Generic to Genius: Crafting AI Prompts for Optimal Results

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 Stelzer, and this is the podcast for marketers, creators, and business owners who want to know how to use AI. Today, I'm going to be joined by Jonathan Mast, and we're going to explore how to craft AI prompts that really optimize the output of artificial intelligence systems. If you've been prompting AIs and you've not been happy with the results, well, today's interview is for you. I think you're going to find a lot of really cool ideas, especially at the end of the interview when we talk about something absolutely fascinating. By the way, if you are new to this podcast, be sure to follow this show so you don't miss any of our future content. Let's transition over to this week's interview with Jonathan Mast.

Helping you simplify your AI journey. Here is this week's expert guide.

Today, I am very excited to be joined by Jonathan Mast. If you don't know who Jonathan is, he is an AI marketing strategist and founder of White Beard Strategies, a consultancy that helps businesses and nonprofits save time, deliver more value, and become more profitable using AI. His course is AI Prompting Mastery, and his newsletter is AI Prompting Insider. Jonathan, welcome to the show. How you doing today?

Well, thank you so much for having me. I am excited to be here, having a great day, and as you know from our pre call, I love talking about AI. So I'm excited.

Super awesome to have you. Today, Jonathan and I are going to explore an AI prompting framework that he's developed that will help anybody maximize their results. Now, before we go there, Jonathan, I would love to hear a little bit about your journey. Tell us about what you did before AI and then tell us kind of how, you know, you came into this world of artificial intelligence.

No, I would be happy to. Well, I started in the digital marketing field and I've been involved in that since I sold my first website in 1995. So I like to have earned all this white hair. I've kind of come by naturally throughout that process and obviously much after 1995. I started using some other AI related tools like Jasper and things like that, and was always pretty impressed, but they never blew my mind, so to speak. And then when ChatGPT released its new interface in the fall of 2022, I started playing around with it and I was like, wow, this is truly an entirely different scenario than what we had before. And I started seeing some of the value of how it allowed me, especially as an entrepreneur with add, to get things done f to be able to whether that was ruining an idea because I found out an idea wasn't any good, which happens a lot, or whether it was actually being able to make traction and move forward quickly because I could actually move forward in it without having to wait, without having to spend a ton of time thinking. And I could ideate and brainstorm, which to me are two of the things I love to do with AI really quickly.

So you started messing around with it. It sounds like shortly after it was released, somewhere along the line you decided, all right, I want to actually put a flag in the ground and build a business around this. Tell us a little about that journey. Yeah.

So again, being an entrepreneur, having been involved in digital marketing, AI just, it was amazing to me how good of content and results I could get out of it. When I started learning how to communicate with it properly, it gave me decent responses without even communicating well. But when I started digging into it, I started realizing, wow, this is actually an incredibly powerful tool. And for me, because of the fact that it's so easy for me to get distracted and see bright shiny object and go, whoa, what was that? AI allowed me again to ideate faster and to be able to come up with things and ultimately, I believe to create better deliverables for my clients. Whether that be me responding to emails, whether that mean helping them with a digital marketing campaign, whether that was creating the right images for something that we're doing, or even, as I've begun recently, working with some educators, helping them figure out how to leverage these tools within education so that we can ensure that our students when they graduate are going to understand how to use the tools effectively because we know that's something employers want. Again, I kind of dove in head first. You asked why did I decide to kind of plant my flag? Having been in the digital world for a long time, digital agency world, I Love that world. But I saw just huge opportunity with AI as far as for not only agencies, but businesses in general. And one of the things that's become a real passion for me, Michael, is actually also talking to their teams and helping their teams realize that this is not a Terminator AI that's here to ruin our lives. In fact, it's an amazing tool that allows us to do our job better. And for me and my background, it allows us to serve people more fully. In other words, deliver more value to them. And that gets me really excited because when I can save time and deliver more value to my customers, that's like, you know, a win win scenario for everybody involved. And AI is fun on top of it. So it's like, why not have some fun?

You mentioned that you have attention deficit challenges, and I'm just curious, how has AI because there's plenty of people that are neurodiverse that are listening to this, like, I have dyslexia. How is it that AI somehow has assisted you in that struggle?

