AI Prompting for Writing: How to Get High-Quality Output

Happy 2026. Hey, how's this for a New Year's resolution? How about you Master AI marketing? The AI Business Society gives you live training with experts who are getting real results. No hype, no fluff, just proven workflows that work. And you've got access to an incredible library of training from many of the incredible guests that you've heard on this show. So secure your AI transformation by visiting socialmediaexaminer.com AI.

Welcome to the AI Explored podcast, helping you put AI to work. And now, here's your host, Michael Stelzner.

Hello, hello, hello. Thank you so much for joining me for the AI Explored podcast brought to you by Social Media Examiner. I'm your host, Michael Stelzner, and this is the podcast for marketers, creators and business owners who want to know how to use AI. Today, we're going to be joined by Moritz Crem, and we're going to explore AI prompting for writing. And we're going to go really deep into how to get AI to do really, really good writing. If you write anything and you're thinking about using AI to assist you in the process, I think you're going to love today's interview. Also, you can find me on Facebook and X. I'm at Stelzner on Facebook and Ike Stelzner on X. And if you're new to this podcast, be sure to follow this show on your favorite podcast app by clicking that little follow button so you don't miss any of our future content. Let's transition over to this week's interview with Moritz Krem, helping you simplify your AI journey.

Here is this week's expert guide.

Today, I am very excited to be joined by Moritz Crem. If you don't know who Moritz is, he's one of the most advanced AI prompting experts that I've ever come across. He's the founder of the Prompt Warrior, which is an email newsletter designed to help creators, marketers and entrepreneurs excel with AI tools.

And.

And his AI Growth kit includes complete templates and prompts for AI. Moritz, welcome to the show. How you doing today?

Thanks. Very good. Thanks for having me on.

I'm super excited to have you here. Today, Moritz and I are going to explore prompting techniques for getting quality output from ChatGPT. Claude, whatever kind of technology that you're using now, before we go there, I would love to hear your backstory. How the heck did you get into AI? Start wherever you want to start.

Cool. Yeah. So I grew up in Singapore and studied electrical engineering in Germany. And, yeah, I got really into entrepreneurship during, and then later, also after college, tried a bunch of things. I built an iOS app, grew various social media accounts, did an agency, did some crypto and none of them really worked out. And at some point I need money, so I just went and got a job. And yeah, I worked in consulting for a while, then switched into product management at a software company. And I was never really happy. I always wanted to do my own thing. And the day ChatGPT came out, that really changed a lot for me. So I was working on a big project at work, I remember that. And was kind of having lunch with a friend and my friend told me about this new app that just came out and we started playing around with it a bit and we basically realized that in just a few minutes it finished a project we'd been working on, which took us months of work, and finished it with a pretty high accuracy and did a lot of the work for us. So after that I got really fascinated by it. I started using it more and more for work, just to do all kinds of things like brainstorming, writing product specs, summaries and so and so on. And yeah, eventually I started writing online about ChatGPT that started taking off and now I write about AI and prompting full time.

So when you say you started writing about AI online, tell us a little bit more. Was it specifically on Twitter, slash accent you were writing, or were you blogging on your own website? Tell us a little bit about that.

I started on Twitter, so I was reading a lot on Twitter as well, just to kind of find out about the various prompting techniques and how to use the tool better. I was consuming content for probably 3, 4 months, so since November when it came out. And then sometime around April 2023, I started creating content myself too. I just basically wrote about how I was using it, what I was learning while using it, posted a lot of prompts, and yeah, people quite liked it. So Twitter took off pretty quickly. And then I think a month in or so I started my newsletter and that resonated quite well too.

So when did you start the newsletter and kind of, what was the catalyst? Because here you are, you're working for a software company and perhaps you're still doing that today, I don't know. But you've got this full time gig and you're a software developer, you discover ChatGPT, you put it to the test, presumably with some code. Is that what you were using it to help develop code or was it something else? Just out of curiosity?

I was actually in product management So I wasn't writing code myself. I was more, you know, doing the, like, research, aligning stakeholders, and then kind of presenting the results stuff. So I was using it a lot for that. And I think ChatGPT is really. It was really awesome for a product manager because you could do your research, which just took you ages before you could do that in seconds. And then as a product manager, you had to write a lot, too. So ChatGPT was perfect for that, for, you know, writing my summaries of meetings or, like, the product specs, or just things that you had to present to other stakeholders.

