.jpg)
The Art of Selling Online Courses
The Art of Selling Online Courses is all about online courses.
The goal of this podcast is to share winning strategies and secret hacks from top performers in the online course industry. We are interviewing successful business owners, asking them questions on how they got to the point where they are right now, and checking how their ideas can help you improve your online course!
The Art of Selling Online Courses
216 How I Hit 7 Figures in 7 Months (Caleb Ulku interview)
🔥 Grow your course revenue up to 30% in 7 days - no paid ads, no sales calls 📈 https://datadrivenmarketing.co/roadmap
Caleb Ulku went from making $5K a month from his course community to hitting a 7-figure run rate in just 7 months. And he did it with only 27,000 YouTube subscribers.
What makes this even more interesting is how unconventional his approach is. He doesn't have a traditional email list funnel. Instead, he sends people directly from YouTube to a $27/month School community. No free lead magnet, no nurture sequence. Just YouTube doing the heavy lifting.
In this episode, Caleb shares how ChatGPT 3.5 completely changed his business direction overnight. He saw what was coming with AI overviews and made the call to pivot his entire SEO agency from informational queries to local SEO within a day. That decisiveness is a theme throughout this conversation.
We dig into his YouTube content strategy, how he uses Claude to write scripts (and why the AI keeps making the same frustrating mistake), and his two-tier School community model where 20% of people who join at $27 end up upgrading to his $197 pro membership.
One thing Caleb said that really stuck with me: "A subject matter expert can replace a team of interns with AI, but an intern cannot replace a subject matter expert." If you're using AI in your business, that's worth thinking about.
🌐 Check out Caleb's YouTube channel: https://www.youtube.com/ @calebulku
🌐 Check out Caleb's website: https://calebulku.com
The growth from there to October happened in seven months. We just hit a seven figure run rate. We were bringing on board like one client every day. Did that, and within a couple of months after putting a$27 price tag on it, the school group was making five or six K a month?
SPEAKER_02:The way you told the story, it sounds like you went on touching the T3 and a half. You came to this decision by the afternoon you switched to Low Flexia. Was it actually as fast as that? Or was that like if it was as fast as that?
SPEAKER_00:I had this conversation with quite a few people. They almost don't believe the course figures because we only had 27,000 subscribers or so. So it's still a fairly small channel.
SPEAKER_03:Hello, and welcome to our online course. We're here to get a winning strategy. My name's John Anto, and today's guest is Caleb Fox. He has earned the full cost of Cooper Union competitive engineering program. Spent over a decade and made the world completely. Climbed the ladder, earned an MBA, got married, had two kids, and he found he's missing out on family life. Then a hip injury in 2015 changed everything, he was forced to work remotely, and he realized how much better life could be at home. So this is how he started his entrepreneurial leap, starting an SEO business on Upwork. Now he has moved over into the space of selling courses. And we're going to be talking to him today about his course business and how he's grown it using YouTube and what's working for him using AI.
SPEAKER_02:Caleb, welcome to the show.
SPEAKER_00:Man, appreciate your invitation to be here, John. It's uh it's an honor and thank you for the introduction. Yeah. So talk us through it.
SPEAKER_03:So, what what are you selling people courses about? Who are you helping? What are you helping them with?
SPEAKER_00:Yeah, so we have uh a full audience avatar so that I know who we're making YouTube videos speaking to, right? Who we expect to be watching the content. So for the most part, it is uh men uh from 25 to 45. Uh they have a full-time job, but they're looking for something else. They're looking to start their own business. And in general, they're actually looking to start their own business in the space of SEO. So my avatar is not like a generalized make money online uh personality, it is very specifically someone who is interested in an SEO agency.
SPEAKER_03:Got it. So they're already at that stage, they've already decided they want to run an SEO agency, but they're still in the job at the job at the moment. They haven't taken the first step yet. Exactly. Got it. Okay, cool. And so what are you helping them with? Are you helping them to learn how to do SEO? Are you helping them to learn how to run an agency? Is it both of those bits? Like, what do you help them with?
SPEAKER_00:Yeah, it's both of those. So the we talk about every video that we produce and put onto YouTube has to talk through one of we our avatar's name is Young Timmy. So he has to work on one. Well, you know, young Timmy can be 45, that's fine. Um, he has a little crotchety broken leg. Anyway, so young Timmy has two key problems that he really, really cares about. One of them, he doesn't know how to actually do SEO, okay? And the second problem he has is he doesn't know how to run an agency. So every video we produce talks about one of those two aspects. Uh the goal, of course, I'm sure you're familiar with some of the fundamental strategies of YouTube, but it's to create enough content around those two problems to create sessions where someone watches a video and then the next video is very interesting. It solves a similar problem, same topic. So they watch the next one and the next one and the next one. And that really helps to build trust with that person, with that viewer. Nice. Cool.
SPEAKER_03:So then what's been your strat? What kind of size is the uh the business at the moment? Share anything you're happy with, revenue or number of students, or whatever you're happy to share?
SPEAKER_00:Sure. So the course side of it, uh, we just hit a seven-figure run rate in October. So that was very exciting. Uh, that's just the course side, doesn't include the SEO agency that I'm still running. So um, yeah, the growth of that has been really uh very exciting to be a part of and to watch. Um my YouTube channel itself, I I I have this conversation with uh quite a few people. They almost don't believe the course figures because my YouTube channel, I'm only, and I say only with a grain of salt, I'm quite happy with where it's at, but I'm only at 27,000 subscribers or so. Uh so it's still a fairly small channel. Most of um, most of the time I get three, four, five thousand views a day, something on that range. Oh, that's pretty good.
