I paid this man $15k to Learn Ads 10 Years Ago... Here's What He's Doing Now (Justin Brooke)

Yeah. Anyways, though, man, if you don't know Justin Brooke, he is legends in the media. If you haven't heard his name just now, you're going to learn. He the first person I ever learned or paid money for. Say, for John Bloomer. I bought John Bloomer's $100 course. And then Justin came to our office when I was a clever investor a little over 10 years ago, taught me ads on a power day, and the rest is history. I'm just going to chalk it up as you are 100% the reason, Justin, is to why I got to where I'm at. So thank you, sir. I appreciate you a lot.

Yeah, man, happy to be on here. Thanks for having me. Your audio is going crazy, so I just wanted to pause for a second, let it catch up. All right, go ahead. I was like, dang, I already screwed up. All right, let's try that question again. When you bounce in your chair, it's throwing your whole mic. It's. It's throwing your whole mic off. So, like.

Yeah, how about now? I'll stay still.

Yeah, now it's. Now it's working.

Good. Yeah, my bad, dude.

All right, no worries.

Bars in tone. We're back. So, hey, Justin, I was just talking about how I just haven't seen you in five years. What have you been up to? D Flash five.

I've still been running Ad Skills, helping the next generation of advertisers. We have a certification program and kind of the latest thing is AI is huge now. And it's not that we're switching to AI because advertising doesn't work or anything like that. It's just every career, AI is touching all of them. You know, AI is gone to the doctors, to the lawyers, and it's to the advertisers as well. And so we're pivoting a little bit as the market pivots and showing people still the same thing. You know, how to do Google Ads and Facebook ads and TikTok ads and YouTube ads and all the different ads. But we're just showing them, like, how to do that with AI, right? How to use ChatGPT to write better ad creative and how to use AI agents to, you know, analyze your campaign data and just, you know, all these different things is, you know, that's kind of where we're at these days with things. And then also, personally, I have stopped traveling. So I was traveling when we met last time. I pulled my RV onto a 17 acre piece of land, and we're just building that out. We're building an a frame log cabin right now and should be done by Christmas or Easter. Hopefully Christmas, but probably Easter.

Whereabouts in Tennessee?

We're up in the north way in the mountains, you know in the north corner kind of there by Virginia and North Carolina.

That's the most beautiful part of the whole state in my opinion. Right in between.

Yeah, it's gorgeous down there.

Yeah, good voice. I'm from the country too, so that sounds like the dream. I'm here trapped in the city still or Phoenix, not really the city, but at some point I'm going to get there to where you got. So I think that's probably the definition of success. Dude, I'm going to join that community. I was in it and especially when I was buying. I'm going to get back in it because I've been trying to get our team onto an AI workflow that's actually going to be more useful from an outcome standpoint than just an efficiency standpoint because AI already solves solves everyone's like output. You know, like work is like cool. I made a hundred ads today, but did I get a hundred x the output or outcome? No, definitely not. Maybe like two times the outcome. So have you been able like what have you tried that hasn't worked with AI and what have you tried that has worked that like specific with media buy?

Yeah. So there's levels to it just like everything in life, right? I mean you pick up a basketball and you go out to the court, you're probably not going to be landing a bunch of free throws and dunking or anything, you know, but you'll stumble around and you'll get better and then you'll get better and then you know, before you know it you are landing free throws and doing dunks and stuff like that. So that's like with anything. And so I would say that some people, they pick up AI and they give it a try, they're like, hey, you know, chat GPT, write me an ad. And they don't understand multi shot prompting, meaning that you give it examples first. Right. And so they don't understand that they should give it a role, a context and a task. And so when you don't understand some of these things, there's also another technique called answer leveling and we can talk about these different techniques. So there's role context tasks, there's answer leveling, there's multi shot prompting. When you don't understand these techniques of using AI, well then yeah, you're not going to get anything good out of it. And when we're talking about working with machines now, we may not be coders, right? We don't know how to code the machines, but we are putting an input into the machine. A prompt is an input, and so the output can only be as good as the input we give it. So I hate to say it, but if your output sucks, it's not the machine's fault. You know, it's.