You launched this newsletter called the Prompt Warrior. And when did you launch that and why did you decide to do that?

To be honest, back then I just saw a lot of people talking about how if you grow an audience, you need to have an owned audience too, and that you should do it as soon as possible. So I. I pretty much created that domain from. From day one. And then I think a month in, or. Or maybe maybe even less of. Of creating Twitter content, I started to get people onto my list as well. And then I just started writing blog posts for it and kind of grew organically from there.

And one of the things that I think you. You come up with these really creative and detailed use cases for prompting. Where does that come from? Because I think you have an approach to prompting that maybe most people don't have.

I, most of the time, just try to solve my own problem. So, of course, I've read a lot about prompt engineering, so I really went into a deep rabbit hole at the beginning. So I kind of know about the different techniques, and I write about that too sometimes. But very often I have a certain problem in my business, which is very largely about creating content, and then the other part of it is kind of strategic stuff. But, yeah, when I encounter a problem, I try to use AI for it and I try to create a prompt for it, and then at some point I write about it.

This is really fascinating for anybody who's wanting to grow an audience. Is that what Moritz is doing is he is basically experimenting with prompts? Right. And the ones that you really feel are really good, they become educational content for you. Right. And you post about them in presumably your newsletter, but also on your blog and also on X slash Twitter. And have you found that that's a way that has grown your newsletter and your reputation?

Yeah, it's my. My newsletter was. I only grew it organically, basically just through plugging the link in my Twitter threads. That's how it grew Organically. And then I did experiment a little bit with ads at some point, but then I decided that it doesn't make too much sense currently.

And your newsletter is about 20,000 subscribers, is that right?

Yeah, it's around 20 now.

Excellent. All right. What I love about the work that you're doing is that you are kind of on the front line in a pretty substantial way, putting prompts to the test. And I believe you've experimented pretty heavily with multiple tools, right? Like ChatGPT and Claude are the two that you mainly focus on, is that correct?

Yeah, in the beginning it was just chatgpt pretty much. And then there were a bunch of other AI tools coming out, and, you know, I was. I was testing them all, but ChatGPT was definitely the dominant one. And then at some point, Claude got really good. And right now, for me, Claude is actually the dominant one because it's just a little bit better than ChatGPT at writing in a more human tone. And so, yeah, it's currently my preferred one. Also, on the rankings that are done online, Claude 3 overtook ChatGPT for a while. I think now, since the new ChatGPT model is out, that's again ranked a little bit better. I expect it to just always be this. Like, when a new model comes out, it beats the other one, and that just changes which models I use most.

Of the time for those that are new, that are listening to using AI in their business and in their work. Why does prompting matter so much when it comes to utilizing these tools? What do people need to understand about that? Why is that critical?

So, first, maybe what are prompts? So prompts are basically the way that we communicate with an AI chatbot, right? It's. It's what we type in as an instruction when we're talking to a chatbot, like Chatpedia or Cloud. And I think it's also important to distinguish between two different types of interacting with the AI, because when people talk about prompting, it's often not quite clear which one of these they're referring to. So one is kind of just talking to the chatbot. So I just call it chatting. Basically, in this use case, the actual art of, or the craft of prompting doesn't matter too much because it's just. You're just having a conversation. It's just like talking to a human. And this is good for just like asking simple questions, going back and forth and kind of exploring a topic. So this is the first use case. The second one is the actual prompting, which is what I mostly do. This is Usually when you want a very repeatable outcome. So it's like delegating a specific task to your virtual assistant or to your team member, where you just want something specific that comes out of it. So for example, a specific blog post, a video script, or a research report. And so in the second case, the way that you instruct has to be in a certain way in order for this repeatable outcome to be good. And here it's where the craft of prompting becomes more important. Yeah, I think it's also useful to think of LLMs, so large language models, which are the underlying technology of chatbots, as next word predictors. The most simple version of the next word predictor you probably know is if you take out your phone and you have your keyboard and then it kind of suggests you like these three words, which are just a statistical prediction of the best next word kind of. And basically LLMs are the same thing, just at a much larger scale. This may be a bit of a simplistic explanation, but I think it's useful to think about it in this way. So, yeah, I think prompting is extremely important, is an extremely important skill to have for the future, because AI will just be. It'll be everywhere. And most things that we interact with will have AI in it in some way or the other. And so prompting is just the language that we use to speak to it.