SPEAKER_03:27,000 subscribers, because the more important thing than the subscriber count is the view count, right? But just everybody can see your subscriber count because it says it at the top of the colour. Because it says it, yeah. So 3,000, so like a what's that, like a hundred to 150,000 views a month within the math?
SPEAKER_00:They're about, yeah.
SPEAKER_03:Last month was 138,000, so spot on. Nice. Okay. And then how are you making sales from that? Are you sending people straight to to the you because you've got actually just take us through what your offerings? How much do they charge?
SPEAKER_00:Yeah. So the videos are all about essentially how to use AI to rank local businesses better, faster, run your agency better, faster. Um, so it's all very much an AI-focused uh take on running an SEO agency. Uh, some of that is because, well, November of 2022, Chat GPT three and a half comes out. Before that date, my SEO agency was largely focused on non-local clients trying to rank for informational queries. You mentioned I had built my SEO agency initially on Upwork, and that is the client base on Upwork, large businesses looking for SEO services, right? So we've worked with Skillshare, Adobe, Stanley Blackendecker, all from Upwork. Um and, you know, November 2022, I start playing with ChatGPT three and a half, and I I always think back to the Google search mortgage calculator. Right. So this is like pre-2022. This is like, you know, 2017, 2018, sometime around then. And all of these companies are spending a ton of money trying to rank for mortgage calculator because somebody who needs a mortgage calculator might be interested in getting a mortgage. Then Google, uh wanting to keep people on their search engine results page for longer and longer, send people to their own properties, they roll out a mortgage calculator built into the search engine results page, right? Great for Google, great for its users, absolutely terrible for the businesses who had spent five, six figures trying to rank for mortgage calculator, their traffic went away overnight.
SPEAKER_03:Especially the guy who was so proud because he just got them to number one like the day before or something like that. Can you imagine being that guy? Like right? How awful would that be?
unknown:Yeah.
SPEAKER_03:You're like, you've hired this agency, you've done your link building, you're moving up the rankings, you get to number one, you're cheering, you're whooping, you're like, oh no.
SPEAKER_00:And then you wake up and your traffic is like zero the next day, and you're like, oh my gosh, what happened? You check your rank position, you're still in the top three, but where why why don't I have any traffic anymore? Yeah. Seeing what ChatGPT three and a half was capable of, uh, I called uh my agency, some of the key employees together, and was just like, yeah, obviously Google is going to do something with this, and these informational non-local queries, I don't think they have like a five, 10-year life left in them. I think we should pivot to focus on local SEO, right? No matter how good ChatGPT 3.5 was, it felt like we were at least 10 years away from it being able to, you know, replace a water heater or something like that.
SPEAKER_03:Um Yeah, I'm thinking about the way I use ChatGPT versus Google. Like I almost never do a search on Google anymore for information. It's all through ChatGPT now. But if I'm looking at the case, I go to Google. Sorry, go ahead. But yeah, if I'm doing something local, I'll go on Google Maps and I'll search for something there. And so it's all still through Google. It's all still like if someone's got themselves to appear, you know, like a the restaurant that comes up as being the best reviewed one or something like that. I don't know how typical of a user I am, but from my perspective, that totally fits what you're saying.
SPEAKER_00:Yeah, you're yeah, absolutely. And beyond that, even if you did do an informational query on Google, you would get now the AI overview. Uh obviously I didn't necessarily see that exactly.
SPEAKER_03:Do you think they're any good? A bit of a tangent, I know, but do you think like compared to a chat because they they come up so fast, are they actually as good as Gemini or ChatGPT or anything like that?
SPEAKER_00:So it's the the they're basically running on almost like a Gemini Lite, if that makes sense. So they're not terrible, but one of the things that makes an LLM, like the current AI models, one of the things that makes them better is basically giving them more time to process, more time to think. So that is one of the challenges of the AI overviews, absolutely, is I don't have that time because no one would wait 30 seconds for Google search results. Um so they only have a second to do the AI overview. But there is clear data showing click-through rates are pretty much massively down when the AI overview shows up. Like uh I was just looking at a study the other day, and click-through rates were on the order of half once the AI overview shows up. Um Wow, that's brutal. It's very brutal. So, and I I've had conversations with multiple people who have talked about, hey, wanting to do things in SEO, uh getting started in SEO, and it's just like, yeah, that's great. Just don't do non-local informational queries. Like you you're not gonna get traffic if you're trying to rank for those queries. It needs to be either local or non-local transactional, non-local, very detailed type of information, right? So uh not like what is a water heater, because no one is going to click through on that. But if you're looking for like a detailed how to replace a water heater, that is still likely going to get a click because the AI overview is not going to answer that. And most likely that click is going to send somebody over to YouTube. Mm-hmm.
SPEAKER_03:Yeah. Got it.
SPEAKER_00:So again, depending on the audience, like what you're trying to get put in front of people, depending on what the search is, SEO still can have a good fit, even for non-local, but it's not going to be general, broad informational queries. That that ship has uh largely sailed at this point. Got it.
SPEAKER_03:Okay. I want to dig into something that you said there. You said you got your key employees together, you said we're gonna switch over to local SEO. Was that like the way you told the story sounds like you went on ChatGPT three and a half, you came to this decision, by the afternoon you'd switch to local SEO. Was it actually as fast as that? Or was there like a No, it was as it was as fast as that.