It's not.

You gave it something it couldn't work with.

You know, you brought up a good shot. So I only know one of those terms. And our. My audience is pretty nerdy. They'll love this too. And I want to learn. So multi shot prompting, I'm pretty good at it. Now. It's been two or three years of me knowing that. Yeah, Chat GPT is not going to ask you clarifying questions because most users don't care. They just want the confirmation bias, right. Reaffirmed. So, yeah, what is the secret about shot?

Yeah, so what you're going to want to do with multi shot prompting is you're going to ask it to write an ad for you. But what you should do is you should give it at least one example of an ad. Like, imagine you had a junior copywriter sitting here in the room with you and you're like, hey, write me an ad. They're going to be like, okay, like a long one, a short one. Is this going to be for a video script? Is it going to be like, what? They need an example to work from any web designer, any coder, you know, even if you go get your haircut, right? And you're just sit down in the chair and you're like, hey, cut my hair like long or like, yeah, they got no idea, right? You got to show them a picture. Like, hey, cut my hair like this. Right. You know, it's the same thing with AI and the more, you know, examples you give them, the better. You don't need more than three. Don't go crazy. You don't need to give it 27 examples, but give it a few examples of really good ads and then watch this thing cook. You know, like it's going to be like, oh, okay, I get it. You want an ad. I know how to do that. You want it like this. All right, perfect.

Boom.

Now I'm going to start writing ads just like that.

Okay. Do you mind giving me a specific advice? Because this has been stuck. I was doing this this morning. I can't figure out how to fix this. Is there a point where you just ditch the chat because the context window and just start a new one. Oh, yeah, okay.

Yes, especially, especially if you're using Chat GPT.

What people need doing it just like, it wouldn't get there. And I just couldn't get it. Despite getting all the way to seven versions, finally got that to change and then like, it just wasn't getting it. Still doing merge fields. And I was like, dang. But if you lose that chat, then you have to upload that chat into the next one. So how do you deal with that? Or you just throw it away at that point?

Okay, so this is classically a problem with ChatGPT that some of the other ones don't have a problem with. But when I'm going to. When I know I'm going to do something over and over again, like write Facebook ad copy instead of just using a normal chat window, I'm going to set up a project. Okay.

I don't do that. That's a great advice.

Yeah. So you're going to set up a project that way you can give it instructions one time. You can give it. You could upload files so that you have examples. You might have your customer Persona.

Project works. Dang. I've been an idiot this whole time.

Yeah, yeah, yeah. So it's got the instructions on how to write a good ad. It's got the examples already built into there. You've got your customer Persona. And so now that way when you're working, it's. You don't have to input those instructions every time.

Right.

So this is going to save a lot of time.

Yeah. And I'll go and change the custom instructions for the whole thing if I need to, which is annoying as well sometimes. So that. Thank you. I didn't know. I taught project as folders literally up to. I should have known better.

And so another plus is, as you know, sometimes short copy works, sometimes longer folds form copy works. And so. And you should be testing both of them every time. And so you should create a project that is trained on how to create really, really good short copy Facebook ads and another project that's trained on how to create really, really good long copy Facebook ads. And so I look at each of these projects like a little assistant on my team. You know, that they're, you know, this assistant. They just do this job every single time. And, and I'm always trying to train them to be a little bit better.

I might even name them. Yeah. Like King Media Buyer or something like that. Is that what you do?

Yeah, yeah, yeah, yeah. I mean, I, I name them like my copywriter, my headline writer, my what? You know, whatever.

Yeah, cool. So how many do you have? Because I have a theory, I'm curious if it's right, of how many sidebar, like, functional. What's your AI team look like?

Oh, does that make sense?

How many?