I think for me, the big difference, Michael, has been the fact that with add, again, I'm wide open to being distracted by bright, shiny objects. And sometimes it's even just dull objects that are different that I want to pay attention to. And so when I get started on a project, when I have an idea, and like many entrepreneurs, I have many of them every day, oftentimes I'm not able to execute far enough on them to determine if they're viable, simply because I get distracted before something happens. And it's. I guess it's a curse. But on the other hand, ADD allows me to hyper focus. And that with AI is a tremendous combination because I can hyper focus, ask questions, get very fast responses. And as I mentioned, one of my favorite things to use AI for is for ideation and brainstorming. And it allows me to validate some of these ideas that I have. Should these be something I invest time in or not? Because not every idea I have is great. And I want to be able to basically ruin that idea or get it off the table if it's bad. So I don't spend any more time on it. If it's good and it's got legs and maybe I can add value and again, improve, improve the lives of those that I'm serving, then I want to be able to dig into it further. And AI helps me, particularly in those areas, do that and get to a point that I can then make a decision, move forward or put it aside. And that's been a Huge value for me.

We're here to talk about prompting today, specifically with AI systems like ChatGPT and Claude and others. For those that are listening that do not understand the importance of prompting, why is this something that's really critical, like help everybody, especially the new people who are listening that don't understand prompting. What is it about this that's so important that we're about to talk about today?

Well, without trying to overdo what AI is, anytime you're working on building a relationship, and I'm going to use a coworker as an example, you need to communicate with them. And I really look at AI as a coworker that's working on my team that I not necessarily need to build a relationship with, but I need to be able to communicate effectively. And that's really all prompting is. Prompting is this, you know, especially we hear this word prompt engineering. It's just a term to mean communicating with AI. And I have found through the last year plus of writing literally thousands and thousands of prompts, that if we can learn how to communicate effectively, we can get amazing results. And that's where I came up with the prompt framework. It's four simple steps. We'll cover those here as we go along. But the key there is, it's learning how to communicate and communicate effectively. Because when we do that, just like with coworkers, we get better responses and we have a better relationship.

Yeah. And if I were to flip this on its head, if you've been experimenting with AI tools and you're not happy with the results, perhaps it may be the way that you're prompting the systems. And with some of the things we're going to be learning today, you can increase the likelihood that you get what you expect and hope out of these things and maybe even get surprised by them. Is that a fair statement?

Absolutely. It's one of the most common complaints I hear about AI is Jonathan. I tried it. It's no good. The results were junk. I didn't get back anything relevant, or it just made a bunch of stuff up. And all of those can be true statements if you're not prompting it well, if you're not asking it the right questions in the right manner. The good news, Michael, is, you know, it's not a hard process, but it's like anything else. If we have a system that we follow, it allows us a huge upper hand in working with the AI models to get those tremendous results that we're looking.

Well, let's start with some of the biggest mistakes that you see people make when it comes to prompting. Let's just address some of those. Off the top of your head.

Off the top of my head. I would say a couple of them that immediately come to mind are one, what we call expecting perfection with one prompt. People expect that AI can read their mind, that's probably a second aspect of that, but that it can read their mind and in the process that it's going to give them the perfect response based on one reaction. Michael. That's like me going to a co worker and saying, hey, I need you to write me a memo. And they go about what? And I go, I'm not telling you. And well, to whom I'm not telling you. And what do you want me to cover? I'm not going to tell you. And walking away. And that in my mind is exactly what. We don't consciously do that, but subconsciously we're expecting that AI can read our mind and do all of that and it's just not the reality, it doesn't happen. So I think that's probably the biggest thing that we see happening. The other thing that I see regularly is that people don't want to provide it with enough information or in my case maybe don't know what information to provide it with. So they just let AI make assumptions. And we know what happens when people assume. The same happens when AI assumes it makes mistakes and sometimes it gets embarrassing in the process.

What about asking way too much in a single prompt? Let's talk about that a little bit too.

That's the other end of the spectrum and that happens as well where people ask 27 questions in one prompt. And you know, I compare that to going to my 17 year old son and giving him a list of five things to do. If I give him a list of two or three, he's going to get it. But if I give him a list of five or more, he's never going to remember everything that I told them. And AI really is the same way when we give it too much to do in one prompt or one communication, it's likely to fail as well because it's going to pick and choose the things that it thinks are most relevant. And much like my 17 year old, sometimes it's going to be wrong in what it does, but it's almost always going to forget if we give it too much in one prompt.

