This Is How Creators Use AI To Hack the YouTube Algorithm

In this episode, we break down one of the wildest shifts happening in content and marketing right now: AI-powered media at scale. After discovering a highlights channel from Alex Hormozi pumping out dozens of videos per day, we reverse engineer how it works—and how you can replicate (and improve) the strategy using AI tools. We also dive into real-world insights from global marketing teams—what’s actually working, where AI is falling short, and why most people are wasting the time AI is saving them. If you want to understand where content, AI, and marketing are really headed and how to stay ahead this episode is a must-listen.

• 0:00Speaker 00

  Finding customer information shouldn't mean digging through emails, call notes, and spreadsheets. HubSpot Starter Plan brings all of those details together, so it's easier to track what's happening with every deal and what to do next. It pulls information from calls and emails. Then it gets to work, logging and summarizing them. Right after your call ends, HubSpot pulls next steps from your call notes and drafts suggested follow-ups. And it stays up to date for you, so you can build demand, win more deals, and spend less time on repetitive work. Your records, insights, and next steps are ready for you to act on. And you don't have to miss a beat searching for any of it. Try the Starter Plan for $10 a month. Learn more at hubspot.com. It looks like he's publishing like 10 videos per hour. Check this out. I just pulled it up. Okay, so Neil, you see on the screen? So six minutes ago, okay, one hour ago, one hour ago, one hour ago.

• 0:45Speaker 00

  Look, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10. And so 10, and then two hours ago, 11, 12. So 12 videos in the last two hours. And so he's publishing all this at volume, and I'm like, what the heck is going on, right? So I take the link from this channel, and I just dump it into my OpenClaw. And I'm just like, dude, what is going on with this channel, right? Is this all AI-generated? And it analyzes the channel through the YouTube API. And it's like 99% chances is all AI, okay? And I'm like, tell me how this channel has been performing over time. So I want to show you my interactions with OpenClaw and what it actually built. Because this week, you heard from Jensen Huang himself, every company needs to have an OpenClaw strategy, okay?

• 1:25Speaker 00

  So I'm like, oh, thank God, last seven weeks of me spending time on this crap has not been a waste of time, okay? So let me, let me, let me show you my screen over here. Okay, you see this? Yes. Okay, so again, I'm like, hey, if you use the API and look at this channel, what are the odds all this clip content is using AI? And how would we reverse engineer this, okay? So it's like 99% chances AI clip. The data is a smoking gun, 74 uploads in a day. No, no human can do that, right? Identical descriptions on every video. Okay, fine, great. And so I'm like, I'm not going to read all that crap, okay? Then I'm like, okay, I'm just not going to read that. How has this channel grown over time? Is this a viable way to grow? And then it's like, here's the full picture. Created in 2025, 58,000 subs, 8.2 million total views. And it's like, it shows the trajectory over time.

• 2:08Speaker 00

  So you can see in February, he published 500 videos, got 4.3 million views from it, okay? And then 341 in March to date. And then it's like, the pattern is clear, here's what's happening, average view is declining. But like, is it viable for growth? And it's like, yeah, volume is lottery ticket. So it does all this analysis. And then I'm just like, okay, whatever. And then I'm like, can you just? Do this for me? And it's like, yeah, you know, if you spend a five to $10 a day, we can do this for you. I'm like, just give me an MVP version. Okay. So Neil, all it does, it shows me this entire process. It signs up for all these different things or it uses them. And it makes me, based on our clips from marketing school, it makes me six clips that have nice titles that are at the right moment as well. And it neatly organizes them into Google Drive.

• 2:52Speaker 01

  I'm curious what the results will be in the long run. My hypothesis with Hermosy's is yes, sure. Every once in a while, there'll probably be a video or two that's a lot of ticket that gets tons of views. But the algorithm is going to say, hey, your channel puts out a lot of stuff. It's not performing well. We shouldn't push you as hard is my guess on what's going to happen.

• 3:12Speaker 00

  There's a good chance you're probably right, but I want to call out the fact that I can literally just take a channel, reverse engineer it, and then it creates a process and it dumps all these clips in here. I sent it to the editing team because Neil, you like, if we made like a highlights channel, your highlights channel, my highlights channel, whatever. It's, it's not, if you have the AI doing all the work, it kind of just doesn't matter. Right? So, because it's a channel that's not affecting your main one. And so it basically creates an SOP for me. And then I'm like, the fact that it mattered that you can sit on your couch and do this, reverse engineer something, have it tie into your goals and have it do the work and like gift wrap it for you. That to me is the unlock here.

• 3:49Speaker 01

  Yeah, I think the technology is amazing. And the cool part about this is you can take this and you can improve upon what was done with that strategy. Because as Eric agreed to, and Eric knows this, and just like I know this, we've both been doing the marketing game for a long time. Typically, algorithms hurt you if you just keep pumping out crap. I'm not saying that content that he's pushing out is crap, but if you're putting out 80 videos a day or whatever the number is, it's hard for every single one to be amazing in quality. So if you can take that strategy and take some human hours, call it even like an hour, two hours and fine tune it a little bit, like Eric even mentioned when he was showing me his, I'm assuming he's not using the same description over and over again. Right?

• 4:33Speaker 01

  It's more of a fine tuned version. So then as you start adjusting, you can take a That maybe I I wouldn't say hermose spamming. He's just mass producing. I would consider it quote unquote some AI slop in theory. If you produce a higher quality version of it, you can actually do something that can gain long term traction and potentially even drive revenue from it.