Well, and I think that, and you can correct me if I'm wrong, but from my experience, the reason why prompting is so important is because the output you get from an AI system is extremely dependent on the prompt. Like just altering a few words can completely alter the output. Have you found that to be true?

Yeah, exactly. And that's why I mentioned next word predictor, because whatever you enter is going to influence what next word kind of is predicted. And so what you input will influence the output. And sometimes if you just change one word in the input, then you'll get a quite drastic change in the output.

That's perfect. And I think it's really important for everybody to understand, especially if you've had a less than desirable experience. Right. Because a lot of people have tried out AI and they don't like the response and they just assume it's not very good yet they hear other people talking about how. How awesome it is. And the reason why a lot of people say it's awesome is because they've grasped what more it is talking about and what we're about to talk about here, which is it's really, really important to get the prompt dialed in because when you get the prompt dialed in, all of a sudden you get something that's really, really useful for your unique situations. Right? Let's explore some of your tips on getting started with prompting to generate content. What is one is one of your first tips.

Okay, so my first tip is to be very clear and specific with your prompts. So you can think of AI as a super knowledgeable, almost genius intern, but it doesn't really have any judgment for what's good and bad, and it's kind of also lazy and just does whatever you say without questioning it. And so when you want something done, then you have to just be very clear and specific with what exactly you want. I'll give you one example. When you say, write a blog post about my morning routine. So let's say you actually set this as a instruction to your actual human intern. Then there are a lot of nuances that go into the sentence, Right? You might have assumed a lot of things when saying this to your intern that you didn't really think of. For example, who would this blog post be addressed to? The human intern probably knows this because you've worked together and you just kind of know this, or how long should the blog post be? It's like you maybe read similar blog posts, similar newsletters, and it's always in a certain length. So you kind of assume that it would be in the same length and the same thing for other things like tone or the structure of it and so on. So these things really need to be. You should put them very specifically into the prompt. So for writing, for example, there are a couple things that are good to include. So one is kind of, you start with the goal of this blog post, or so you want the AI to write a blog post, and it should be about this and this. And then an audience is a good thing to include. Then what's really good to include is also writing guidelines to kind of define the tone. And then what I think is the most important part, actually, and that's what makes the prompt very specific, is to include a kind of template. And a template is just basically, it defines almost for each line what the AI should be writing there. So if you put a template in that you know is proven to be good, then it's almost like foolproof for the. For the AI to write something that you totally don't want or is totally unexpected.

Okay, there's a lot that you threw out here, and I'm just going to go over some of it. First of all, you said be clear and specific, and you said start with the goal goal. So if I'm a marketer, for example, which I am, my goal might be to persuade somebody. Right. But if I'm a blogger, my goal might be to entertain somebody. Right. So I like the idea of what the end goal for me is. But when you said start the goal, did you also mean what the goal for the AI is, which is to output an article, or did you mean the goal that the writer is hoping to achieve when someone reads the article? I'm curious on that.

I was referring more the goal for the AI. So at a very high level, the goal is to write the blog post.

I see.

So that's like kind of the first instruction set. But yeah, you can obviously go down this chain and define the goals, what the audience should then be doing after they read this, or what they should be feeling, and so on. So you can definitely go down this ladder and make it more specific. But then I think you also have to be careful because let's say, for example, you put in the goal that you want it to be high converting, which is, I think, something that people often say, and they want their blog post to be high converting. But that will very likely just make this blog post very marketing and salesy and just sound very off. So this one line can be a little bit dangerous in this case.

Well, in that case, it might be better for you to take the output that you love and then just have it slightly modify it down the road. Right. And we're going to get into iteration a little bit here. Writing guidelines. What kind of guidelines do we need to give the AI? Specifically, when it comes to writing guidelines, are we just talking about the length of the output or what are the kind of guidelines do you recommend?

That's part of it, but also the style and the tone are very important because you may have noticed with ChatGPT, its default writing style sounds very robotic and it often sounds a bit too formal. And especially if you write digital marketing assets, it sounds very salesy and.

And repetitive too.

Yeah, exactly. And so what I usually put as writing guidelines is to make it sound a bit more casual, conversational, just more human. And then what I actually have is like, I think it's a 10 points, bullet points set of, of guidelines which are like, your language should be punchy. You should often ask rhetorical question, you know, questions that, that flow into a sentence, like does that sound good? Or the good news with like a question mark that you should break long sentences apart, things like that.