Perfect. I mean I've experimented probably not nearly as much as you have with prompting a lot. And I find that definitely like if you, especially with things like creating images, if you just you just like add one extra piece of detail in there, all of a sudden it can go nuts because it thinks that's the most important one and it is one of those little art and science kind of projects. So that's why I'm excited to dig in right now with you. So let's explore the very first step of your framework and maybe define what it is, why it matters, and then we can dig into some of it a little bit.

Well, the first step is really simple and it's simply telling AI who it's supposed to act like, what type of Persona we want it to take on. If we remember, Michael, that AI is this vast encyclopedia, probably the largest set of encyclopedic information we've ever had access to with almost instantaneous recall. It helps us to remember that if we can define it better and at least get it into the right, maybe the right building or the right book or the right section in the library, it then can do a much better job of giving us responses because it allows it to dive deeper. So simply starting by telling it how we want it to act, I want you to act as an expert consultant in a particular area, will do wonders to get you better results.

By the way, we didn't even talk about the name of your framework. What is the name? Do you have a name for your framework? I think you do, right?

We do. We call it the Perfect prompting framework just because I'm so humble. But we really do find it works well and we had a number of people in our AI group group help us with that and that's the name that they came up with and I didn't mind it. So we, we call it the Perfect Prompting framework.

Okay, so the very first part of the framework, and this is, I believe a four step process, is defining who you want it to act like. So tell us a little bit more about like this, like let's dive in deep on this with some examples and stuff.

I think, you know, the, the best example, or a simple example would be let's just say that we wanted to write a social media post for our business. If I want to write a social media post for my business, I want to start by telling ChatGPT I wanted to assume the role of an expert copywriter who is either award winning or proficient or whatever you want to say. It doesn't matter per se, but that is an expert at writing engaging social media posts. By just doing that, you are an expert copywriter who specializes in writing engaging social media posts. Now, the AI model, whether we're using ChatGPT or Gemini or Claude or any of the others immediately knows the route or the path that we're headed down. I now just need to give it some additional detail to get the best possible response.

One of the examples we talked about when we were prepping is, and I'll read it to you and you can respond to this. Act as an experienced digital marketing strategist and AI technology expert or in the AI technology world. Right. So with a background in content creation. So how creative can we get here with this? Should we be broad? Should we be specific? Because, like, when I think about your example of a social, let's say you said social marketing expert or expert copywriter. Like as a copywriter myself, I could say you're an expert copywriter who specializes in writing email copy for B2B. I mean, can we get as narrow as we want to get? What's your thoughts on that?

I think there's a balance, but yes, you can get as detailed as you want. The best results I find are when we get it in the right vicinity of where we're heading. In other words, I don't necessarily, if I use a city analogy, I don't necessarily have to get it to the right house on the right street, but if I can get it to the right part of town or even to the right street, it's going to help make sure that I find what I'm looking for a lot faster. And so you absolutely can dig in with more detail. But keep in mind that the deeper you dive on defining who AI is, the more specific it's going to respond from that perspective. So if I were to, in this example, say you're an expert digital marketer who specializes in, you know, writing compelling social media posts, and maybe we even go further and expand it for a particular industry. Now if I ask it something that doesn't fit with that, we may have a higher likelihood of confusing the AI. So it's probably the old 80, 20 rule that applies in so many areas of life. Michael, we want to get it close, but we don't necessarily have to make it perfectly precise.

Okay, so as I'm listening to you and we're starting with this first step in your prompting framework, which is to identify the role that you want the AI to essentially model. There's so many applications here, right? Like, let's say that you just got a contract from somebody and you wanted the AI to analyze it. How might you explain the role in that situation? Would you say something along the lines of you well, I'll let you answer that.

At that point I would probably say, you know, you're an expert contract attorney. I just recently closed an opportunity and I need you to review that for me.

Got it.

Something along those lines.

So you can be really generic if you want to. It's funny because I've seen so many different applications like you're an expert in travel planning or you're an expert, like whatever it is you want to do before the prompt is, you let it know that this is its domain expertise. Now let me ask you, does it matter how we word this? Should we say act like or should we say you are? What's your thoughts on that?

