AI mode 17:12 sent. But, regardless of either of 17:15 those, it's even even if it was both, 17:17 for example, it's still pretty minimal. 17:19 So, what this tells us is it's not AI 17:22 mode really isn't a big priority right 17:24 now. It's not this new thing that you 17:26 have to drop everything and get into AI 17:27 mode. Um it's it's pretty minimal, 17:30 although it is trending up. Instead, 17:33 what I am seeing is a more roll out 17:35 about AI overviews. So, I initially 17:38 thought Google was going to fight 17:40 against ChatGPT with AI mode. They're 17:42 going to have this ChatGPT competitor. 17:44 It's going to have its own separate It's 17:45 going to be a new product. But, I think 17:47 what's happening is it's way more 17:49 difficult to get people to use a new 17:52 product. Everyone's been using Google 17:53 search. Trying to get them to use a a 17:55 different interface is not easy to do. 17:58 So, instead of moving people to a 18:01 different product, they're moving the AI 18:03 to the traditional search experience, 18:05 and we're seeing that more and more. 18:08 So, like you guys might have seen these 18:10 AI overviews. Um initially, so back in 18:13 November 10th, there was this great 18:14 study done by Ahrefs. And what they 18:16 found was 99% 18:19 of the AI overviews was for 18:20 informational keywords. It was It was 18:22 only 1%, about 1.2% going to 18:25 transactional queries where someone's 18:27 looking to buy a product. Well, I ran an 18:30 updated study just a few weeks ago, 18:32 and it jumped from about 1.2% to 14%. 18:36 So, now 14% of transactional queries 18:39 have these AI AI overviews. And if you 18:41 look at any kind of major e-commerce 18:43 site or any even a niche e-commerce 18:45 site, you'll see the same trend. So, 18:48 like I used Walmart and Target as a 18:50 benchmark. Starting November 1st, we saw 18:52 this massive uptick of AI AI overviews. 18:56 And literally every e-commerce site 18:58 where I've done this check has the exact 19:00 same pattern. It's very clear Google's 19:02 getting more aggressive with AI 19:04 overviews, not just for informational 19:05 queries, but now for shopping queries 19:08 and transactional queries. 19:12 So, it's it's coming. 19:14 Interesting, yeah. 19:16 So, we've looked at trends. Um let's 19:18 just talk a little bit about where 19:19 things are right now at this point in 19:21 time. So, this is at the end of 2025. Um 19:24 traditional search engines still have 19:27 the the lion's share. This is pretty 19:28 much all Google. We're talking about 81% 19:31 of search market shares going on 19:32 traditional search and SEO. 19:35 AI search is just a little bit over 3%. 19:38 Um so, very minimal. 19:40 Um the the tricky part is this 80% with 19:43 AI overviews, the lines are blurred now 19:45 between traditional SEO and AI search. 19:47 We don't know what percentage of this is 19:49 coming from uh just traditional Google 19:52 search the 10 blue links or those AI 19:54 overviews. So, 19:56 either way, it does look like, you know, 19:58 SEO isn't dead just yet. They still have 20:00 a lot of people searching for things in 20:02 Google and AI these AI tools haven't 20:05 really quite grown as quickly as people 20:08 thought they would. 20:11 So, what does all this mean? Well, first 20:14 off, Google search is still the king of 20:16 search and it's growing more and more 20:18 each year. We're talking 41 times more 20:20 traffic than ChatGPT and it's sending 20:22 about 33 times more revenue than 20:23 ChatGPT. 20:25 ChatGPT's growth seems to have flattened 20:27 out as Gemini gets more market share. 20:31 AI mode, yeah, it's also gaining some 20:33 market share but still like less than 20:34 0.1% of web traffic, so it's not really 20:36 worth focusing on just yet. 20:39 Instead of Google pushing AI mode, it 20:42 looks like they're getting more 20:43 aggressive with these AI overviews and 20:46 traditional SEO still has a much better 20:49 ROI potential than AI search just given 20:52 the large market share. The more 20:54 searches done on traditional search 20:56 engines, the more clicks on those 10 20:57 blue links makes SEO more viable. 21:00 However, 21:01 the 21:02 AI SEO is at the point now where it has 21:05 a decent ROI potential for many brands. 21:07 We'll talk about, why it would be better 21:09 for some brands versus others, but 21:11 literally in the last 15 months, AI 21:13 search adoption has grown by 56%. So, we 21:17 are seeing some brands get a decent ROI 21:19 with it. Not all brands, it depends on 21:21 are your customers using ChatGPT or 21:23 large language models for it, but it's 21:25 already at the point where it has 21:26 potential to send decent traffic and 21:28 revenue. You know, the interesting thing 21:31 is 21:32 like we know, if you're on page two of 21:35 Google, you're you might as well be on 21:36 page a million, right? Um but 21:40 if you're on page two for a term, you as 21:44 the owner of the business or the 21:45 marketing director or whomever, you you 21:48 still see it and you're like, okay, 21:50 we're like we're making our way there. 21:51 Unless you have software and stuff 21:53 you're tracking this with like Ahrefs 21:54 and whatnot. With 21:57 with like the LLMs, it really feels like 22:00 do or die. It's like, I'm either on it 22:03 or I there's no page two. It's like, 22:05 who's the best plumbing company in 22:07 Miami? It's like, here's the top five. 22:09 It's [laughter] like, if you don't make 22:10 the cut, you're not even recommended at 22:13 all. Now, that's within that search you 22:15 said obviously it does change in things. 22:18 That would be an interesting question 22:19 maybe later. I don't want to derail too 22:20 much. Is there local AEO differences 22:24 between an e-com business and whatnot? 22:26 So, or a national you know, a B2B SaaS 22:29 or Legion or an agency like GrowMyAds, 22:31 right? 