SPEAKER_00:So everything that we had built in the agency, everything was based on Upwork. Um Upwork is not a great place to find local SEO clients, right? Joe the plumber isn't posting SEO jobs on Upwork. Um so we had a conversation how to get clients, what to do. I know a lot of people in the space. So what I'd heard a lot of people having success with was meta ads, based Facebook ads. So by the next day, we had a Facebook ad campaign running. I mean, it wasn't anything great or fancy. We spent an hour on it, uh, wrote some copy with ChatGPT three and a half, uh, got some images thrown together with Dolly. Um, and we're running ads targeting plumbers. And what I really love, we did this on purpose when we put the ad together. Um, we're targeting water heater replacements, uh, because that's what plumbers love, right? They make a thousand dollars with a few hours of work to replace a water heater, but we called them hot water heater replacements because plumbers hate that. Um, they're obviously not hot water heaters, they're water heaters. If the water is already hot, it wouldn't need a heater, right? Yeah. Um, so it's just a way to get under their skin because we're trying to create engagement, trying to get them to comment on the posts. And man, our ads had hundreds of comments. So it was like so many plumbers were upset about it. And we had an AI generated image of some plumber replacing a water heater, but it didn't make any sense. Like the wrench was in the wrong position. I mean, this was back in the early days of Dolly when the AI image generation wasn't very good. So it just looked ridiculous, but it was early enough that you know, now if you look at a ridiculous image, you'd say, Oh, well, that's AI generated. I mean, this was early 2023. So that wasn't like the first thought, especially for the target audience we were running ads to. So we were getting a bunch of comments like, what the heck is that plumber doing with that wrench? It's in the wrong position, it's you know, it's not galvanized or or what so we had all of this engagement, and uh the ad performance was, you know, I we spent so much effort split testing this ad, and nothing that we did would outperform this sad, stupid ad that we threw together in an hour. Um yeah, we started running that. And to be honest, John, uh at one one day, like a couple a week later or something, uh the my phone sales guy he he calls me in the morning. He's like, Hey, Caleb, I have the first call this afternoon with a plumber from the ad that we ran. I was like, that's awesome. I'm glad to hear it. And he was like, Yeah, what do you want me to talk to him about? We didn't have an offer, we didn't have a price point. We because the first step is always, you know, can I get a local business owner on the phone? And if I can, then we'll figure out the next step. And if that works out, we'll figure out the next step. But I hate, you know, back in the early days, I've been in the online space for about 10 years. Back in the early days, we used to do a whole bunch of upfront planning, but then if step one doesn't work, you have to throw everything else away. So now it's let's get step one, let's see if it works. And if it works, we'll figure out step two. So step one was can we get people on the phone?
SPEAKER_03:And we had a similar planning meeting. Uh me and uh Yosip. Um, I'm just trying to figure out is it me and Yosip or Yosip and I? Yosip and I. Yosep and I had a planning meeting early this uh this week, and we went through and we figured out like this is the thing we need to do for the next step. And I was like, okay, and then what happens when that works? What happens? And he was just like, to know, future us will figure it out. That's like that's that's a problem for two weeks from now, us. Two weeks from now, us is they're really clever, those guys. They are gonna figure out, but and it if it if it doesn't work, it doesn't matter. They don't have to solve that problem, do they? So it doesn't it doesn't matter at all.
SPEAKER_00:It's like it's no big deal. Exactly. There are few ways, few better ways to waste massive amounts of time than to plan everything out ten steps in advance, and then the first step doesn't work. Yeah. And the the first step often doesn't work, in my experience. It often doesn't. Yeah. We got lucky with this one because it did, but yeah, often the first step doesn't work. Okay. Um let's take a week.
SPEAKER_03:You've you've got the you've got this process now working, you've you've changed it to local SEO, um, you're getting people coming in through that. What point did you start selling courses? When did the YouTube channel start? Because that's like this was even before YouTube. Okay, yeah. This is back ChatGPT three and a half. This is what is 2023?
SPEAKER_00:November of 22.
SPEAKER_03:Like November 30th, I think. Okay. Got it. Three years ago.
SPEAKER_00:It's weird that I know that date off the top of my head, but yeah, I think it was November 30th of 22. Yeah. Um could be lying. There's no way for me to check. Okay. Yeah. Well, so uh we started landing uh SEO clients. It took us a few months um to figure out what offer was working, uh would work. Uh, but after boy, it was probably 40 or 50 sales calls. Um, where you know, every 10, 15 calls, we we would change the offer, change the offer, change the offer. Uh, we finally got one that worked, and in the next um like couple of months, we were bringing on board like one client every day. Wow. He was averaging a close every day. So we put together this system of how to actually manage all of these clients. Uh, we were using ChatGPT three and a half to write content. We were using a website builder. We're basically leveraging as much AI as we possibly could to get these local businesses to create new websites for them, to structure them in the right way, and all of this. And I was talking with someone, um, a mentor of mine, and he was like, Hey, Caleb, what have you been up to? And I explained, here's what we're doing with AI and SEO and all of this. And he was just like, Is it working? And I was like, Yeah, absolutely. I mean, the they're all right, not all of them, but they're ranking, like the ranking is getting better, the rank maps are training green, really happy with it. So he was just like, Well, can I sell your training to my online group? He has his own online course, he has like 10,000 members or 12,000 members, something insane like that. So he's just like, Would you be willing to come on and sell your training? Because I built all of this and training for my internal team. He's like, Would you be willing to sell that to my online group? And I was like, sure, we can do that. So we go, uh, did a pitch. I walked through what it was, uh, what it was designed to do, how it used AI, all of this, and it sold really well. Um, like it was uh six figures in sales uh for that pitch. And he was just like, Caleb, you need to start a YouTube channel and uh start selling this to a wider audience. And you know, over the years I have uh ignored his advice um more often than I should. So uh for this time I did not. I was just like, you know what? I'm going to take your advice and and and let's let's do it. So it was January of 23. So right around that time, that I started to like, okay, I'm gonna get serious about YouTube. Uh so I started posting regular YouTube content. I created a school community, uh S-K-O-O-L, I whatever. Alex Harmozy can call it whatever he wants, I suppose. Um and I started having a the call to action for my YouTube videos was hey, I'm gonna show you a bunch of AI prompts that we use for SEO. If you want those prompts, there's a link in my description to the school community. Um, I kept it free for a long time. Basically, I had the the free version and then like all of my actual like over-the-shoulder recordings that I had originally done for my