I don't think it ultimately matters a lot in the response, although the AI models will change their response slightly and what I mean by that, the quality of their response should be the same. But I have found if you say things like you are an expert travel advisor who specializes in helping business people plan trips to Mexico, you'll say, I went there. Right. It's more likely to come back and say, I'll be happy to act as such a person, but I am not one, because AI has enough filters that are being added into it that when you say things that it doesn't deem to be accurate, it's starting to tell us that. So I think you're going to get the same end result. But if you want to mitigate or minimize some of the hassle in between, I would definitely focus more on act like versus you are.

Got it. So the key take home here is act like a subject matter expert. Right. In this domain who maybe specializes in this particular thing. Right. Like you're a travel agent or a travel expert who specializes in traveling to Mexico City, for example. Right?

Correct.

If that's where you wanted to specifically have it go. And the reason why this is so important, just to reiterate, is because these AI models have a very broad education and when you give them a narrow focus like this, they're going to focus their results in that particular funnel almost. Is that correct?

Absolutely. So like you mentioned with if I mentioned Mexico City, now when I'm looking for that travel advice, it's going to focus on things that are related or around Mexico City. If I on the other hand, said I was going to go to, you know, Los Angeles, California, it would now focus on Los Angeles or things like that. And so that just helps define where we're going. Again, one of the things I want to encourage everybody is iterate and try this out, find out what works best for you? Because what works best for me, the model I know will work, but you'll find how you communicate what your style is, and ultimately that will lead to the best results and make it as easy as possible for you to do the prompts.

Perfect. Okay, let's transition over the next step in your framework.

So next, as we've told AI, you're supposed to act like this expert. Then the next thing I want to do is I need to act and remember the same thing I do. Michael, if you and I were collaborating on a project, now it's time to give you some context. I need to explain in general what I'm looking for you to do. So, going back to our example of the travel advisor planning a trip to Mexico City. Now, I want to say I want you to help me plan a three day excursion to Mexico City where I'll be meeting with clients, but I'll also be bringing my family. And so I'd like to identify some extracurricular activities that my family could do while we're in Mexico City.

Okay, this is all in the same prompt or is this the second prompt?

I mean, it's all in the same prompt. Yeah. So it's, you know, we've started with you're an expert travel advisor who specializes in business to business travel in Mexico City. Right. And then the next sentence is, I'm going there to visit some clients, I'm bringing my family, and I'd like you to help me identify some extracurricular activities that we could do as a family while we're in Mexico City.

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. As far as this context stage that we're in right now. There's more than we can do with this, obviously. So let's explore some of the other options beyond travel. Just help people understand, like the kinds of context and the depth of context that we need to provide, because that Mexico City one is pretty straightforward. But there might be some business applications that might be a little bit more complex.

Let's go back to the contract application you mentioned. Let's say we've just signed a new contract. We can at that point, after we've told it you're an expert contract attorney who specializes in contract reviews, then I might say, I've recently signed a contract, or I'm about to sign a contract with a new partner. I need to review the contract, and I'm looking for you to identify any aspects of the contract that need to be amended or that should be amended in order to, for example, mitigate my liability.

That's interesting. So how are we providing the contract then in this situation? Are we just attaching it or how's that work?

In most cases, you're either going to upload a file, such as a PDF or a spreadsheet, a Word Doc or something like that, or. Or in some cases, you can actually copy and paste the text in there. On a contract, I'd probably upload it as a PDF.

Okay, so in the case of the contract, I would imagine you could also ask it, hey, what's the worst case situation that could happen for me if I were to sign this contract? And it would probably tell you, well, these clauses aren't very friendly, right?

Absolutely. Well, and again, that's where the context becomes so important, because if I define that I'm looking to have it review the contract to help me mitigate my liability or mitigate my exposure, then that's the lens or the filter with which it's going to take a look at that contract. And that can be helpful. I still need to tell it what I want to do. And that's step three, which we'll get to in a minute. But the background is providing it context around whatever aspect we're looking for. So in the case of a contract, we may ultimately go through this multiple times with the AI looking at different aspects. We may talk about first, you know, what about potential exposures? And we may want to talk about financial liabilities. What about, you know, is there anything in the contract that's going to be a risk to my reputation or to my brand? And by defining that context of I want you to review the contract in the context of protecting my brand, it's then going to review it based upon that role. If I said I want you to review the contract based upon any potential financial liabilities I may incur, it's now going to have a different filter. Same contract, relatively same process, but a different filter by which it's going to analyze that information.