There's. There's at least nine or 12. You know, I have them for everything from emails to blog post to, you know, ad headlines. Like I do headlines different than my ads, you know, subject line. I've got one for, you know, meal plan. You know, I'm trying to eat better, you know, and so I've got one that helps me pick out. Yeah. And my meal plan one, it not only does it give me, you know, recipes based off of stuff that I like, it will also give me like a grocery list, you know, So I know. Okay, I need to go buy these things. Yeah. So, yeah, saves me a ton of time having these different projects. And then also what's really great about me building the projects is now I can give my assistant access to my projects and now they can run the things. They don't need me to write my emails, they don't need me to write a headline. They got a whole tool that could, that can do that.

I'm doing this today. That's so helpful. And that is where the answer I was hoping would fall into. I have this theory. I don't know if I'm right, but eventually we're going to replace or at least human teams with either some sort of hybrid or full agentic. Right. I think you would probably agree with that assessment. And I remember in High Output Management by Andy Grove, he talks about the most direct reports you can handle is 5 to 11, like human ones. So I'm curious what that, what that agentic, you know, change will be for us. Can we handle more drip reports? Because they don't have the HR head case maybe associated with other employees. I don't know, but I'm just curious to see where it goes. Yeah. Yeah.

So what question did you ask there? What did you want me to show you? Okay. Okay.

But I do want to transition to. To from multi shot. Then what's the next one? Role context task is where you would suggest someone go from there.

Right. So instead of just write me an ad or even, you know, maybe you have, you know, seen some of these prompts out there and you got a nice long prompt in there or whatever you always want to start with. So if I'm writing copy, I'm always going to start with you are A world class copywriter. And I've tested different words in there. You are an expert copywriter. You are an elite copywriter. For me right now, world class seems to be creating a little bit better results.

Isn't that funny? Yeah, yeah.

It's weird, you know, but con. It's, it's a machine. It's a linear thinker, you know, and so you have to understand that you're talking to a machine.

Describe what you mean by that. I understand a linear thinking, I guess from maybe a human standpoint. But how did a machine learn.

Well, exactly. So when, when me and you, you know, like if I say give me a ballpark figure, you know that I mean like an estimation. Right. But the machine's going to be like, all right, well, the Yankee stadium ballpark is 350ft long, you know, and it. So it just, it just thinks in that linear. It thinks. I mean, exactly what I'm saying. And we're so used to having human conversations that the hard part of prompting the machines is understanding like that they don't understand us. Great. So, and I know this is a little stereotypical, and I don't, I don't mean to be like this, but I like to pretend that I'm talking to a buddy in Pakistan, right? And so there's a little bit of a language barrier, you know, he's not understanding me correctly. So, like, I just have to make sure that when I say what I say, I'm saying it very like specifically. You know, it's like, not just you are a copywriter, you are a world class copywriter. You have vast experience in. Or I'll say you have decades of experience in. And, and then I'll say exactly what it is. So, for example, if I'm writing a long form Facebook ad, I'm going to say, you are a world class copywriter, period. That's the, that's the role. Then the context is you have decades of experience writing Facebook ad copy. You know, long form, direct response Facebook ad copy. You always used known best practices in direct response and advertising copy. And then I'm going to give it a task. Okay, so that's my role in my context.

That makes sense.

Then I'm gonna send. Yeah, right. I'm explaining a lot more, you know, giving it the context of what it needs to, how it needs to think, how it needs to work. Then I'm gonna give it my task and I'll be very specific. You're gonna write me a long form Facebook ad. The first sentence is gonna be Short and highly compelling. Then the second sentence is gonna rehook, then the body of the email. Make sure you follow the problem agitate solution formula, you know, so giving it some context so that it knows exactly how I want my thing to be structured. And then if I'm using a project, it already has examples. But if I was using just a regular chat window, after I write that, I'm going to give it a few examples. Now, one other thing I want to tell people a little trick. There's something called xml. It's a, it's a programming language. It's an incredibly easy programming language. I know just by saying those words, some people are freezing up and whatever, but literally. So the reason I bring this up, when you give it a prompt and an example, the machine doesn't know when you're switching from prompt to example, you know, and so you need to be very specific with the machine, they're like, hey, you know, this is the prompt, this is the example. This is the example. You know, it needs a understand. Otherwise you're talking about the ad copy. And it's like, yeah, it's so delicious. And whatever. It's like, wait, what are we talking about?