Very interesting. This is getting a lot More nuanced and almost is a little bit of an art form, you know, than I think a lot of people realize. Okay, we're going to come back to this template thing you talked about in a little bit, but let's talk about your next tip, which you and I were prepping and we were talking about context. So talk to me about the importance of providing context to the AI and how that's different than some of the things we've been talking about.

Yeah, yeah. So I think it's definitely related. So the way I often also think about AI is kind of this, this huge cloud of knowledge. And it knows a lot. It has a lot of knowledge in there, but obviously it doesn't know everything. It only knows the things that it's been trained on. So, for example, it wouldn't know what my particular preferences are. It doesn't know what business and product I'm working on, what target audience I have, maybe which methods I've tried already in the past. So your job is to really inject this information because with this information it can produce better outputs. Yeah, common ones are what I've already mentioned, like things about my background, things about the product or business. I'm working on the target audience also. And what I recommend is to take these things and kind of pre write them and just save them as text chunks. And then you can just always add them into your prompt, into the sections where you need them. One other way is to also just save them in the custom Instructions feature in ChatGPT. And it's kind of just pre saved every time you start a new chat.

Okay, first of all, what are custom instructions? Because not everybody knows what that means. And then secondly, if you have an example of one you use and you could dig it up and tell us, that'd be awesome. But let's start with what are custom instructions.

So custom instructions is a feature that ChatGPT rolled out, or OpenAI rolled out sometime last year. It basically lets you write a few things that are saved and that are used every time for context when you start a new chat. So you can put things in there, like my background, so I can put my entire backstory in there so that it knows a bit more about me and where I grew up. I can put in there the description of the product that I have or the business that I have, and also like information about my target audience.

For example, is this something that is stored in like the settings or something like that inside of ChatGPT? So we have to go in and.

Look for it yeah, exactly. You go into settings and then you just write it in there and then you can. There's a toggle where you can say, use it for every new chat, or you can also turn it off and then it won't be used.

And obviously, if we're using claude, we're going to just include this equivalent in a prompt, is what I'm hearing you say. Is that correct? Yeah, we'll just be writing up this kind of context paragraph. Do you have one handy that you use in a lot of your prompts that you could read us, or is it really long, or can you just give us a little sample?

I can, yeah.

While you're looking for that, what I love about what we're talking about here is that. And I've. I've said this to some of the other people that I've interviewed, this is just like training a new employee, right? Like, you're giving them as much information as possible so that they're able to do their job with excellence. Because if you don't give them any insight, and especially with AI, they're just going to make lots of leaps and assumptions and you're going to say, this is not relevant. But when we follow your model, we're giving them so much insight that we allow them to narrow the window down into which they're creating content. Correct?

Yeah, yeah, exactly.

Did you find something that you can read us?

I found my old ChatGPT custom instruction set. Since I've been using Claude a lot, these might be a bit outdated, but I just. I kept it quite simple. I just said I'm a solopreneur and content creator on Twitter, and then described my content and kind of what it helps my audience do and describe my audience a little bit more.

It's like a paragraph or something like that.

Yeah, just like a few lines, so not that much. And then there's also a question which says, how would you like ChatGPT to respond? And you have to be a little bit careful with the custom instructions because. Or in my experience at least, I was turning them off quite a few times because again, we're going back to this. Whatever comes in as the input will influence the output. So what I really like to do is to experiment a lot and see how my input changes the output. And then if I have something in the custom instructions that are just always there, then I don't really like that because then it kind of always puts the output into a certain direction. And so I prefer to have it just manually or just use it manually.

Yeah, because I would imagine the more you use ChatGPT, the more you use it for other things and you don't want that custom instruction to necessarily impact that other stuff.

Stuff.