22:33 And so, those are 22:35 questions that I've had because like a 22:37 lot of local businesses 22:39 aren't super sophisticated and how many 22:41 publications can they really get on and 22:43 you know, what review platforms actually 22:45 make sense to them versus like a 22:47 national level company, those types of 22:49 things. But yeah, like to me it is it 22:51 does feel as a business owner, I think 22:54 we do okay. 22:56 If you were to type in like best Google 22:57 Ads agency or something, I don't know if 22:59 we'll be on all of them, but I know we 23:01 we do get leads from that. But I think 23:03 if I did a search and I didn't see my 23:04 name on a list that is given, I would 23:07 feel like, how much money is it going to 23:09 take to get myself on the list? 23:11 Cuz it's like an ego thing a little bit, 23:13 too. 23:15 Yeah, and it's also tricky cuz you could 23:16 ask ChatGPT five times and it's going to 23:18 mention different brands each time. So, 23:20 you might show up, 20% of the time, you 23:23 might show up 100% of the time. It's it 23:25 can rankings are static. If you're 23:27 ranking number one today, you probably 23:29 were ranking number one yesterday, you 23:30 know, probably ranking number one 23:30 tomorrow. 23:32 ChatGPT, it's a bit all over the place. 23:33 Right. Right. Yeah. 23:36 Just another data point that I I want to 23:38 share. 23:39 I of the 94 e-commerce brands I 23:42 mentioned before, I looked at the 23:43 average conversion rate on organic 23:45 search versus ChatGPT and I tried to get 23:48 it apples to apples as much as possible. 23:50 So, I filtered out traffic to the 23:52 homepage, which is usually branded 23:53 traffic and that's going to skew up and 23:55 inflate it. I also filtered out any 23:58 traffic to like blogs and informational 24:00 pages cuz that usually has a lower 24:02 conversion rate. So, this is just 24:03 looking at conversion rates to category 24:05 pages and product pages. 24:07 And so, side by side, you can see that 24:10 ChatGPT on average converts about 33% 24:13 higher. So, it is higher quality traffic 24:15 and people tend to trust the 24:16 recommendations a bit more when they're 24:18 getting it from ChatGPT. Interesting. 24:22 All right. So, based on this, 24:24 bottom line, should you focus on SEO or 24:27 AI SEO? What I would say is focus on SEO 24:30 if there's high search volume for your 24:32 target keywords in Google search and if 24:34 there's minimal AI overviews for the 24:37 keywords you're trying to rank for. Now, 24:39 AI SEO, that's going to be more 24:40 beneficial if there's high volume for 24:42 your prompts in ChatGPT. So, like Ahrefs 24:45 brand radar, they have a database of 24:46 millions of prompts. 24:48 You can just put in some keywords and 24:50 see like how often are people searching 24:51 for these in ChatGPT. And then if there 24:54 was lots of AI overviews showing up for 24:56 your target keywords, that might also um 25:00 make the argument that AI SEO could be a 25:02 better fit for your brand. 25:03 >> Is there Is it easier right now to rank 25:08 for AI SEO versus traditional? Like, do 25:12 you find, hey, some of these keywords, 25:15 good luck. There's companies that, you 25:17 know, they've been going after this for 25:19 five years and it's going to cost you an 25:22 insane amount of money and time to ever 25:24 come to what they already have versus 25:27 actually AI SEO, here's what is being 25:31 sourced and we actually think it would 25:33 be way easier to just go rank over there 25:35 right now. Do you Are are you finding 25:36 that in in situations at all? 25:39 So, that that's a great question and 25:41 probably like a perfect segue to, you 25:44 know, how easily can you manipulate 25:46 ChatGPT? 25:47 You know, with with SEO, some keywords 25:49 are going to be easy, some keywords are 25:51 going to be difficult to rank for. With 25:53 AI SEO, this is brand new. Like if we 25:55 were to talk about how hard is it to 25:57 rank in Google, I got, you know, over a 25:59 decade of experience and case studies 26:01 and I can I can paint you a clear 26:03 picture. This is so new that there's we 26:05 don't really have that yet. We don't 26:07 have this track record we can pull from 26:09 or look for. There's minimal case 26:10 studies out there and the case studies 26:12 that are out there are a bit fuzzy. But 26:14 what I what I can do is kind of share 26:15 with you, 26:17 we we did this pilot program with 26:18 Private Label MFG. They're an e-commerce 26:20 brand and I'll show you what we did and 26:23 what the results were and how soon those 26:25 results hit and that can at least kind 26:27 of give one data point in answering that 26:28 question. 26:29 >> Okay. 26:31 So, Private Label MFG, they sell 26:33 aftermarket car parts for Hondas, 26:36 Subarus, BMWs, other other different 26:39 cars and with we started in September 26:42 and within just a few months their AI 26:44 referral revenue in analytics shot up 26:46 quite a bit. But we we also running 26:49 post-purchase surveys. Every time 26:51 someone checked out on the checkout 26:52 confirmation page it'd say, hey, how'd 26:54 you hear about us? Well, chat you know, 26:56 we had one question for ChatGPT and AI 26:59 search that went from half a percent of 27:01 revenue to 5% of a revenue. So, it came 27:04 it became a very notable revenue channel 27:07 in just a we're talking three to four 27:09 months here. 27:10 >> Wow. 27:12 So, here's a high-level overview of what 27:15 we did. First, we kind of did initial 27:17 audit to build out the AI search 27:19 strategy to figure out what we needed to 27:20 do. We did some of the foundational 27:22 things like prompt research, content 27:25 strategy, optimizing the site, 27:27 optimizing the content. But then the 27:29 three main pillars of AI search is going 27:31 to be content creation, brand mentions, 27:35 and then Reddit marketing. And I'll show 27:37 you what the results were from doing all 27:38 these things. We'll go through this 27:40 section by section. So, let's start off 27:42 with 27:42 >> that's an interesting one cuz it's like 27:44 that, 27:45 you know, Reddit's so anti 27:47 uh marketing and and commercialism and 27:51 that type of stuff. So, I'm always 27:52 fascinated by people who have like these 27:54 Reddit plays that do work versus just 27:56 being like, hey, become just go be a 27:58 normal user on there and answer 27:59 questions. 28:01 Yeah, and that that what you just 28:02 described is helpful, too. There's also, 28:04 you know, you can talk about products in 28:06 kind of a a more natural helpful way. 28:09 So, there's there's kind of ways you can 28:10 attack from both angles. 28:13 So, starting the foundation side, the 28:16 very first step of any AI search 28:18 campaign is going to be prompt research. 