internal team, they were in uh an unlock section that they had to pay money to get unlocked. Um and once the free group got to have almost it was like 1800 members, 1900 members, it started to be just a massive headache to manage because all of these scammers were joining the group and they would post all of this crap in there. So I was just like, okay, I'm not gonna hire a moderator for my free group because that doesn't make any sense. And by the way, like this is this is probably like 15, like 12, 15 months into my weekly YouTube journey where you know I'm I'm paying an editor, um, I'm paying like it's all these uh outgoing expenses, and yeah, uh to show for it, I have a growing free group, like you know, whoop-de-doo. Yeah um but it yeah, it takes time. So if I made the free group paid, um it's still paid, obviously. It's a whopping$27, and people were joining it quite regularly. I think if you land on the school about page, it doesn't tell you how many members are paid and how many are free. It was just like, hey, this group is$27 a month and there's$1,800 members in it. Uh so that's just an insane amount of social proof. Um, it doesn't say this group has three paying members and 1,800 free members. Uh you want to be the fourth paying member. It doesn't say that, which is very nice. Um, but yeah, that group. Is now uh just approaching 3,000 members, which is very cool. And it's been the same, you know. Here's a video, we're gonna use AI to get clients, we're gonna use AI to help manage your agency, we're gonna use AI to do local SEO. Here are the prompts I'm using, join my school community, uh, and you can have them there. I always show the prompt on screen too. So it's like, yeah, feel free, pause the video, you can transcribe the prompt, or you can get it in my school community. Um and then, yeah, eventually redesigned the real course, the that was before just internal team training, uh, redesigned it into something that made more sense as a course, uh, re-recorded it, uh, released that, and that's now my second school group. So the first one at 27, it gives sort of the why and what to do. It gives a little bit of details on how to actually execute, but most of the time that the material in that course, of course, will say, hey, you know, if to get these like detailed over-the-shoulder trainings on exactly what we do at my agency, it's over in this pro community. You know, you can find out more information here. Um, so yeah, that's basically the structure. And um probably, I say probably, I know it's about 20% of the people who join my$27 group end up joining the pro group, uh, which is$197. Um, yeah, I mean, that's basically the funnel I've built. I don't have anything above the 197. So I still have my local SEO agency. Uh, as I'm sure you know, John, the profit margins on running an agency are not anything like the profit margins on running an online course. So I joke that you know, the reason I keep my SEO agency is to get make sure the content in my pro group is actually valuable and useful. So a quick example: my content director for my agency, uh, she'd been with me for six, seven years, and she had an opportunity come up, uh, was excited for her to get it. So she left to pursue that, uh, was happy for, but now I needed to hire a new content director. Um, so my process of how I found the right person, how I trained that person, what that person does, how we manage client content on an ongoing basis. Uh everything that I did as part of that is basically in the pro group now. Like I just recorded myself going through the process of hiring a new content director for my SEO agency. Uh, and yeah, people really responded quite well to that type of content, which obviously I it would have been really hard to do if I didn't actually still have an SEO agency. Yeah.
SPEAKER_03:Yeah, yeah, yeah. Like the testing ground. Yeah, it's like I get um people from my team from the agency on on the podcast that discuss well, what did you do and what worked, what was the result. Because I haven't I I don't do a lot of this stuff personally myself, so I have to be like constantly checking, am I up to date? Show me the checkout page. What ones the uh the in the A-B test, which one did better? Why did that one do better? What's better about that? So I can make sure that the content I'm sharing is like as accurate, you know, as as up to date and as um helpful as it could possibly be. Um okay. So uh what is the process? I I I heard you describe it, but I just want to check the process from someone watches YouTube video to becoming paying customer. It sounded like, if I understood right, someone watches the video, maybe they then find that useful, go on and watch another video. But within it, you will at some point have here's how I'm using AI, here's the prompt. You can pause the video to transcribe the prompt, or you can just get this and everything else in the membership. Is that is that right, or are you also pointing people in the description? Are you pointing, have you got uh are you building an email list? Like what's what's the process overall for Yeah?
SPEAKER_00:So the the that's correct. There's a link in the description also. Um I have an email list, but honestly, the only way that people can get on the email list is by joining the school group. And I've gone back on this, back and forth on this several times. Like, should I have an email list and then try to convert that email list to the school group? And every time it's been like, you know, the school group is$27 a month. They've watched a handful of my YouTube videos, and one of the great things about YouTube as uh a venue for this is I'm sure most people have experienced, if you watch a video from a new channel that you never watched before, suddenly your recommended page is all that new channel. So it it YouTube almost takes the role of an email list because I don't need to reach out and do all the touch points that I would normally need to do to convert someone. YouTube basically does that for me. And if they're interested in it and they watch more and more videos, then that serves as the email list. The community itself, the first step is$27. So part of it is, man, if if they if they're not going to convert at$27 as the first step, then do I really care if they're on my email list? And email, it's very effective, but there are a lot of people that it isn't really going to work for. So how many people am I losing by dropping an email newsletter sign up instead of here's the link to my school community? So I haven't split tested it, split-tested it to be honest. Uh, we do send weekly emails to the whole list, but most of those emails are going to be focused on upgrading to the pro membership or what's going on in the world of AI SCO, all of these types of things that those people would be interested in. I don't pitch the$27 group to that list because the only way you can get on that list is by having joined the$27 group. So I know it's a little weird, but I don't know. It just it felt it felt problematic to force people to join an email list and then try to get them to join school from email. Because remember, the school group was free initially. So that's sort of how I fell into this. Um and my plan was okay, I'm gonna make the group$200 a month, put everything in that group instead of having this separate group, and then I'm gonna have the email as my call to action for my YouTube videos. Uh, but I was I was chatting with someone else who had a school group, and he was like, Man, just make your school group like$27 or$37 a month. So I was like, okay, so I did that. And within a couple of months, after putting a$27 price tag on it, the school group was making five or six K a month MRR. So I was just like, Well, I don't really want to turn this off. Uh, I'm gonna go create a second school group and we'll just leave this one at 27.