But it sounds like something like this, what we're talking about, some sort of a background on, on who it is that the, the AI chatbot is doing the work for and the company and the audience is really important to kind of allow it to zoom in. Quick question. How many podcast episodes have you recorded? How many webinars have you hosted? How many client calls or team meetings have happened that are now just sitting there in your files? Here's the big unlock. Those aren't just old recordings. They're full of insights about your audience, your brand voice and your customers actual objections. You just need to know how to extract them. That's exactly what AI educator Erica Stanley is teaching AI Business Society members on January 8th. By the end of this training, you'll have a repeatable workflow that turns every past recording into a content and data powerhouse, one that keeps paying dividends long after the mic's been turned off. This training is exclusive to AI Business Society members. But here's the good news. You can join us right now by going to socialmediaexaminer.com AI and start transforming your old content into new marketing gold. SocialMediaExaminer.com AI to join right now. Earlier you talked about a template. Tell us a little bit about what that is and let's explore that a little bit. Like what do you mean by providing a template to the AI?

I just think of it as a very specific instruction for writing a blog post, for example. It can also be a tweet, a LinkedIn post, and it would really define line by line what should go in there. So it can be a template that is a little bit more general. Then it would be like the first line, write about a problem. The reader has the second line, agitate this problem and then write about the solution. So that would be a more general template. And then a more specific template would be something like this tool. And then tool is kind of like in a bracket so that it can be replaced. This tool is so much better than this tool. And then this tool, that tool would be. So this way you can kind of, if you want to write a thread about one tool beating the other tool, then you can use this template and just plug it in. And then you can use it for many different topics. But it's kind of always outputting something very similar.

So as a writer myself, and I think about My style of writing, oftentimes I will open with a shocking or compelling part of a story from the arc of the story. And then what I'll do is I'll go ahead and I'll explain more of a story, and then I will introduce whatever I want to introduce, like, for example, the new podcast. And then I will have some sort of a lesson or learning point. And then it will end with some call to action, encouraging people to comment. And I would imagine if I really studied my style of writing, I could break it down in a way where I could tell it, hey, this is the way this should be structured. And then what I can do is I can input, like, the story is about X is what I'm hearing you say, right? Could I take my writing and paste it in there and then give it this instruction and have it make it better? Do you understand what I'm asking? Instead of just like having it write something original from scratch, can I have it take something and recraft it in this model?

Yeah, and I actually do that quite a lot. So you can, for example, take just a blog post that you've written in the past and copy that and just say, okay, write me a new blog post that is just like this one, but about this new topic. And then it will just take the old blog post and exchange a few things so that it matches this new topic. The other way that makes it a little bit better, in my opinion, is that you take the original blog post and run it through a prompt which templatizes it.

That's where I was going, that's cool. So you can tell it, hey, write me a template based on the style of writing in this blog post, and then it will come back with something that it's interpreting that you can use to guide it. Is that what I'm hearing you say? That's so cool.

Yeah, exactly. And then you have this template and you can reuse it to write new topics, and they'll always be in a very similar style and therefore also be of pretty good quality. Because, you know, this is like your style and this is how you've written your blog post in the past.

Let's talk about the process of iterating. Let's say we don't like what we got on the first go round. What are some things that we can do? Because I know so many people listening maybe have tried some of the things we're talking about, but they just aren't getting it the way they want it. So what is iterating and what can we do to kind of like, maybe run multiple variations and hopefully get what we're looking for?

Yeah, I think iterating is super important because as you said, a lot of people just try it once or twice and then kind of think the output is not good enough and then stop using it. And they should rather just have this mindset that they need to experiment a little bit and test it a few times until something. Something good comes out. So I, for example, if I do a prompt, I usually test at least 10 times. Whoa, whoa, whoa, whoa.

We gotta talk about that for a second. So when you say test 10 times, what do you mean? Do you mean you're actually running the prompt 10 different times and then you're looking at the variations and picking the best of the output? Is that what I'm hearing?

You say that also? I'm doing that at a little bit of a later stage. But what I would, for example, do with the template creator prompt that I talked about earlier. So in the initial version of that, it would just be create a template from this post, and then I would run that once and kind of look at the output and see what was good and what's bad. And then I would make an iteration of that prompt. So I would then say, okay, step one, I would define the instructions a little bit better. I'll say, in step one, you asked me to paste some content, and then your job is to transform that content, so just to make it a little, little bit clearer. And then I would take out some of the mistakes that I noticed it doing in the first round and try to piece that into the prompt also. So, for example, what I noticed it doing in the beginning was that it was very much kind of thinking that the content of the original post is important and that you need to make the template very similar to that original post, but that makes it kind of unusable as a template. And so I then wrote, note that the actual topic of the pasted content is not relevant. Your job is to generalize the template so that it can be reused by plugging in any other topic? And then it did a bit better, but it made a few mistakes again. And then I would say, for example, give it some examples, and that, like, slowly then makes it better.