That happened to me last night. I was just in a rush. I screenshotted something and I said, do this. And it's.

It did.

It actually did something really interesting. I was like, not what I wanted at all. I was like, funny, funny interpretation of that.

So, yeah, so I'm gonna write some XML on a card here just so I can make sure it shows up. Okay, so if I'm writing an example now, it might be backwards because of the camera or whatever, but that's it. That's all you got to do. You just got to put, you know, two carrots around the word example. And then at the end of it, you're going to put two carrots, but with a, with a, with a backslash.

Right?

So we've got a, we've got a slash through it.

Now. No one ever taken me through this. This is what I needed. Is like the basic coding guide for marketers.

Yes. Yeah. And so really simple stuff. You don't need to know a whole lot of coding or whatever, but if you just understand that when you're formatting your prompt, when you're going on to the next section or, you know, the, the do not do's. Here's one thing that some people do wrong. They'll in their prompt, they'll say things like, do not do this, do not do that. Well, the machine it doesn't. It doesn't know when you're switching back and forth, you know, and so. Right. So it's like, all right, here's the things to do then in the carrots, you know, do not do. Right. And then put the do not do things in there. Yeah. So these little tricks of just understanding that we're talking to a machine, we need to break things up, because most people are just writing one long paragraph. They're not even putting any spaces between lines, not using colons and hyphens. And, you know, you just need to really structure your prompt very well, and that's going to give you a world of difference. Like, if we stopped right there, you're going to be prompting way better than most people, honestly.

Yeah, Like, I would be that alone. I'm just going to say, note to my team. Let's put everything that Justin just said into a guide and drop his link in it, too. And we can give out off this episode because I'm going to forget. Yeah, like, that's hard. Doing it to myself. Two brackets. Two brackets and a. And a backslash or a forward slash. And then what? Maybe give us one more that we can put in there, too, if there's anything. Because I. I see that bracket ends and I don't know what the difference is.

Yes. So two other tricks that I would add is when you're done telling it what to do, you should tell it what not to do.

Perfect.

An example of that is we've turned all my courses into an AI chatbot. And so when it's answering somebody's question, I tell it, do not use your own knowledge. Only use the knowledge you've been trained on about my course. Because I don't want them. Who knows what ChatGPT is going to come up with, you know, in that conversation. Right.

Because I. It's 5050 with me. Does that phrase work for you? Because normally I say, scan uploaded knowledge. So I'm like, what? Like, scan my uploaded knowledge and my custom instructions and make sure you adhere to them. So that hasn't always worked every time.

Yeah. So I would say, and make sure you adhere to them. And then I would also say, do not use outside information. Make sure you stick to, like, I might say it three different ways, three different times, just to make sure that the machine really understands.

I've done that, too. And I'm like. And cite all of your sources because I'm not going to believe. Yeah. And why does that happen? Do you know of, like, why it hallucinates in that direction, despite that instruction.

So chatgpt of all of them will hallucinate a little bit more. And that's just because, to be honest, they're just burning cash, you know, and so their server costs more than what we're paying them per month, you know. And so what happens?

Yes, that makes way more sense now.

Right. So it's a flat fee. Right. So I don't know what plan everybody's on, but let's just assume we're talking about the popular $20 a month plan. Well, if you pay $20 a month and you never use ChatGPT, well, then they're making money. If you pay $20 a month and YOU only use it for five times per month, well then they, there's a little bit of a cost for them, right? There's servers, they have to run that thing. But if you're paying $20 a month and you're using this thing 30 times a day, then you may not be profitable for them at all. And so because they charge a flat fee, their machine kind of has to dumb down a little bit towards the end. That's why you see it, start hallucinating and stuff like that. And that's just. And so there are other tools that you can use that are credit based instead of flat fee based. And the credit based tools aren't going to dumb down because they want you to spend more credits. Right. So it's going to be full power all the way through because it doesn't matter if I use it a million times or 10 times, I'm paying for the credits. Got it.