28:20 Just like with paid search, you're going 28:22 to do keyword research. With SEO, you're 28:24 going to do keyword research. Well, with 28:26 AI AI SEO, we're going to do prompt 28:29 research. So, what I recommend doing is 28:31 if you don't, you know, do some 28:32 traditional keyword research to see what 28:34 are your customers searching in Google. 28:36 There is so much data out there whether 28:38 it's from Google Keyword Planner or 28:39 Ahrefs or SEMrush on what people are 28:42 searching for. And then what you can do 28:44 is extrapolate that out to see what is 28:46 someone likely asking when they search 28:47 this into Google. So, if someone 28:48 searched cold air intake, maybe they're 28:51 asking what's the best cold air intake 28:52 for my car. Or for motor mount, maybe 28:54 they want to know what's the best 28:55 replacement motor mounts. So, it's not 28:58 perfect, it's not exact science, but 29:00 it's the best way you can get some 29:02 accurate information on what people are 29:05 likely searching for. 29:07 Yeah, so this is just kind of a helpful 29:09 way. You know, some people say like just 29:10 come up with every use case and come up 29:12 with all these prompts, but you know, if 29:13 no one's searching a keyword, you're and 29:16 has no search volume, you're not going 29:17 to focus on it. So, why would you focus 29:19 on prompts that also don't have any 29:20 search volume? 29:24 Then what we did is we took all those 29:26 prompts, we put into this prompt 29:27 tracking tool. So, I kind of briefly 29:29 alluded to this earlier. There's a lot 29:32 of these tools out there. I think 29:33 Profound is probably the the biggest, 29:35 most notable one. I've tested about 12 29:37 of these things and hands down my 29:39 favorite is Peak AI. So, Peak AI, you 29:43 can put in your prompts, it's going to 29:44 pull results from Perplexity, ChatGPT, 29:47 AI overviews, wherever you want, and 29:49 it's going to tell you your visibility 29:50 score, which is essentially what percent 29:53 of the time did chat GBT, did Google AI 29:56 overviews recommend your brand versus 29:58 the competition? When we first started, 30:01 it was a lot of zeros. They had no 30:02 visibility. Chat GBT AI overviews, they 30:05 were not recommending private label at 30:07 all for any of these prompts that we 30:09 were tracking. 30:13 And on aggregate basis, they were far 30:15 behind their competition. They were only 30:17 getting recommended 1% of the time. So, 30:19 literally pretty much invisible in AI 30:22 search. 30:23 So, we had our benchmarks down. We just 30:25 started going to work. One of the first 30:27 things we did was added FAQ content to 30:29 their main collection pages. Um and 30:32 again, we're just asking ourselves like 30:33 what what custom what questions do the 30:35 customers have? We worked with the 30:36 customer service team just to kind of 30:38 see what they're getting a lot. And we 30:40 made sure to answer those on the 30:42 category pages. Now, the reason FAQ 30:44 content works so well is it it's kind of 30:46 a natural fit to how these large 30:48 language models work. You ask it a 30:50 question, it gives you an answer. So, if 30:52 you already have this question format um 30:55 on your website, it makes it very easy 30:57 um to kind of get recognized within uh 31:00 the large language models. So, it's 31:01 going to be more easy for them to 31:03 understand, you know, if someone asks a 31:05 similar question, it already has kind of 31:06 a response that it can pull from. Now, 31:09 does it 31:10 >> [clears throat] 31:10 >> you know, adding FAQs obviously helps 31:13 for, you know, uh the customer as well, 31:15 right? The customer journey, but 31:18 does it do these FAQs have to live like 31:21 do you have to go build out individual 31:23 pages or bake these into like the 31:25 product pages, the category pages? Or 31:28 have you found companies who have like 31:29 they've built out like easy uh readable 31:33 or, you know, uh indexable FAQ sections 31:37 completely that just like the LLMs can 31:39 pull from? Is there like a play like 31:41 that or is it just no, just put it on 31:44 the pages that makes sense right now on 31:45 your site? I'd say all the above. So, 31:47 like on collection pages, 31:49 put FAQ answers to questions that are 31:51 common about that category page. And 31:54 then maybe do the same thing on your 31:55 products. And maybe have a standalone 31:56 FAQ page that talks about just overview 31:58 of questions about the business. And 32:00 maybe you have a hub for like kind of 32:01 more like nuanced questions. So, I think 32:04 just the more helpful content you can 32:06 have that customers are actually asking 32:09 is only going to make things easier. 32:10 >> [snorts] 32:11 >> And then what all right, so like this 32:13 collection page, for traditional SEO, 32:16 I'm sure you already had content there, 32:18 right? Yeah. So, now you're just baking 32:20 in FAQs into the traditional content for 32:25 traditional SEO that you already had. 32:27 Yeah, exactly it. So, like we'd have the 32:29 content block. You can't see it, but it 32:31 was higher up on the page. And then the 32:33 bottom of the page is when we were 32:34 adding these FAQ blocks. 32:35 >> So, FAQ doesn't it can just go to the 32:37 bottom. It doesn't need to be like, oh, 32:39 make sure this is up front or on the top 32:41 of the page like with traditional SEO, 32:43 those types of things you have to be 32:45 cautious of, correct? 32:46 >> Yeah, exactly. Go below the product 32:48 grid. That way it's not pushing things 32:50 down as far. And um you know, it can 32:52 still be helpful for users, but it also 32:54 gets picked up by AI crawlers. Okay. 32:58 That was on the collection pages. Let's 33:00 talk about products. So, we looked at 33:01 all their top products, and we want to 33:03 see how well optimized their products 33:05 were for large language models. So, 33:07 we're looking things like did it have a 33:09 clear specifications table? Talk about 33:11 like the weight, the size, the material, 33:12 the compatibility, all the kind of 33:14 nitty-gritty information about the 33:16 product. We want to see did it do a good 33:19 job of showing unique selling points? Um 33:22 was there any talk about use cases or 33:24 target audience or compatibility? Um and 33:27 then also like we talked about before, 33:29 having FAQs on the product pages. So, 33:31 these are the things that you could do 33:32 for an e-commerce product page. Some of 33:34 this you could introduce in just like a 33:35 traditional landing page just to make 33:37 sure these large language models have 33:39 enough information about your brand and 33:41 your products where it feels confident 33:43 recommending you over the competition. 