Okay, so what I think I hear you say is, like, take any instruction that you want to create a template out of so that this is a reusable. Because this is what we've been hypothesizing from the beginning is prompting, is creating some sort of a system. So that you can have repeatable outcomes happen over and over again. And if you can get this prompt stuff figured out and dialed in, then you can have it every time, create a high quality piece of output for you. And in your case, we've really zoomed in on this template that we've been talking about, and what you're saying is it's. You got to iterate, you got to keep experimenting until you really dial in that template so that it produces the kind of output that you want to produce. But I would imagine that not everyone is going to do a template, right? Some people are going to, but I know that you suggest that they do do that, but let's just say that they've created some sort of prompt with the hope of getting the output that they want, but they don't like it. So is the process of iterating looking for what's wrong with it, and then telling the AI, try it again, but don't do this or do more of this? Is that kind of what we're talking about when we talk about iteration?

Yeah, that's what I mean with it. So just tweak it in a way that it will match what you want. And, and I would also add that sometimes you keep trying and you keep tweaking, and it still doesn't really output something that you want after a lot of trying. And then I think one of the things that you can still do is just try a few different AI models. So this was happening a lot with ChatGPT, that the writing was just not perfect. Even though the prompts were pretty good, it was just not great writing. And if you use that same prompt, then in Claude, then the writing is a lot better. And I think this will be more and more the case in the future where certain models will be better at certain things that you sometimes just need to try a different model for your task.

Excellent. And then when we're iterating, is this also where we might want to experiment with replacing just a few words? Is this where we have to get a little creative? You know, like if. If we say that your job is to be a creative writer, and then we realize it thinks a creative writer needs to be all flowery language. We might want to remove the word creative and say, like professional writer or nonfiction writer or blog writer. I mean, is this where we want to try subbing out some of those kind of words to kind of like, see if it varies the output?

Yeah, exactly. And I guess sometimes it's a bit hard to identify which words are causing this. But yeah, your, your example is really good. Or often if you use a role and you kind of say you are a copywriter, which is what people often do when they, when they start these prompts. It's kind of like a recommended prompting technique. But the problem here is that this word copywriter just turns it into a kind of very marketing salesy thing. So, you know, you have to be a little bit careful with that too. One thing I also really like to use is what I call the lever technique. So I think I invented that and don't have the perfect name for it yet. But it's basically when you get a certain output, you want to go back to the chatbot and kind of rate it. Let me start. You're writing an email, and the email that comes out is very formal, because that's just like the default of the chatbot to write a formal email. And then you say, hey, on a scale of 1 to 10, 10 being very formal and 1 being very casual, how would you rate this email now? And then it would maybe come back and say it's a seven. Then you can go and say, okay, now turn it into a three to make it a bit more casual. Because what usually people will do when they get a very formal email, they will say, okay, make it more casual. But then what happens is it gets way too casual. Yeah, exactly. It swings far, far too much on the spectrum. So this way you can fine tune it.

Okay, you talked about examples. Talk to us about how we ought to consider including examples and what kind of examples we should include, because you have a unique process for this which includes not just good examples, but bad examples. So talk to us a little bit about that.

So I think this is probably one of the most important unlocks to get really good results of using examples. And it also ties a bit into the iteration process. Yeah, AI is quite similar to a human in this regard in that it's much easier for it to understand something if it has examples. So as you iterate your prompt, you always want to bring in the results that it has given you and bring it back into the prompt to improve it. If we take the template creator prompt again right now, how my template creator prompt is looking, it basically has this initial instruction part, and then I have five examples of transformations. I would say example one of the original piece of content and the example of this content being transformed into a template and kind of put that into a prompt and say, this was a really good transformation and give like four more of these examples and say, these are really good transformations. This is how you should continue doing things.

When you say transformation, can you clarify what you mean by that?

So a transformation in this case is taking the original piece of content and turning it into a template. That's one transformation. So I would just paste the original and the end result and put both of them into the prompt again and say, here, you did a good job. Keep doing a good job. Like this.

So we're. We're talking. These are mega prompts then, right? If you're talking about five examples of, like, original output translated into something better, or here's examples of work that you've done based on past prompts that, that I loved. And here's one I didn't love by adding the one that I didn't think was very good. Why, why is that important? Does that help the AI kind of understand where to go with all this?