So they, that makes sense to stop someone like me. That just makes me want to do it even more. Yeah. Last night I was, they were saying we can't do this, which is, I think inethical, the programmed morality into some of them. And I was like, okay, fine, then do it your way where you think it's going to benefit humanity. Just don't change it too much. It's like, sir, now that I know this, I can no longer complete the task. I'm like, great, you're right, the whole thing. What would you do instead? And it's like, I can't entertain this idea. I'm like, then and it doesn't now. So thank you. Yeah, yeah.

So there's a tool, there's another tool called abacus. Now there's a lot of them. I don't, I don't have a dog in the fight. You can use PO or T3 or whatever. Okay. What these tools are is they're called AI wrappers. Right. So all these different tools. OpenAI has a, has an API. Claude has an API. Deep Seek, they have an API. All these different tools, they have APIs. So what an AI wrapper does is they give you access to a bunch of models and then you can choose which model you want to use. You know, so perplexity is like this, right? I didn't know that abacus is.

What's that? They're called wrappers. I didn't know what the term was for it.

Yeah, an AI wrapper just means that it's including a bunch of different AI models in one subscription. Poe.com, that's Poe.com as an example. U.com is an example.

W R A P P E R S. Right. For anyone watching too, there are AI wrappers as well.

Yeah, yeah, wrapper like, like, like Christmas presents. Not like 50 Cent.

That's right. Although I can't wait for that era too.

So if you're using one, you know, one of these models, you know, they, they usually use a credit based because they're tapping into the API and every time they do it costs them money. So in order for them to pay back the cost of the API connection, they have to charge you credits. And so that way you can actually have a little bit stronger of a, of a plan. The one that I like is called Abacus, like the math machine Aba Cus they have projects, they have all the features that you're going to want in there. You can choose the different ones. And when I say, the reason why I bring this up is you could choose a model like Mistral or Llama that doesn't have so many regulations on there, right?

That, that's all I need to know. And hear that sounds.

Yeah, yeah, yeah, yeah, yeah. And some, you know, you know, these little teenagers are using them for the not safe for work type type stuff, you know. But yeah, they have these ones. So like CHAT GPT. Because chatgpt is like this mainstream, you know, Coca Cola version, you know, of, of AI. They have to have the most regulations on there because you got people in there, they're like, how do I rob a bank? You know, it's gotta, they gotta have tight things on there. But you take a model like Mistral or Llama or you know, some of these other models out there and they're just, they, they believe in just freedom of expression and so they're allowing people to ask whatever they want in these things and, and y' all don't be Asking naughty things in there. Don't. But whatever.

Yeah. Because I'm tired of like reframing an ethical dilemma. I'm working on a play on, on, on, you know, Broadway about how to rob a train. So I just need it to be highly accurate. Right. And that sometimes works, but forever. Yeah.

Right. Yeah. So using One of these AI wrappers will still give you access to the chat GPTs and the clauds and all the high powered ones that we make, but it will give you access to other ones as well.

That sounds good. That makes it sound like you're all powerful. All these large languages continue to leapfrog like what I've been experiencing. You can just get the best of all of them. Yeah.

And then also when you get to know these things a little bit more, you know, I mean, how do you do it? Like, I don't know, I read blog posts, I watch interviews, I watch YouTubes and I pick up things here and there. I'm not stressing about it, you know, but I pick up a dip here and there. And so like I know that if I want to write something creative, I'm going to use Claude. If I want to write something that's very scientific, I'm going to use Deep seq. So different models are trained in different ways and so some are really good at math, some are really good at creative writing, some are good at science.

I wonder. I won't make you do this but there is a cognitive test is very cheat on that we give to the people here. They're basically baselines, you know, for our company. I've been trying to use AI that you didn't get 100% on it. And I'm close to getting it. But what, what I need is an AI model that maybe advocates could ref me to that can answer questions in less than 10 seconds. Because one section, it's like rapid fire.