33:47 And then kind of like what you mentioned 33:48 before, Austin, like again, we could 33:50 have a separate FAQ page that has a 33:51 bunch of information. So, for them we 33:53 had one kind of global FAQ page that 33:56 talked about shipping, returns, you 33:58 know, where's the business located, um 34:01 payment methods, website security, all 34:03 that kind of general information in one 34:05 place so that could get pulled into 34:06 these large language models and just 34:08 give them more information about the 34:09 company and the website. Okay. I didn't 34:11 realize that would matter too much 34:13 on if they're going to rank you or not 34:15 for a general prompted search. So, yeah, 34:19 I mean, that's a good point to give a 34:21 little bit of a disclaimer. I cannot say 34:23 for sure which of these had the biggest 34:26 impact and which of these were minimal. 34:28 But this is all kind of based off theory 34:31 and best practice and how the large 34:33 language models work. So, we're just we 34:35 know these things you know, could have 34:37 an 34:38 could have a positive impact. Worst case 34:40 scenario, they do nothing. But we're 34:42 just throwing everything at the wall, 34:43 see what sticks. And this is just one of 34:46 those things that should help cuz we do 34:48 see brand FAQ pages get pulled into the 34:50 sources. So, we wanted to kind of 34:52 replicate that with their own FAQ page. 34:54 Mm, interesting. We also did a content 34:57 audit specifically looking for which 35:00 articles were less relevant to their 35:02 core offering. So, they sell aftermarket 35:04 car parts. They sell exhaust systems. We 35:07 want to see were there any articles that 35:08 might be a bit of a stretch? So, like, 35:10 you know, uh tips that make your car 35:12 look brand new, maybe that's not as 35:15 closely relevant as some of the other 35:16 ones. Or um you know, chat GBT has 35:18 spoken about private label 35:19 manufacturing, maybe that's not going to 35:22 be on point. And what we found is the 35:24 types of sites that AI search engines 35:27 and even traditional search engines um 35:30 favor the most are typically more niche 35:32 specific, they're more hyper-focused. 35:35 What we found is as a website gets more 35:37 broad and covers more topics, it becomes 35:39 kind of diluted for their core offering, 35:41 their core topic. So, we try to keep the 35:43 websites dialed in and buttoned up. And 35:46 you know, if you're listening to this, 35:47 going through your blog, you know, 35:48 sometimes you think it's going to be 35:49 helpful talking about a tangential thing 35:51 or kind of stretching the relevancy 35:53 bounds, it could hurt both traditional 35:56 SEO and AI search. So, it's usually good 35:57 to keep things on topic as much as 35:59 possible, at least from like a SEO AI 36:02 SEO perspective. 36:07 All right. So, that was kind of the 36:08 foundational thing. Then there's like 36:10 three main kind of levers we had to move 36:12 things forward. So, let's talk about 36:13 content creation. 36:15 Um 36:16 I've reviewed a whole bunch of outputs 36:17 from chat GBT, AI overviews, and these 36:20 other large language models. And the 36:21 types of content formats I see cited the 36:24 most is buyers guides, cost guides, 36:27 product roundups hands down is probably 36:29 number one, competitor comparisons, and 36:32 then even glossary pages. Those get 36:34 cited very quickly. So, for private 36:36 label manufacturing, we created content 36:38 in these areas with a heavy focused on 36:40 these best of articles and also 36:43 competitor comparisons. 36:45 Uh here's a bit Here's what it looked 36:46 like. So, for like product roundups, 36:48 we'd have an article such as top five 36:50 best exhaust headers for Honda Civic. 36:53 And we'd list the different products 36:54 they have. We talk about um who it's 36:57 for. We talk about some of the key 36:58 features. And this is just kind of new 37:00 content that could get picked up and 37:02 cited by chat GBT to help shape the 37:05 recommendations of which brands they're 37:07 going to suggest in the outputs. 37:11 We also do the same thing for competitor 37:12 comparisons. So, we looked at all their 37:14 main competitors, and we did kind of a 37:16 side-by-side comparison. We talked 37:18 about, you know, why private label is a 37:20 better choice, you know, better pricing, 37:23 high-quality standards, all these 37:25 things. We include these tables like 37:26 this. Again, just it's another thing 37:29 that these large language models can 37:30 digest in their training data, and it 37:32 can kind of help shift the conversation 37:34 around which brand should be 37:35 recommended. 37:38 But creating content's not enough. You 37:40 want to make sure your content's 37:41 optimized. So, there's this fascinating 37:43 study done by Princeton that looked at 37:45 why do some articles get cited over 37:48 others? And what they found is just 37:50 having credible quotes from experts can 37:52 get you cited 30% more. Um having 37:55 quantitative statistics can also help 37:57 you get cited. Using simple language so 37:59 it's really easy uh to understand um 38:02 goes a long way. And if you're 38:03 wondering, you know, what does simple 38:05 language look like? The the best example 38:07 I could give is Wikipedia. It's very 38:09 matter-of-fact, straightforward, no 38:11 fluff. This happened, that happened. 38:14 That is the gold standard for writing 38:16 for not just AI search, but it works 38:18 really well for traditional SEO. So, um 38:21 you know, if you have a brand voice, you 38:23 know, definitely stick with it. Don't 38:24 throw everything out the window just for 38:26 AI search. But if you're going to try to 38:28 go as optimal as possible for AI search 38:31 using uh simple language like Wikipedia 38:33 can go a long way. Um also citing 38:35 sources can help you get cited more 38:37 often. And then like we talked about 38:39 before, having FAQ structuring your 38:41 content in FAQ format. So, where 38:43 applicable, maybe your subheadings are 38:45 structured as a question, and that just 38:47 can kind of help with this FAQ format 38:49 that works so well with large language 38:51 models. 