Okay, so the level of context that we provide here, help me understand what the options are. You already mentioned that you can upload a PDF. What are the things can we provide the AI for context in this case.

Where we're trying to do a simple prompt, I think the best thing to do is any documentation that's part of that. So again, in a contract, obviously we'd upload that contract. If we were talking travel to Mexico City, I might want to upload some brochures of some specific things that I was interested in and see how that comes about. So there's definitely different, different things I could upload there. If I was doing some data analysis, maybe I had a large spreadsheet and I wanted to analyze that data, I would upload that spreadsheet and maybe, let's just say, for example, there were guidelines I needed to follow while I was doing that data analysis. I can actually upload both the spreadsheet data and the PDF of the guidelines and have it review and just say, as part of that, that context, I need you to review this spreadsheet, but I need you to also review the guidelines in the attached PDF to make sure that we're following protocol or following the guidelines that have been laid out by whatever governing authority we may be working with.

I've also found, and I don't know if you've done this or not, but I've also found that I can paste in links to relative resources that it will go read. Have you experimented with this at all?

We have, and this works in many of the models, but it does not work in claude, which is one of the major models, because Claude doesn't have Internet access. But you can absolutely do that in ChatGPT. You can absolutely do that in Google Gemini, and they'll go out and they'll reference those websites. One thing to be aware of with that, Michael, is that some websites are blocking AI tools from viewing them. And so it's not a perfect science. It won't always work. Sometimes you'll get a message back that it can't view the data on that website. And there are some workarounds if that happens. But yes, it's great. For example, if you and I Were working on developing a tone and style parameter that we wanted to use as we were creating content. I could have it go out to Social Media examiner and go to your author page and say, you know, I want to help define a tone and style for Michael. And it could analyze that page and based upon that author information, and then maybe I'd give it three or four of the blogs you've written or a podcast. It would analyze all of that information based upon my parameters and say, okay, we've determined Michael's likely this personality type, and here's his tone and style. So it absolutely can infer a lot of information from those links that you provide.

I mean, I feel like Claude may have resolved the link thing, But I'm not 100% sure. But I do know that in situations where you can add a screenshot if I'm not mistaken, and you can correct me if I'm wrong, but the vision capabilities of these models can read the text inside of screenshots. Is that correct?

100%? No. In fact, it's one of the favorite things I love to tell people when I'm speaking, when I actually show the prompt framework and some other things, I'm like, this is the one you want to grab an image of and just upload it to ChatGPT or Claude or Gemini, and it will literally read the image and give you all of the text on that. That can be super helpful for that type of thing where we need basically optical character recognition. But also the vision capabilities are really amazing that I can take a picture of something and ask it to identify what it is, and it will search its data, understand that in most cases, and come back and give us relevant information about whatever topic that is. We can even go so far as to say, tell us what's happening in a photo, and maybe if it was a photo of a race or something like that, it'll define what it sees. And we can literally create an entire story based upon a photo by just starting by uploading it.

Okay, so so far, the first two parts of your framework are. Step number one is to provide the role. Right. And we talked about act like you're a blank. Right. Who specializes in blank. And then step number two is the context. Right. And in the context we're really talking about, here is some information, basically that might be very important for you to do what comes next. Right. And the next thing is the third step. Do I have that right?

So far, you're absolutely on track. Yep. And the third step is what most people would consider to be the prompt. So most people, if they were starting off with this, would say, I need extracurricular activities to do and mechanics Mexico City. That's all they'd say. Or I need you to look at this contract. That's all they'd say. They wouldn't give you relevant information surrounding that. And that's step three. Step three is your. The simplest part. It's your question. Whatever it is that you want that AI model to do in this particular response.

Okay, so talk a little bit about why putting these first two things before this question makes all the difference. Like help everybody understand that. Because a lot of people just go right to, here's my question. Right?

They do, and I've done it as well. And AI will give you a response. The one thing that's really important to remember here is that the developers of all of these AI models, obviously from a business standpoint, they want us to use the models because we then subscribe. And when we subscribe, they make money, and that's what they're in business for. As a result, they've tried to make these models as easy as possible to use. And that means that you can just ask a simple question like, give me activities to do in Mexico City, and it will give me activities to do. Whether or not they're appropriate for my family, whether or not they're in the right part of Mexico City, whether or not they're at the right time of day, none of that will be part of it because we haven't done it. But AI will make assumptions to do that. And again, that's been programmed in.