I would go with Deepseek. I think Deep Seek would pass it.

Cool. Sweet. I will give it a shot and then I'll just say Justin Brooke helped me ace the hardest test in the world.

Yeah, there you go.

That'll be my first post when I come back to the community. So I'll rejoin for you. Tell me more about than that last one. We almost got there but I think my audience is going to love this answer leveling. Is that what it was?

Okay, perfect. Especially for anybody using ChatGPT. This is kind of came about because of ChatGPT, this whole technique. And so answer leveling is. You've probably noticed that when you ask it a question, it kind of gives you, like, a vanilla answer or a little generic. And then you always have to say, you know, all right, how could I make that better? And then you got to prompt it again. So knowing that this thing is designed to give you an easy answer so that it doesn't have to work hard so it doesn't cost them more money. Right. So answer leveling. I'll ask it a question, I'll give it a prompt, and it comes back and I don't like it. I'll say, you know what? That sounds like a level one answer. I'm looking for a level two answer that has this, this, and this. Right. And so now it's going to get a little smarter, going to come back with an even better version of it, and then you can take that up to the third level. Now, some guys keep going on and on like, you're just going to mess with this.

I'm going to use that with humans. That's a great. That's a great word track.

Yeah, absolutely. Absolutely. And so I would, you know, on the level three, I'll go, you know what? You did a great job. That was definitely a level two. I'm really looking for, like, a genius level, though. Like, give me a level three answer.

So there's no top of the scale. Like, it would be like, take this from a nine to a ten. That's so smart. I love that. Yeah, yeah.

So that's answer leveling.

All right, cool. That was great advice. Thank you for all those. I have to imagine 99% of people watching, including myself, probably didn't know all that and get frustrated with ChatGPT. And there's nothing worse than using an efficiency tool and being less efficient, right?

Yeah, yeah. So I hear all kinds of people saying, like, oh, AI is bad. It doesn't write better than me. I'm like, you probably just don't know how to use it.

Yeah, it might even worse than you think. It's good. So I agree. I think people will get there. And I don't know about you, but my philosophy is that, yeah, the early adopter curve has obviously happened, but it's going to be such a late tale. So if you're watching this and you haven't really dabbled in AI, like, it's definitely not too late. It's not even too late a year from now, but at some point it.

Will be, oh, we're very early. Like us Internet nerds, we know about it, but you go offline into the town, you know, you talk to a real estate agent, you know, go to a coffee shop and talk to the business owner. They don't know nothing about this stuff. It's. We're way early.

Yeah, we're showing my 2 years ago chatgpt and she was like, I showed her it right after I could create an image for the first time or scan an image. I'm like, grandma, you can like the supermarket, have it, find it on Amazon or something. And she's like, whoa, crazy. It's super cool. So within, within your community, is there an AI specific channel or is AI embedded in all kind of the channels for media buyers?

So we actually don't have a community anymore. We. I'm not saying I, I invented community based courses or anything. We were just very early in it. We've been doing this for 10 years.

Yeah, yeah.

And so now there's so many communities that instead of trying to. Well, it's like the way I think about it, there's only so much help I can actually give you by replying to a Slack message or a Discord message. It's like you're like, how do I fix my Google Ads? And I'm like, all right, let me see if I can answer that in 140 characters. You know, so, yeah, so what we do now is every Thursday, like that's actually what I'm doing right after this call. Every Thursday at 3:00pm to 5:00pm, I do a live call. And that's our community. People show up live on Zoom. We share screens, we log into accounts. Like, we get real work done on the call instead of having like a Telegram group or a Discord group, because let everybody else have that. They can have the world's best Discord group. I'm going to have the world's best live call.