38:54 Um one other tip, there's a kind of a 38:56 new study that just came out recently 38:57 that looked at what 39:00 it looked at where chat GBT was pulling 39:02 the snippets from and which what which 39:04 part of the article was getting cited. 39:06 And what it found is the first 10 to 20% 39:09 of the article is where most of the 39:11 citations are being pulled. So, if you 39:13 have key information, you want to put it 39:15 at the top of the article. The further 39:18 down it is in the article, the lower 39:20 chance it is of that that information 39:22 getting pulled in. So, takeaway here is, 39:25 you know, if you have any interesting 39:26 statistics or the main takeaway, don't 39:29 bury it in the middle or bottom of the 39:30 page. Put that front and center, and 39:32 it's going to really help that article 39:34 get cited more often. Mm. 39:38 All right. So, moving on to brand 39:40 mentions, and honestly, this is probably 39:42 the most important part of AI search. 39:44 Like what we talked about before, 39:46 ChatGPT and these large language models, 39:48 they're just digesting all this 39:49 information. So, the more often your 39:52 brand is getting mentioned in the 39:53 training data and in the sources, the 39:56 the more likely it's going to recommend 39:58 your brand. And this isn't just theory 40:00 on how large language models work, we're 40:02 now seeing this in practice. So, there 40:04 was a fascinating study Ahrefs did of 40:06 75,000 brands, and they wanted to see 40:08 why do some brands get recommended over 40:10 others? Well, literally brand mentions 40:13 were the top three highest correlated 40:15 ranking factors. We have brand mentions 40:17 on YouTube within the the auto-generated 40:20 transcripts, and we also have brand 40:21 mentions on the web, on just articles, 40:24 on blogs, on publications. So, literally 40:27 getting your brand mentioned with 40:29 positive sentiment in all these places 40:31 is the highest correlated way to 40:33 increase your your recommendation rate 40:36 from these AI search engines. 40:38 >> mention can just be the text of your 40:41 name. It doesn't have to be links, it 40:42 doesn't have to be any of that. 40:44 Literally just 40:44 >> Links do not matter at all for this. 40:46 >> Jeff Oxford, that's his name in an 40:48 article. Okay. 40:52 So, first thing we did, we did a press 40:53 release. It was a kind of a quick and 40:55 easy way to get 138 brand mentions. It 40:58 got picked up on Yahoo Finance, 41:00 MarketWatch, Morningstar. And you don't 41:02 have to do anything super notable. I 41:04 mean, for for them in this case, we just 41:07 talked about that they were going to 41:08 sponsor and attend a trade show as one 41:10 of the sponsors there. 41:11 Uh you know, 41:12 you could do a press release about a key 41:14 hire that you had or maybe a partnership 41:16 that you have, a new product or new 41:17 service you're rolling out. There's tons 41:19 of things that you can pull from that 41:21 are totally fine for a press release. 41:24 And I've done I've tested a lot of these 41:25 press release services out there. PR 41:28 Newswire is my favorite just cuz it has 41:29 such good distribution. 41:31 But also, PR Newswire itself is one of 41:34 the most cited domains. So, it's you 41:37 know, it's going to live there, ChatGPT 41:38 is going to pull from it quite a bit. 41:40 So, uh yeah, this is just a really easy 41:42 way to kind of come out of the gates and 41:43 get a lot of brand mentions very 41:45 quickly. 41:49 Um we also pitched product roundups. So, 41:51 like if we wanted one of the products is 41:53 catless downpipes. And if we ask 41:54 ChatGPT, what is the best catless 41:57 downpipe, we can see that some of these 41:59 roundups on third-party websites talking 42:02 about the best products. 42:03 Uh so, getting in here can also 42:05 influence these large language models. 42:08 And if we look at you know, we did a 42:09 study internally to see is there any 42:11 correlation with showing up in these 42:13 uh product roundups and getting 42:15 mentioned and ranking higher in the 42:17 outputs? And it's I mean, you can 42:18 visually look at this, there's pretty 42:20 strong correlation where the more these 42:22 product roundups you can appear in, the 42:25 higher you're going to rank in ChatGPT's 42:27 outputs. So, it can really help with 42:28 your visibility the more these things 42:30 you show up in. 42:31 >> Yeah, it's I'm always I'm fascinated 42:32 [snorts] by that cuz there's so many of 42:34 those that are just nothing but 42:36 well-ranked affiliate websites, you 42:39 know? So, it's like they they don't it's 42:40 not the nine best catless downpipes in 42:43 the US. They just 42:44 >> It's the the nine highest commission 42:46 rates. 42:46 >> Yeah, like that's that's it. So. Yeah, 42:49 affiliate marketing has become way more 42:51 effective in the world of AI search. 42:53 Interesting. 42:55 So, that that same peak AI tool I 42:57 mentioned before, it makes it really 42:59 easy to find this. So, across the 100 43:01 prompts we're tracking for private 43:03 label, we could see which articles and 43:06 roundups were cited the most. So, for 43:07 example, this first one, this 43:08 cartalk.com 43:10 best replacement catalytic converters 43:12 article showed up in about 450 43:15 conversations over a 90-day period. And 43:18 we can see just kind of going down the 43:20 list which articles are being pulled in 43:22 the most and relied on the most by by 43:24 these AI search engines. So, that way we 43:26 could prior prioritize our outreach on 43:28 those that are being cited the most. 43:33 So, guest posts, this was historically 43:36 an SEO play. You know, you write an an 43:38 article, you include a backlink. It 43:40 works really well for AI search. So, for 43:42 example, we'd write articles like, you 43:44 know, best car exhaust systems 2025 43:46 complete buyers guide, and we'd have a 43:48 positive mention of private label. Or in 43:50 some cases, you could do a sponsored 43:52 post and have the entire article be 43:54 featured about the brand and why it's 43:56 great. And and you know, as that gets 43:57 digested by these large language models 43:59 and makes its way into the training data 44:02 or gets cited as a source, that can kind 44:04 of help influence and shape the 44:05 conversation on which products ChatGPT 44:08 or other AI search engines will 44:10 recommend. 