SPEAKER_03:Yeah, I mean, normally everybody I talk to, I'm like, okay, well, you you're not making nearly as many sales as you could be doing because you've not got an email list, you need to be running your email promotions. You've just hit a seven-figure run rate. I'm like, maybe stick with that. That's pretty good. Like, that's nearly all of this podcast is about here's how you get more people on your email list, here's how you get that email list to buy regularly, here's how you increase your revenue per sale. But uh I don't know, it sounds like you're doing really like maybe you could do way better, right? If you use the systems that we normally use, but it's doing really well. Like that's a that's a lot of money, seven figures a year, you know?
SPEAKER_00:It is, it is, and yeah. And I mean, I'm looking at the metrics for my group, the conversion rate to the group is nine percent, which is almost certainly part of it.
SPEAKER_03:Nine percent of work.
SPEAKER_00:So when people land on the join this group page, nine percent of them actually end up joining. Uh percent high.
SPEAKER_03:Do you know how many people get to that page a month versus how many views you have a month? I do. Well, I used to.
SPEAKER_00:They changed they changed the metrics recently. Um past 30 days, no. Oh, here we go. So in November of 25, I had 212 people join the group. Okay. So that's 10%, which is around 2,000 people or so uh landed on the about page.
SPEAKER_03:Out of it was about 138,000 website for YouTube views, right?
SPEAKER_00:And 133, yeah.
SPEAKER_03:So you had 1.45% of people ish, uh so one and a half percent-ish of people from YouTube got as far as the sales page, and then 10% of those bought. It's pretty strong.
SPEAKER_00:Yeah. And if the sales page were like a free email sign-up where I would email the prompts to them, I would assume that the sign-up rate would be probably north of 60 or 70%. Uh-huh.
SPEAKER_03:Yeah, that's the normal when you're coming from YouTube, yeah.
SPEAKER_00:Yeah. But then if if the sign-up rate is even 70%, I mean, I would need my emails to be able to convert. Well, I'm I'm I'm quickly running out of the ability to do this math and those-ish. Yeah. 15 yeah, you're right. 15% of those. And I mean that's a that's high.
SPEAKER_03:Yeah.
SPEAKER_00:That's a high email conversion rate.
SPEAKER_03:Yeah. No, I think stick with it. That's fascinating. Okay, well, let's look at then at what are you doing with YouTube? Because um, as a as someone who's using AI everywhere, can you talk us through like what's your process for using AI within your YouTube video content creation? Or wherever?
SPEAKER_00:So it's um we used to use a lot of outliers, uh, the outlier strategy where you look at other channels and look for the outlier videos. But in the SEO niche, the vast majority of outliers were basically rank really fast. Uh and I did that video a lot. So I'm trying not to do it as much because it gets really boring.
SPEAKER_03:And you say that, but your the your video from last week is I spent 10 plus years ranking businesses number one, this ranks the fastest.
SPEAKER_00:Yeah, yeah, yeah. That was definitely a rank fast video. Absolutely. Yeah, absolutely. I'm st it's like it's like this this drug, and I'm trying to get off of it because they they work well, but uh, if you look at the comments of especially the over the last couple of months on the rank fast videos, I'm getting like long time viewers who are commenting on the videos, like this is just the same video over and over again. And it's just like you're like, I mean, you're not wrong.
SPEAKER_01:It kind of is.
SPEAKER_00:Um so trying to do different things, different types of videos. So, what we've been working with is we have a prompt that uh gets Claude is almost always the model that we use for work-related things. Real quick side note I think it's hilarious that in November, right, Microsoft had uh invested$13 billion into Open AI, but in November, Microsoft integrated Claude into its Office products because Claude's output was better than ChatGPT, even though Microsoft had invested$13 billion into ChatGPT. So I just it's kind of hilarious. What's what's Copilot normally Copilot's their thing, right? Microsoft? Copilot is their thing, and copilot uses ChatGPT for most of the queries. But if you're using Copilot with Excel or PowerPoint, it's going to be pulling on Claude and not ChatGPT.
SPEAKER_03:I thought it was that a whole other tool that they'd built, but it was bait is based on ChatGPT. Okay, got it.