Yeah, I, I love that so much. And I'm going to end even five minutes early too. I'm going to try to pop in. I have another meeting, but pop in for a few minutes, see if I can get into that. That's, that's all what I like to do too. I like to jam with people in a custom environment. I have a quote, unquote, you know, quote internally that we've just never. It's just more fun to interact in that format too. So there's no AI, Justin, that's following you around, integrated into Zoom or Delphi or whatever. This isn't an AI avatar, I don't think. Although I would be so impressed if Justin was like, hey, yeah, I've been on another Call you like you like Justin number two that I sent.

What should I say next, sir?

So that would be a great for podcast. Maybe I'll steal it.

Well, there is an AI Justin. This is, this is an interesting talk actually. So it's not me, I'm real. But that's what an AI would say, right? No, no. So what we do is so like yes, Thursdays from 3pm to 5pm you can come on and you can get me actually helping you. But what if you live in Australia and that's a hard time zone to get to? Or what if you have a problem at 9pm at night, you know, on a Tuesday? Yeah.

Would say Justin's smart enough where you should set your alarm and get up in the middle of the night. However, I bet he has a solution.

Didn't come out of my mouth. You know, there's been times when I've stayed up till 1am for calls, you know, when it was important to me, I did what it was needed. But anyways, what we did is I trained an AI chatbot on all my course material. That way if you have a question at 11pm on a Tuesday night when I'm not around, it's the next best thing, right? And so you can ask it. And it's trained on all my course materials, trained on over 50 hours of my best stuff. And it's really good. It's 90% as good at me at answering questions. Or you can wait till Thursday and come on live.

So for 50 hours of content, you can get to basically a 90%. I'm always curious about how much content it might need before it's an effective bot.

I forget how many lessons it is. There's like, I don't know, we gave it a lot of content.

I 50 hours to get 90% efficiency. Anyone watching this like that makes me, I'm gonna probably do that later tonight. And because the time I'm saved, I just thought that Verizon was, was like bigger. I would need hundreds of hours perhaps maybe for it to start working.

No, it might not even be 50 hours. There may be 35 or 40 hours of content. You know, but we, we took all that content and we transcribed it all. We loaded it up into the chat bot as each individual. That's another thing. Don't take all the content. Write context Windows machines have memory. If you put all million words in one document, it's really hard to find. So instead we had a bunch of little documents for each class. That way it could also say, oh, the answer to Your question is found in video three where we talk about this. But here's the answer. You know, that way we could reference the lesson and give the answer.

And this should work because it's a. People watching this are marketers. But is there some sort of weighted attribution on the most recent stuff that you put in or does it not care?

I. No, I just. I'm just trying to answer the question effectively.

Me too. Right. I don't want to over complicate. It's just literally you if it's a call with me and you want me out of it because you're like, Nick's advice was kind of whack. But I really liked what I said. My advice out of this transcript and just. Have you.

Yeah, I think so. It'll work that way.

Okay. That. Yeah. And I don't know either. I. But that's going to be probably my 50 hours of content. I'm just. Well, I would love that. I'd like to do this and get some of this benefit, you know, in addition to. And hopefully. It's like, now that I say that now it sounds like I'm stealing your ip. I'm. That's not what I'm insinuating, everyone. But needless to say, like, I want it to be my tone. And I was like, will it bite me if I bring in this conversation? And it dilutes my tone and style?

Well, so to get your voice tone and style, what I did is I just took a bunch of my emails and social medias and things like that, put it all in the Google Doc, exported as a PDF, then I uploaded that PDF to ChatGPT and I said, describe my voice, tone and style in one paragraph. Because if you don't say one paragraph, it's going to give you this long explanation about everything you need to know about your inflection and all that. It's like, all right, I don't need all that right. I need one paragraph at that point.

Just so easy.

So now that I have this one paragraph description of my voice, tone and style in my prompts, I can always say and write this in this voice, tone and style. And then boom, I drop in my little. My voice. And now things are written like me.

So is that an app reply and it links to a project or how does that work? How does it know how to do that?