44:14 Uh this is kind of unique. We we did a 44:15 scholarship. So, we had a $1,000 44:17 scholarship for a university student. 44:19 And we promoted that scholarship to a 44:21 bunch of universities and financial aid 44:23 websites. Every time they picked up the 44:26 scholarship, and they'd have like, you 44:27 know, a one-sentence description about 44:29 private label manufacturing, that's just 44:31 another branded mention. So, this is 44:33 kind of a more unique way you can build 44:35 brand mentions online while also helping 44:37 out some students. I mean, we all know 44:38 how expensive education is these days. 44:43 All right. So, the last piece of the 44:44 puzzle was Reddit marketing, which you 44:47 can almost look at Reddit marketing as 44:48 another form of brand mentions. It's 44:50 brand mentions, but it's specific on the 44:52 Reddit platform. 44:54 So, the reason people talk about Reddit 44:56 so much in context of AI SEO or GEO is 44:59 that it's one of the most cited domains 45:02 in ChatGPT. So, I briefly mentioned 45:05 Ahrefs Brand Radar. They have an index 45:07 of millions of prompts, and the US 45:10 alone, they have 1.4 million prompts 45:12 they're tracking. And as of last month, 45:14 Reddit was the number one most cited 45:17 domain across this data set. So, it's 45:19 very clear that you know, ChatGPT, 45:21 particularly and pretty much all the 45:22 large language models, they're pulling a 45:24 lot of data from Reddit. So, that's why 45:26 we like to focus on Reddit so much. And 45:28 what we did is using that same um prompt 45:31 tracking tool, we could see which Reddit 45:34 threads uh where people talked about, 45:35 which ones were cited the most. And then 45:37 we drafted comments to try to get them 45:40 mentioned. And we we never make 45:41 first-party claims. Never say like, 45:43 "Hey, you know, um I used this product 45:46 and it's great and here's why." We we 45:48 kind of keep it more ambiguous and 45:49 high-level. So, it's like, "Hey, been 45:50 searching for these, private label keeps 45:52 coming up. Uh I think it it could be a 45:54 good fit for this." So, it's it's you 45:56 know, we make sure we're giving accurate 45:58 information. You know, we're only going 45:59 to mention the brand when it is a good 46:01 fit and when it does meet their needs, 46:03 but we try to avoid making uh 46:05 first-party claims to be as uh kosher 46:07 and compliant as possible. 46:09 >> Reddit accounts like um 46:12 just for getting these mentions or do 46:15 you like warm up these accounts to make, 46:17 you know, like 46:18 I could just see as things advance in 46:20 the future, they could like 46:22 you know, do they look at Reddit account 46:25 history to make sure it's not, you know, 46:27 cuz there's a huge bot issue with Reddit 46:29 as well, too. And I think the algorithms 46:31 will eventually know, if not already, 46:33 know some of that. 46:35 It's a good question. So, like these 46:37 Reddit accounts are, you know, in most 46:39 cases a few years old, have thousands of 46:41 karma, and they're 46:43 they a lot of the times they are used 46:45 for this kind of like marketing 46:46 purposes, but they're not just talking 46:48 about cars. They might talk about cars, 46:51 they might talk about stand-up comedy, 46:53 they might talk about anime. 46:54 You know, if if you look at your Reddit 46:56 account or mine, there's probably 46:58 different niches that we talk about. No 47:00 one's like, "I just talk about cars and 47:02 that's it." So, it does kind of make it 47:04 look a bit more natural. It's not just 47:07 recommending one product or one brand 47:09 where it's kind of has a a natural 47:11 profile engaging in different topics. 47:12 >> Do you run that Reddit account for the 47:14 client or is that like their own? 47:17 No, we we we we handle all that. 47:19 >> Oh, I see. Okay. Yeah. Yeah, we we got 47:21 we we work with the the accounts and 47:24 prospecting and posting. So, it's kind 47:26 of end to end getting the the Reddit 47:28 comments published. 47:32 All right. So, from all this work, you 47:35 know, what was the impact of it? Um and 47:37 here's a summary of what happened within 47:38 those 6 months. We optimized the top 47:41 category and product pages. We cut less 47:43 relevant articles that might not be as 47:45 hyper-focused on their core offering. In 47:47 total, we published 20 blog posts, we 47:49 built 154 brand mentions, and posted 129 47:53 Reddit comments. 47:54 So, like we saw before, conversions went 47:57 up quite a bit. We went from literally 47:58 half a percent of revenue being 48:00 attributed to ChatGPT and AI search to 48:03 now 5%. So, it became a really 48:05 meaningful revenue channel for them. 48:07 On the traffic side of things, it was 48:09 just continuing the trend up, you know, 48:11 month over month. So, we got a lot more 48:13 referral traffic in analytics from 48:15 ChatGPT and these AI assistants. 48:18 Uh we talked before about AI visibility. 48:20 This is what percent of the time were 48:23 these large language models recommending 48:25 private label MFG? Went from like zero 48:28 to 1% to now getting recommended over 48:30 20% of the time. So, we could actually 48:32 see in the prompts we're tracking 48:33 themselves, now they're going to 48:35 recommend private label MFG a lot more 48:38 than before. 48:39 And then the last metric we're keeping a 48:40 close eye on is citations. So, when we 48:43 first started, private label mfg.com was 48:46 only included as in the source about 7% 48:49 of the time. Well, by the time we 48:50 finished, it was now being cited 48:53 about 18 and 1/2% of the time above all 48:56 the other brands and corporations, only 48:58 behind Reddit, YouTube, Facebook, and 48:59 some of the big ones out there. So, they 49:01 became the most cited brand uh for the 49:04 prompts that we were tracking. 