It has the same effect, I think, as providing a good example. It's just basically like saying, okay, do this, don't do this. So one simple way I use this, for example, is I just say, don't use puffy words like game changer, master skyrocket, and so on, which AI is known to use. So you just say, like, don't use these. And then it doesn't. So it's a good negative example in this case.

Okay, cool. Now, when we're providing examples, is it true that we could just upload PDFs instead of actually inputting it into the prompt? Could we say, like, these five PDFs that are attached are all great examples of. Maybe they were or were not written by AI, but they could be models that the AI could follow.

Yeah, absolutely. So you can have it in any kind of document and upload it along with your instruction set. That works just as well.

Okay, looking into the future, I'm curious, where do you see AI and content heading? Obviously, you're creating content with the assistance of AI, and it's an important part of your business, but you're probably on the front line seeing things that the rest of us do not see. Where do you see it all going?

Yeah, I think it's a really interesting question, and I, I do think about it a lot. So I think there are right now kind of two extremes on social media. So one side is saying, you know, these are like the classic marketers. They're basically saying you shouldn't use AI and stay away from it because it's. Right now it's still really bad. Whenever you use it and create writing with it, it just lacks the quality. So you should just learn this stuff yourself and stay away from AI. And then I think on the other side there are these people that are developing the tools that are really automating the entire content creation. And as a result then they're partly flooding the Internet with low quality AI posts. So if you think about some of the SEO blog post tools out there that just you can just generate a bunch of blog posts just to Rank higher in SEO or like these LinkedIn reply generators where you can just click a button and it'll generate some generic reply and they feel really soulless and just like re regurgitated. And so I think both sides are right and wrong somehow. So one, the people that say don't use AI, I think the problem is that you'll just be less creative and slower than other people that do use AI if you go down this route. And on the other side of the extreme, the problem is that currently AI doesn't really have unique perspectives. So the quality of the models is just not there yet that it can create the entire content for you and have it be high value and good quality. So I think what you need to do right now is to have a balanced system. And what I often say is to use AI as your editor and not as the creator. And so I think right now and in the short term future, the human really needs to stay in control of this process and they should see AI just as a tool that they're using to be faster and take some shortcuts in their content creation. But they're still the central part of the content creation. And I think what's going to happen long term is that as the models get better and better, humans will shift more into this role of just supervising these systems and just providing the right inputs here and there and maybe also the right designed prompts in the right way. And yeah, just supervising the system and creating content this way. And I do think there will eventually be a kind of crossover where people don't really care if something was AI generated or not anymore. Like right now people are very angry about this when they see AI content. But I think in the future people won't really care because you just want content to provide you value, right? And that's at the end of the day, what you actually want. So I think the interesting part is actually that we see part of this future already playing out. So I like this quote, the future is already here. It's just not evenly distributed because we're seeing part of this with the AI influencers on Instagram already and the AI girlfriends. If you look at these images and read the comments, most of the people, they either don't care that this image was AI generated or or they can't really tell the difference. And yeah, I think the same thing will pretty much happen with text content and also video content.

Moritz Kram, thank you first of all for this fascinating exploration of the importance of prompting in AI. If people want to learn more about you, it sounds like Twitter is your preferred platform. So where do they find you on Twitter X and then also where do they find out more about your newsletter and all the great stuff you've got going on?

Yeah, on Twitter you can just find me moritzkremp. So the handle is just one word together and my newsletter is called the promptwarrior.com Moritz, thank you so much for.

Sharing your insights with us today.

Thanks for having me. It was fun.

Hey, if you missed anything, we took all the notes for you over@social mediaexaminer.com a9. Be sure to follow this show on your favorite podcast app. And would you tell a friend about the show? If you want, you can tag me. I'm Elsner on Facebook and ikestelsner on X. And do check out our other shows, the Social Media Marketing Podcast, the Marketing Agency show, and the Social Media Marketing Talk show. This brings us to the end of yet another episode of the AI Explored Podcast. I'm your host. Michael Stelzner will be back with you next week. I hope you make the best out of your day and may AI continue to help you be more successful.

The AI Explored podcast is a production of Social Media Examiner.

AI Business World early bird pricing ends on January 16th. You can save $300 on your tickets. Register at AI BusinessWorld Live today.