• 4:54Speaker 00

  Yeah, I actually think I will say I watched some of the videos and I think they're pretty good because Neil, you know, he does the the workshops in in Vegas, right? So there's a lot of content he can pull from that. And then you and I obviously have a lot of a big library of content we've done or, you know, events we've spoken at, right? And I would say even the early days, the first three, four years of marketing school, that's probably the us front loading our marketing knowledge. And now this is more like news related, right? So let me show you an example here. My point of calling this out, like I was just showing Neil the the video on the moments it's pulling. It's not doing a bad job. And I I showed my my internal content team this and they're like, this is actually really good. It gives us like a lot of, we call them atoms, like the right moments. But Neil, after that, I was like, oh, well, you know, we kind of don't do it.

• 5:35Speaker 00

  Well, let me finish this part first and I'll answer your question. So like, but then I was like, hey, we can do it for the the the the mid form stuff, but why don't we do it for short form? There's a lot of stuff we can do for short form because I don't think we do a good job of finding the right moments. And so my point is you watch something, you refine the idea to Neil's point, and then you can even make idea babies from it. But go ahead, Neil.

• 5:53Speaker 01

  So if you can go back, Eric, for any of those listening, what Eric was showing you is a clip of him. There was this chart on X that actually shows AI penetration and it's actually very low compared to you know, what people actually think. And it shows, look at how far we've gotten from AI with so little penetration, like what's going to happen when it actually gets penetrated? I believe that was what the clip was from from last week or the week before marketing school. You can correct me if I'm wrong, Eric, but it sounded like that.

• 6:22Speaker 00

  Hotel, do you look like you're in here? Is it the Hong Kong one?

• 6:25Speaker 01

  Uh, Hong Kong. Yeah, I think I was in Hong Kong. Yeah, but the the point I'm getting at is like it was a good clip. It sounds like AI went through the whole video to find the clip. Is that correct, Eric? Pulled it out and then what did it put as a description or title or anything?

• 6:42Speaker 00

  The title is eighty-four percent of people have never used AI.

• 6:45Speaker 01

  Yeah, so that's a great title. Um, and yes, it could be improved, but you have to keep in mind this is AI doing it. So it's not a bad first take. Does it do a description for you as well? If not, you can create one.

• 6:55Speaker 00

  I think it does a description, but Neil, let me just show you this. So after it did that, it made an S. I had it make an SOP for me and there's like this is it's like a seventeen page SOP. So you can literally dump this into whatever, you know, AI tool that you're using.

• 7:10Speaker 01

  So you want to break down what the SOP is for everyone listening?

• 7:13Speaker 00

  Look here, let's have Gemini summarize summarize this document. See, this would be a good moment, Neil, where it's like, hey, this would be a good lead magnet. But like, I don't want to spend the time to make a lead magnet.

• 7:28Speaker 01

  Yeah.

• 7:28Speaker 00

  So.

• 7:29Speaker 01

  So Gemini did a good job of titling it. It said, this is a SOP for end to end. I think YouTube publishing using AI clipping or something like that.

• 7:38Speaker 00

  Yeah, so this pipeline has been validated by MVP results, which produced fifteen standalone clips for three sixty-minute leveling up episodes in about thirty minutes wall clock time. The strategy is based on the success of Hormozzi Highlights Channel, which achieved eight point two million views and fifty-eight thousand subscribers in six months by publishing approximately six videos a day, a lot more than that now. Um, so the process, I'll just make this really easy to not bore people. First step is to download. So download the YouTube URL to the to an MP four file. Okay, cost per episode free. I like how it has the cost here too, Neil. Number two, it transcribes, so it uses Whisper, which I don't know what that is, by the way. Transcribes the MP four file to a JSON with word level timestamps. This is free, local CPU. Okay, segment, then uses the Claude API. This is the secret sauce.

• 8:21Speaker 00

  This costs fifty cents to a dollar per episode, which is actually pretty expensive. I hope it doesn't cost that. And then analyze the transcript to identify the three to five best segments for clips, applying criteria like a clear hook, complete thought, and an ideal length of three to fifteen minutes. So this is like more like a mid-form clip. And then it cuts it up using FFmpeg. That's free. And then uploads YouTube data API. That's also free. So I do think with this Claude API one, these costs are going to come down over time too. Or you can use like a like a, you know, like, like an open source model if you want.

• 8:50Speaker 01

  Yeah. By the way, you read something in there. It says 30 minutes of wall clock time. What the heck is wall clock time mean?

• 8:56Speaker 00

  Yeah. So this is interesting because I, whenever I work with an AI, I tell it to give me two timelines. So Neil, I tell it to give me the human timeline and an AI timeline. So wall clock time is really just human timeline because

• 9:08Speaker 01

  Ah, okay.

• 9:09Speaker 00

  You know how the AI is trained on human work, right? It's like, oh, this is going to take like, you know, two weeks or whatever. I'm like, no, stop giving me AI timeline. Stop giving me human timelines. Give me AI timelines. How long is it going to take? It's like 15 minutes. I'm like, great, do that.

• 9:21Speaker 01

  Because I was like 30 minutes. I'm like, this couldn't have taken AI 30 minutes to do this. It should have been quicker.

• 9:26Speaker 00

  Yeah. Well, I don't know. I mean, it does look through a lot of different things, but who knows? Because it pulled 15 clips and then like all that stuff like so. Anyway, I just wanted to show that. All that to say, guys, that is it for today and we'll see you tomorrow.