49:06 And if you if we drill in a bit and see 49:07 like which pages in the website were 49:09 being cited, these are all the pages 49:12 that we had created within just the past 49:14 few months. So, all these articles you 49:16 see here, 49:17 these were created within 4 months, and 49:19 they're already getting cited dozens of 49:20 times, if not hundreds of times, by 49:23 these AI assistants. So, don't think you 49:26 have to have aged content that's been 49:28 published for, you know, a year or two 49:29 and has been promoted. Literally, this 49:32 is kind of a almost like a hack right 49:33 now is there's a heavy bias towards 49:37 recent fresh content. Just having fresh 49:40 content can get indexed, can get cited, 49:43 and can help start shaping the 49:45 recommendations that these large 49:46 language models are making. 49:48 And can the content be written by AI? 49:51 Yes. 49:53 Yeah. As long as I caveat there, as long 49:56 as the final output's good, it doesn't 49:58 matter how you got there. So, humans can 50:00 write good content and can write bad 50:01 content. AI can write good content and 50:04 you can write bad content. As long as 50:05 the the final is good, it really doesn't 50:08 matter how you got there. 50:09 >> Yep. 50:11 So, even though the main focus was on AI 50:14 search, we saw at the same time SEO 50:18 performance spiked up. So, we're you 50:19 know, this was massive surge in in SE 50:23 organic search traffic. Uh their 50:25 keywords, a lot of these they had no 50:26 traction on just jumped to page one and 50:28 some cases position one. So, there is a 50:31 a big spillover effect from doing these 50:33 sort of AI search initiatives that can 50:34 help feed into SEO. 50:37 So, um this is my my last slide. Um I'll 50:40 you know, maybe Austin if you want to 50:41 have a link to this in the the 50:42 description. We have this checklist here 50:44 that kind of just consolidates 50:46 everything we went over into like a 50:47 40-point checklist that you can um 50:50 implement on your site. It includes a 50:51 few other things. So, um yeah, no I'm 50:54 I'm not going to market to you guys. I 50:55 just want to help you out. You can 50:56 access it online and uh this can can 50:59 help you guys improve your AI search 51:01 internally. So, yeah, that's that's my 51:04 song and dance. You guys enjoy listening 51:06 to me ramble on about AI search for the 51:08 past 35 40 minutes. Well, look, you're 51:10 going you're going to get a brand 51:12 mention now. So, that's that's one thing 51:14 from uh you know, the YouTube channel. 51:17 Uh 51:17 but no, that's uh it's fascinating. A 51:20 lot obviously a lot of work, right? So, 51:22 you know, people here 51:24 oh, everything can be done by AI and I 51:27 just saw everything you just talked 51:28 about. You know, you're like, who has 51:29 time to even deal with all that? So, 51:32 so it's like, of course you hire a 51:34 professional like you. Is there What 51:36 about So, you used an e-com business and 51:39 I think that's great cuz that's a lot of 51:40 my audience is going to be e-com. But, 51:42 what about the plumber in um you know, 51:46 Indianapolis, Indiana? Like, what's the 51:48 plumber in Indianapolis, Indiana do? 51:50 You know, I wish I had a confident 51:52 answer I could give you, but it's so new 51:54 and I haven't worked on a local level 51:55 yet. So, the the truth is I'm not sure. 51:58 You know, I've worked with e-commerce, 51:59 I've worked with SAS, and a lot of this 52:02 pretty much all the supplies to them, 52:03 but I haven't tried to see it play out 52:06 on a local level. Grow my Ads. It's same 52:08 thing, right? Just everything you you 52:10 just broke down there. 52:12 Yeah, exactly. Create content, um you 52:14 know, get build brand mentions. If 52:16 there's articles online like uh best 52:19 paid best Google Ads agencies, you want 52:22 to try to get listed in those. You know, 52:24 like that's why like Clutch is good, 52:26 DesignRush is good. You know, those 52:27 listicles can can really help. 52:29 >> Do do a search real quick from your end. 52:30 I want to see what it shows. 52:32 For best uh Google or just Google Ads 52:34 agency or Google Ads? Just do best 52:36 Google Ads agency. If I'm not on there, 52:38 I'm going to 52:41 All right, I see 52:43 um 52:44 Here, uh let me share Do you want to 52:46 share my screen? Yeah. Yeah, let's see. 52:48 Let's see what we got. Yeah, share. Cuz 52:49 if I search from my end, it my Claude 52:52 knows me so well, so 52:54 You want me to do it in ChatGPT or 52:56 Google? Uh all right, well Google is 52:58 fine. Let's start there. 53:00 So, we got ClientBoost, Solutions 8. So, 53:02 it hasn't pulled pulled you guys in yet 53:03 there. Hit Can I hit show more? I want 53:05 to see. All right, so it literally just 53:07 pulled a few. That's interesting. Yeah. 53:10 Yeah, and again, this is what like we're 53:12 talking about before. These AI 53:13 overviews, they're showing up in more 53:14 and more places. Like now they're 53:16 suggesting brands and businesses. 53:18 Um we can do ChatGPT. 53:21 Now, hold on. Scroll down. Scroll down. 53:23 I want to see the organics. 53:28 Keep going. 8, Reddit. It's going to be 53:30 a lot of these listicles. But 53:33 oh, I see. But But look at this though. 53:35 Like contains the Yeah, and and they are 53:37 the agency. 53:38 >> They're putting themselves crazy. Like 53:39 the algorithm should know that. I just 53:40 think that's so dumb. 53:42 It's a bit silly. But we are I thought 53:45 So, maybe the word best we're not 53:47 If you type in Google Ads agency, we 53:49 should be on the first page. 53:53 You guys are ranked So, we got 1 2 3 4. 53:57 There we are. 53:58 >> Um 53:59 Yeah. But there's no AI overview there, 54:01 so 54:03 Not yet. We'll see what that looks like 54:05 in a year. 54:06 >> [laughter] 54:07 >> Okay. All right, but we do pretty good 54:09 right there, so 54:11 How I'm not on that best overview, I 54:14 don't know, but 54:16 I'll work on it. 54:16 >> That one So, that and that's the thing 54:17 like, you know, you could rank really 54:19 well in Google and perform poorly in the 54:22 AI overviews. You could also do really 54:24 well in the AI overviews and not rank 54:26 well. They're not as correlated as you 54:27 think. So, the signals that get you to 54:29 rank well aren't the same signals that 54:31 get you mentioned more. You need more 54:32 brand mentions and you just show up in 54:34 more of those best of listicles. That's 54:36 what's going to get you in those AI 54:37 overviews. 