SPEAKER_00:Yeah, Microsoft doesn't have their own uh model. They're working on it, but they don't have one. Apple doesn't have one. Pretty much all the other major tech companies do. But uh yeah, because I mean Amazon um is a part owner of Anthropic, which does Claude. So we can say Amazon has one. Um but yeah, got it. So we have a prompt that we use with Claude to get it to look at Reddit, uh, look at local forms, look at Facebook groups, and try to find uh questions about local SEO that people are asking, and then just uh work to start to record videos that answer those questions. So the process, as with most YouTube, we always start with the title, the upload title, and the thumbnail, of course. Uh then we do the 30 seconds, the first 30 seconds of the video. Um, and then after that is done, I will do a brainstorming session. And I almost always use chat GPT for this because I have the voice chat. Claude also has a voice chat, but it's really clunky for some reason. Um, but the chat GPT voice chat, it honestly feels like a phone call with someone. Like it's just really good, like almost in a creepy way. Like the Chat GPT agent will sigh and you know, clear its throat. It just it anyway. So my team loves that also because I'm I'm the type of person I like to brainstorm out loud, and I can brainstorm out loud at ChatGPT instead of my employees, which is very nice for everyone. So, you know, we we have the title, what we want to talk about, I have the conversation with Chat GPT, everything that I want to say in that video. Um, then we have a prompt for uh I'll take that transcript, I'll give it to Claude, along with a prompt to have Claude organize that into an outline. Um, we'll review the outline, make sure like the AI is really bad at the setup payoff structure that's important on YouTube videos. What it wants to give away the payoff uh at the beginning of the section instead of you create that setup payoff structure, right?
SPEAKER_03:It wants to write like a newspaper instead of like the most important line, describe everything in the headline, then describe everything in the first line, and then a bit more detail in the first paragraph, like reverse pyramid. And what you really want is pyramid with with YouTube, yeah. Okay, fast. And it's Claude that does this wrong, is it?
SPEAKER_00:And it's Claude that does it, and I've tried fiddling with the prompts so many times to get it to generate the actual pyramid to create the mystery and the retention, but it just keeps defaulting to writing like a newspaper. I know that so ChatGPT was basically trained on the internet. ChatGPT, starting with ChatGPT 4, it was trained on the internet. And when you train an AI model on the internet, uh you very quickly end up with a model that will devolve into you know very problematic type behaviors because it was trained on the internet, and that's what a lot of the internet has. It's gonna be so much anger and Snide comments.
SPEAKER_03:Is that why it's open Snide?
SPEAKER_00:Yeah. Do you really want to interact with a model that was trained on Reddit and it's going to behave like the worst examples of Reddit? So to fix that, uh OpenAI had to use this technique. It's called like RLHF. So it's like reinforcement learning with human feedback. So they literally hired a bunch of contractors in the developing world to review Chat GPT-4's output and provide better output, uh, feedback to fix the output, basically to try to install these guardrails to train out of it the internet. Uh, the founder of Anthropic, uh, Dario Amade, I think is how his last name is pronounced, he used to work for OpenAI, and he had a fundamental difference of opinion versus Sam Altman on how models should be trained. Um, he thought the training set should be much smaller, much more high quality, and not requiring as much of that uh human feedback to get high quality output. So that's how Claude is built. Its training set is like one to ten percent of Chat GPT's training set size, and it uh it was trained on a lot of like high-quality written content. So I think that's one of the reasons when I'm trying to use it for YouTube script writing, that it's writing like a newspaper writer because it was trained on so many newspapers, because that's fairly high quality content as far as internet content goes. Um, so you know it's hard to get it to break that habit. So that's why the outline still needs help, still needs uh adjustment and work. But once we have the outline, uh where we're happy with it, I have another prompt with Claude that writes a word-for-word script based on that outline. Uh and that one usually works really well. Um that one is you know, I gave it dozens of transcripts of prior videos and tell it, you know, write in a tone, cadence, style similar to these transcripts. Here's the outline. You know, don't add anything, don't add summaries, don't add introductions, just follow this outline. And that it can do pretty well. So usually from outline to finish script is pretty quick. The bulk of the work is in uh developing a good outline for the script.
SPEAKER_03:Got it. Okay. So you're using I think you s no, which tool did you say you're using for the going through Reddit and everything in the first place for brainstorming? Claude to go through Reddit and other places to brainstorm ideas. Yeah. Then what are you using uh Claude or ChatGPT or anything for the title and the thumbnail and the in the first 30 seconds?
SPEAKER_00:No. Um they're in my experience, they're quite bad at that.
SPEAKER_01:Okay.
SPEAKER_00:Um so what what we'll do there is, and I know we briefly talked about outliers, so we'll use outliers not necessarily in the SEO niche, uh, but we'll look at outliers just in general for broader channels to see what type of titles are working really well, and then we can use that for the SEO niche. So, a common example, you see videos like this all the time on YouTube because uh Leia Hormozy did one that says, like, give me five minutes and I'll teach you about XYZ. Uh that title is everywhere now because that video performed really well. Um, so it would be looking for something like you know, uh uh title structures like that that have done really well on Outlier and similar thumbnail uh inspiration for outlier type videos that have done really well, and then we can take those structures and apply them to our video content, if if I'm making sense with that.
SPEAKER_03:Yeah, yeah, absolutely. Okay.