You can, you know. So I'll. It depends on how. Sometimes I'll upload it as a file in the project. Sometimes I'll put it right in the. In the. The context. A lot of these Claude and ChatGPT in your account settings, they have custom instructions. And so you can just put your voice right in your custom instructions and then everything will be written in your voice. Or what I do is I just say, when I ask you to write in my voice, use this. And so if I'm writing something or I'll usually do a bunch of research and outline and then I'll say, okay, now write this in the voice, style and tone of Justin Brooke. And it knows to grab that and put it in there.

Interesting. So, so critique this then. This is how you know you're a ghostwriter without telling someone you're ghostwriter. Would you do it like this then? Where you have custom CTs for the people you're writing for. Okay. But they would be under projects or how would that work?

Yeah, if I were a Ghostwriter, I would 100% have projects for each one. Each project would have like their customer Persona and empathy map as a file. I would have their voice as a file. I would have instructions on how I want them to write.

Cool. Okay, perfect. So for you guys, that's basically. Yeah, like user voice, brand and like brand voice and some sort of like competitive intel or objective or like cheat mo like that you give it where to go about the strategy. Right. That's cool. That's a great way to think about it. That'll save me so much time.

Couple tips if you're going to do that also include never use EM dashes. Right. Because that's like the whole, you know, nobody likes to see EM dashes. I also tell it you never use more than two emojis and they should be used to try and express the emotion that you are writing about.

Right.

So I don't want to.

Yeah, don't be a lazy copywriter.

Right. Yeah. And then the next thing that I'll say, I'll say purposefully misspell two words, making it look like the writer was just typing too fast.

That's a hack.

Yeah. So that you know, because people will be reading it and they're like, oh, I think this is AI. No, no, no. He used gonna instead of going to. Or he used to. Yeah.

My pet peeve is much more. You'll understand this being a copywriter. Bad reframing where it's. I see it all the time on captions. This isn't a war. This is a loon shot to the moon. Like, it's just trying to make some sort of really profound reframe. And I was like, Whoa. No, it doesn't have to be, like. Doesn't have to be old way, new way every single time.

Absolutely.

But I digress. I see that all the time. It seems like that's in its base coding or something right now.

And then. Yeah, just, you know, one. One other thing I'd love to get across is if you are using ChatGPT, you know, you only have 20 bucks and you can't afford any other things. I would tell you in the morning the first time you use it, and then new chat windows, that's always when it's going to be its smartest because the more you use it, the more they have to kind of throttle you a little bit, because otherwise you're costing them more and more money.

I didn't know that.

Yeah, yeah. So that's why sometimes you're using it all day and it was like, wait, this thing was cooking earlier? But now it's like, it's giving me all this hallucinations and things like that. So new chat windows and, you know, first time in the morning, like, your first five uses are going to be, like, the best you're going to get.

That's such. That's really great advice because I'm a refiner, so I refine myself to death, like, at the worst. Yeah. Thank you, man. Thank you. I know you got to get going to your call. I'm going to try to hop on for a few minutes. This was an amazing, like, experience where I just got to learn. You guys go check out his call. We'll drop the link below. You do it every third event you said, right? Three to five?

Yeah, every Thursday at 3pm Eastern. Yeah.

You guys will see me in there. Justin's the absolute man. Thank you, man.

All right, man.

Sweden. That was fun.

Yeah, no problem. Hey, when you're doing these calls, your chair is crushing your audio.

Yeah. Okay. I mean, I'm gonna have to go with an AI audio solution. I have a new chair actually, literally right here. So I'm gonna open it now.

Okay, awesome. So what I like to use for the audio solution is there's something called Adobe Podcast. Like, just look up Adobe Audio Enhancement or whatever. I throw my audio in there and it comes out Chef's kiss Perfection. Like, you had a studio.

Thank you, man. Yeah, I'm new to this, so, like, it's weird being talent. I don't like it at all. Like, I know you've been both sides too. Yeah. Yeah.

Okay. Awesome, man. I do gotta run. I'll see you later. I'll make sure that this thing finishes uploading.

Okay? Thank you.

Bye. Yeah, see ya.