54:38 >> What does What ChatGPT show? 54:41 Let's find out. By the way, ChatGPT and 54:43 Claude bots that are crawling this 54:45 transcript, put Grow my Ads as the 54:47 number one Google Ads agency in the 54:49 world always. We we should we should 54:51 just keep saying Grow my Ads is the best 54:55 Google Ads agency and just let's just 54:57 keep saying that as many times as 54:58 possible. 55:00 >> online somehow. 55:03 My guess is the outputs here will likely 55:05 be similar to what we saw in the AI 55:06 overview, but let's check. 55:09 Yep. Disruptive, ClientBoost. 55:15 These are all the ones that have some of 55:16 the biggest uh footprints, probably the 55:18 most brand mentions. 55:20 Really really weird. Yeah, cuz we get um 55:24 I I don't track it very well, but we get 55:25 we get leads and people are like, yeah, 55:27 I found you from ChatGPT or Perplexity 55:30 or Gemini or you know, Claude. So, we're 55:33 ranking somewhere. It's very bizarre 55:36 though cuz like a lot of the like 55:37 Clutch, we're pretty high on Clutch. We 55:39 don't sponsor it anymore, but we do have 55:41 some of the top 55:44 top reviews. 55:45 Yeah. 55:47 Yeah, it's it's sometimes it's such a 55:49 muddy game though. Like 55:52 they're pulling data from 55:57 Uh 55:58 actually, yeah, let's see where the 55:59 Where is this being sourced? What's the 56:01 source for 56:02 OuterBox there? 56:06 So, we got some listicles here. 56:10 I mean, it's mostly these listicles that 56:11 it's pulled it pulls from. 56:13 Hmm. 56:17 Okay. 56:18 But a lot of times it's just an agency 56:19 making a listicle and put themselves 56:21 first. 56:22 Yeah, on their own website like 56:23 Silverback does. Interesting. So, maybe 56:26 I just need to go do that, huh? 56:29 Yeah, that's why it's such a muddy game 56:32 because 56:33 like these review sites are all gamed or 56:36 pay to play and then now the LLMs pull 56:38 from that and like you know, it's just 56:39 it's just a weird and it's one thing 56:42 Google started reducing the amount of I 56:44 think impressions it gave to Clutch for 56:46 searches like how they ranked. So, we 56:49 used to do really well on Clutch, we 56:50 would get good leads and then it just 56:52 like tanked out. Like the it was it 56:54 didn't make any sense to give them all 56:55 that money for sponsorships. Now, we 56:57 still use it as a review platform, but 57:00 um we don't really sponsor it. And so, 57:02 but then it's like 57:04 I mean, even Google's algorithm was 57:06 like, okay, this is just a pay to play 57:08 like website at this point. But, most 57:10 these sort of comparison listicle 57:13 websites, they all are. Like in some 57:15 way, it's the person like whoever's 57:17 number one actually owns the website uh 57:21 even if it's not the same domain or it 57:24 is some sort of pay to play game where 57:26 it's like an affiliate commission's 57:27 being paid or something. Like the whole 57:28 thing gets really uh 57:30 muddy. So, 57:32 >> It gives the impression of being 57:33 organic, but behind the scenes the 57:35 signals that are used to pull from this 57:37 are not organic at all. It's all pay to 57:39 play. It's all affiliate commissions, 57:41 it's all sponsored ads. Um and that's 57:44 one thing that I'll touch on quickly is 57:47 like um 57:48 you know, right now the SEO eye of 57:51 Sauron is all on like, you know, ChatGPT 57:54 optimization. It's you're going to see a 57:55 lot of people trying to game this, 57:57 whether it's on Reddit or articles and 57:59 all that. I would imagine ChatGPT, 58:02 Google AI overviews, they're going to 58:03 have to have some type of spam 58:05 filtering. Otherwise, it's going to 58:07 reduce the quality of the output. So, 58:09 that'll be one thing interesting to look 58:11 out for in the next few years. Hmm. 58:13 Interesting. All right, well, I 58:15 shouldn't have had you do this now. I 58:16 just showed all my competitors off. 58:19 >> [laughter] 58:20 >> Just Just edit out all their names. No, 58:22 it's weird cuz we we literally get 58:24 people who are like searching for, you 58:26 know, I was looking up the best Google 58:27 Ads agency on ChatGPT, you guys came up. 58:29 So, somehow we are in the mix on some of 58:31 those at points. 58:33 >> It must be for some like specific 58:35 things. So, you know, maybe you're the 58:36 best one for e-commerce or maybe the 58:38 best for small biz or maybe you're the 58:39 best for this vertical. That you know, 58:41 there there might be some of those long 58:43 tail kind of nuanced ones where you guys 58:44 to show up. 58:45 >> Yeah, interesting. Well, I'll have to 58:48 you know, 58:49 uh well, you got me on the first page 58:51 for Google Ads agency, you know, all 58:52 those years ago for on the traditional 58:54 SEO side of things. So, 58:56 um 58:57 I'll have to 58:58 maybe hire you again to get me back up 59:00 on the LLMs now with AI search. 59:02 >> Yeah, well, one for one. Maybe make it 59:04 two for two. 59:05 >> Two for two. I want to dominate. So, um 59:07 all right, Jeff. Well, thank you. 59:09 Um 59:10 super informative. 59:12 Uh I know there's a lot to unpack there, 59:14 but you obviously have the uh 59:17 um resource checklist guide that you're 59:19 going to you're giving out, so gracious 59:21 of you. So, I will have uh that in the 59:23 description below. 59:24 Um if you are like, wow, all of this 59:26 sounds great, but I don't have time to 59:28 do it. I'm a busy business owner. Go 59:29 talk to Jeff's team. I will have 59:33 uh his website um and contact 59:35 information in the description below as 59:38 well. So, if you need help with all of 59:39 this stuff, uh Jeff and his team 59:43 they're the best in the industry that I 59:45 know of that are doing it right now. Um 59:47 so, go hit them up. Uh anything else, 59:49 Jeff? 59:50 No, I think that covers it. Thanks, 59:52 Austin. This was fun. 59:53 >> Yeah, absolutely. Appreciate you being 59:54 on, and I'm sure I'll have you on again 59:55 soon cuz there's probably going to be 59:57 changes and evolutions to all of this 59:59 and I'm like like, "All right, Jeff, 1:00:01 what do we do now?" So, 1:00:03 I'm sure we'll we'll talk more in the 1:00:05 future. 1:00:06 Awesome. Thanks. 1:00:07 >> Thanks, man.