SPEAKER_00:But I have heard Gemini does an excellent job with this. I haven't played around with it yet, uh, especially the new Gemini 3 Pro that came out like a week ago, because Gemini 3 was trained on YouTube. Of course, Google owns YouTube, so it was trained directly on YouTube content, which upset a lot of uh YouTube creators. So I I I I am looking at for the next batch of videos, uh, using a little bit more Gemini in that first process for thumbnails and um video uh upload titles because it should do well in that. And it's kind of fascinating, right? Because Gemini was actually trained on the video of the YouTubes, like Chat GPT, they downloaded all these transcripts and trained the model on those transcripts, which you know we don't need to talk about copyright violations that OpenAI did. But anyway, uh Gemini wasn't trained on transcripts, it was trained on the actual videos, uh, which is fascinating. Um so yeah, I'm looking forward to giving that a try, playing around with some prompts to have Gemini like analyze video patterns, upload titles that worked well uh that we can apply to our stuff. Um I find that in developing prompts, so when I talk about using AI, uh I the biggest mistake that I see a lot of people making when they're using AI is they don't know enough about what they're doing to be able to catch AI's mistakes. And AI makes constant mistakes. I think even the best models are like 92% accurate, is the most recent number I saw with Claude, uh, which means 8% of what you're reading. In Claude's output is just completely made up and wrong. So the way that I always think about using AI is a subject matter expert can replace a team of interns with AI, but an intern cannot replace a subject matter expert with it, right? You need someone who knows enough about that topic to be able to recognize when AI is getting it wrong. Like even I described how Claude is incapable of structuring the pyramid that you need for retention on YouTube, it just can't do it. If I didn't know anything about YouTube scripting or YouTube retention, I would just take Claude's outlines, record the videos, and then wonder why my channel wasn't growing. So what I like to do when I'm developing a new prompt is I will tell the AI exactly what I want it to do and have it write the prompt for me. Then I'll run that prompt, I'll see what the output is, and I'll copy the prompt and output, give it back to the AI, and then give it my feedback on the output. Is it what I wanted? What's wrong with it? What's good about it? Uh, how do we need to adjust it? And then ask it to fix the prompt so that the output is closer to what I want to see. And then doing that, you know, usually three or four times, end up with a prompt that's really, really solid. Nice, nice. That's a hell of a workflow.
SPEAKER_03:I love it. All right, cool. What's next? What's next? Yeah, boy, are you there? You've hit seven figures, done. Business done, you know.
SPEAKER_00:Yeah, um, man, uh, I mean, the big thing that I've been focusing on over the last couple of months is honestly, so I talked about the young Timmy. Uh that work we did back in March. Um, and in March is when I got really serious about the YouTube channel. So that was like seven, eight months ago. Um, my my higher ticket community didn't exist in March, and my low ticket community was bringing in like 5k. So the growth from there to October happened in seven months. And a lot of that goes down to I basically cleared my whole calendar in March, and I bought every single course on creating YouTube content that I could find, consumed all of it, um, fed a lot of information to the to Claude, to different AIs, looking for how can I improve the channel? How can I make better videos? How can I do all of this better? Um, and that really worked. The channel growth started to take off uh in April, uh, started to get a lot of people joining the school groups, as we know. But in the last couple of months, the it's not declining, but the growth has notionally plateaued. So my focus for December is to get back to the basics. Okay, now I need to make another change, step change in video quality. And I'm thinking to do that second step change, I'm I'm probably going to be looking for experts because I spent so much time and energy learning the basics, how to do it myself back in March. I don't know that I'm I'm an engineer by training, right? I have no experience in production or script writing or anything like that. So that's why I'm I'm I'm I've I'm trialing. Um, I've hired, I don't know, half a dozen people, uh, script writers and YouTube strategists, and I've hired them all at the same time, just trying to figure out you know, what can we do to keep the channel growth going where it had been, basically making slightly better videos. I've I've watched a lot of interviews of Mr. Beast, because obviously he's very good at YouTube, uh, regardless of your opinions, he's very good at YouTube. And one of the things he said that really struck me is if you make a video that's 10% better, it doesn't get 10% more views, it gets a hundred times more views because YouTube is always looking for the best possible video to show the audience. So if yours is a little bit better than your competition, you get all the views. Um, so that's notionally what we did back in March-April, like that little bit of let's make the videos a bit better. And I'm trying to repeat that, but as you know, John, right? Depending on what level you're at, the next little bit of increase just takes a massive amount of energy.
SPEAKER_03:Yeah, I mean, I guess it's gonna depend on on which which area in your videos is the weakest. Like looking at it from kind of bottleneck uh not exactly bottleneck theory, but like this approach that we take with funnels is like which step in the funnel is furthest from the benchmark? Work on that one because you're most likely to be able to improve it. So 100%. Is it your retention? Is it your click-through rate? Is it your you know, first 30 second retention, whole video retention, watching the next video? Which step is it that could be improved most easily? Is it you know, is it the thumbnails, is it the titles, whatever? And then any work done on that is m is probably most likely to. Do you have a sense of what step that is?
SPEAKER_00:Yeah, so my overall video retention is it could be better, like most things, but um on most of the recent videos, it's in the 40 to 45 percent. From what I've heard, like 50% is the gold standard. Um, but that's close. I think click-through rate is my biggest downfall. Most of my new videos will barely crack 10% CTR, even in the first few hours, which is when YouTube is showing it to people who watch my content the most. Um, my under you probably know better than me. My understanding, the gold standard, especially for a brand new video that you just published, you should be easily 10 to 15% click-through rate with long-term CTRs in the eight to nine, and I'm not there yet. So most of my focus has been on uh thumbnails and upload titles and the first 30 seconds. Um, to the point where I've I've almost I don't want to because people already complain that it's the same video over and over again. But you almost want to take the same script and just play with thumbnails and titles and the first 30 seconds, like three or four different times, upload all of them over a couple of weeks and just see what structure is resonating with people. But I I think that's hard to do when people are already complaining. I have the same stuff.
SPEAKER_03:Do you do a lot of A-B testing of thumbnails or or titles?
SPEAKER_00:Yeah, so the titles is brand new, right? We just started that like a couple weeks ago.
SPEAKER_03:Yeah, yeah.
SPEAKER_00:Yeah. Um, and it's funny, there's one video that wasn't performing, and uh I we we did uh an A-B test on the thumbnails, and I remember the one that I was sure was gonna be so one of the core concepts of my process for local SEO, we call it the core 30. No one knows what that means because I coined it uh in my agency. So we have a thumbnail, like an upload title, and it was core 30 method was in the upload title. And I was like, Well, that's not gonna work because no one knows what that means. That title crushed it. So I was just like, right? You know, like exactly, exactly. But it's like obviously I don't know anything about thumbnail titles because the one I almost didn't run the split test on. I mean, it wasn't even close. Like it was like 20%, 25%, and 45% were the watch time metric. So it wasn't even close. And it was just like, man, I barely even wanted to test that one.
SPEAKER_01:Yeah.
SPEAKER_00:So that's where a lot of the energy has been in uh like thumbnails, title tag, thumbnails, up. I say title tag, an SEO guy. So that's where the focus has been on the um upload title, the thumbnail, and and the first 30 seconds. I um there's a YouTuber that I watch pretty often. Uh his name is Arvin Haddad. He uh does mansion tours and critiques, not related to this space at all. Uh, but I'm in the process of uh building out like a YouTube studio so that I can uh focus on it a bit more. So I booked time with him because he talks about resale value of houses, things like that. So um we ended up talking about YouTube, not what I had booked at at all, but it was great. And he was showing me some of his channel metrics, uh, which I won't repeat, but my 30-second retentions, which I thought were pretty good, they're they're nothing close to what he was showing me. So I think it's really in that yeah, upload title, thumbnail, and intro hook. And that's where so much of my effort has been focused and will be focused in the next month or so to try to improve those aspects the most.
SPEAKER_03:Got it. What's yours at? What what kind of a regular video for you, retention-wise, first 30 seconds?
SPEAKER_00:Uh my 32nd one, and it's always confusing because a lot of it depends on um how many views it got overall, but most of the solid ones are in like the 63, 64 percent.
SPEAKER_03:Yeah, I'm getting about that, and I'm not that shit hot on YouTube, so I feel like you should be doing better than yeah. I hear 70% is like if you can get to 70, that's really solid. I can normally get like occasionally I get to 70. A lot of them are about 60.
SPEAKER_00:Uh my best ever is not 70. Yeah. So that and that's part of the thought, also, because I I've had the experience where I've watched a creator's video on YouTube, like the video, YouTube recommends more stuff from that creator, but the video it recommended is like a year or two old or something, and their background is different, it's not as good. And I don't know, I I feel so shallow admitting it, but it's just like, boy, I don't want to watch this. It just doesn't look as good.
SPEAKER_03:Yeah.
SPEAKER_00:So, you know, working on like the background setup, I have a bunch of stuff planned for that, but yeah, just like that 30-second because people will make their minds up, you know, is the second they see you. So making sure that first impression is strong, that obviously the content is strong, that it exactly follows with what they're expecting based on the thumbnail and the upload title. Yeah, and then the rest of the script is yeah, gravy if you can hold them and uh 50% overall comp because I think the the overall retention is based on the watch time versus the video length. So if they're watching at 1.5x, which most people do on YouTube, uh you won't ever get more than like a 66% retention because that's the whole video uh at 1.5x. I think so. You what how's what what's your process like?
SPEAKER_03:So we're not so YouTube isn't a focus at the moment because I'm doing the the podcast is the big focus, and podcast videos don't do as well on YouTube generally. Now there are exceptions, right? Modern wisdom uh is a massive success, and that's the podcast one.
SPEAKER_00:But um if you could have Taylor Swift appear, then you would do very well probably.
SPEAKER_03:The audio podcast does much better than the the YouTube one, and so I have done work on YouTube and then like started to get the hang of okay, this kind of video with this kind of script, with this kind of titles, does well and had a couple of successes with that, but it's not exactly our target audience. Like our target audience is such a very, very specific type of person. It's a course creator who has got really good courses and has got a YouTube audience who has already built up that YouTube audience to quite a good size, but doesn't know enough about email marketing and funnels. And it's like there's a relatively small number of those people in the world, I think. And so YouTube doesn't seem to be so far the best way for us to target them. So uh I've kind of backed off. I want to work on YouTube. That's that's like my my little hobby. Like, I want to go work on my click-through rates and my my intros and all of this. And uh my COO Yosip's like, come on, John, we've got this other stuff that's working. We need to be putting all of our time and effort into the thing that's actually practically working. I'm like, okay, Yosip. So I'm not I'm not an expert on YouTube. I've kind of taken the courses, learned about it, but only started to get the hang of improving it. Whereas most of my clients are like absolute wizards at this, you know.
SPEAKER_00:And there's a lot to be said for that. Understanding where to focus your time and energy as a founder and on what's working is a pretty good place to start.
SPEAKER_03:So I'm regularly reminded by Yosef when my ADHD brain wants to go off and do something fun and new. Yeah. Yeah. This has been awesome. I absolutely love this. I love where you're up to so far with the business. This is like such a success story to have grown to basically to seven figures in in seven months. I mean, I know it's not exactly you were at something before, but like that's astonishing.
SPEAKER_00:Eight years of pre-planning, yeah.
SPEAKER_03:Um, if people want to go check you out, they want to check out your YouTube channel or anything, where should they go?
SPEAKER_00:Yeah, YouTube channel is Caleb Ulku. Uh last name is U L K U. It's a weird last name, so you'll find me. Uh yeah. And then uh my website is also just calebulku.com.
SPEAKER_03:Beautiful.
SPEAKER_00:All right. Caleb, thanks so much for coming on. Really, really appreciate your time.
SPEAKER_03:John, it's been an absolute pleasure. Thank you for inviting me. And as always, thank you so much for listening. Love you guys. Talk to you soon.