So let's talk about how to write a good question, because obviously, unless it doesn't matter. I mean, I'm just curious what your thoughts are on this step when we're asking the question.

There's a balance really through all steps of this, Michael. Obviously, the more time we take to put into it, likely the better response we're going to get. At. At the same point, I don't want people to get overwhelmed with the fact that, oh my gosh, I need to spend 10 minutes putting together a prompt in order to get a good response. They don't have to do that. What I find for most people works well is if they simply follow the act as an expert in. Here's the background information and here's my question. Just by doing that, they'll improve the responses they're getting out of AI by hundreds of perspectives. Just, it's a huge, huge difference across the board.

I totally am tracking with You. But I just am curious if there's any tips on how to ask the question that you've learned after having asked a million questions. You know what I'm saying?

Yes. I think being as direct as possible and remembering that we want to give it one thing to do, one, one question to answer here. I don't want to ask it for things to do, restaurants to eat at, flights to take, hotels to stay at. And oh, by the way, what are the street vendors I should be talking to? I don't want to ask all, all of those in one question because it's going to get mixed up. So when we're asking the question, we want to focus on that one thing that we want the AI to answer. And I want to encourage everybody that's watching and listening to remember that we can ask follow up questions. So it's not like we're limiting ourselves and going, wow, but I need to find out about flights and I can't do that. No, you can. But just make the question about your flights to Mexico City a separate question, a follow up question to what we're talking about or the street vendors or the hotels or any of that. One question per prompt will get you far and away the best results.

Love it. Okay, now we're getting into the ninja sauce stuff here. The fourth part of your framework is really exciting. So why don't you share a little bit about what that is, please, and then we'll dive in.

So when I identified this, this was probably one of those biggest aha moments that I've ever had with AI. And as I mentioned earlier, AI has been programmed to make assumptions and to fill in the blanks. And a lot of times I wouldn't know what information to give it. And that means that by default, even without knowingly doing so, I was making AI make some assumptions. And then I discovered the fourth step. And if you take nothing else, folks, remember this. After you've asked your question, give permission to the AI to ask you clarifying questions. And you do this as simply as typing in, Please ask me any clarifying questions that you need in order to give me the best possible response. That one tip will by orders of magnitude improve the responses because it will when it needs more information than ask you those questions. And when you respond and give it that, it's going to be that much more precise and relevant in its responses back to you.

It's fascinating because I use Claude and ChatGPT a lot and I was using Claude literally yesterday and I don't remember exactly what I was using it for, because I use it for a lot of different things, but it did unprompted ask me for clarifying questions. So my original prompt, and then it said, okay, here's four pieces of information, five actually, that I think would help me. And I answered the ones that I thought I could answer it. Have you started seeing it prompt you for clarifying questions? Because that's generally not the operating procedure for AI. Generally it just tries to give you the best output in one shoot. Right.

So I originally started seeing this with Perplexity, which is an AI based search engine, about a month ago. I have not had it happen to me in Claude, but in all candidness and maybe because I add that, you know, ask me any clarifying questions to virtually every prompt I put in, but I have not seen that. I would not be at all surprised though, that that becomes the norm because again, it does allow the AI models to respond so much more relevantly to that question. So I'm really glad to hear Claude doing it. Not surprised that they were the first ones of the major models to do so. And hopefully we'll see that with Gemini and with chatgpt as well.

Okay, so when we were prepping, I wrote down an example of something you said and I'm going to just read it out loud and then we can kind of dissect this. Conduct an interview with me to gather insights for creating a compelling social media post about why digital agencies need to embrace AI for marketing content creation, ask thought, provoke provoking questions that will help uncover the key benefits, challenges and future implications of AI adoption in digital marketing agencies. I didn't come up with that one. You did. So let's break it down because this is a more sophisticated ask right of the system, much more.

So the biggest difference here, and it is one of the quote, unquote superpowers of AI in my mind, is to ask it to interview you. Most people have never thought about the fact that we can literally ask AI to interview us to determine the information it needs to give us the best response. And that's really where this particular prompt starts at. It starts right off by saying, I need you to interview me. And then we go into greater depths of talking about why.

Okay, so this is slightly different than what we were talking about before. Right? So before we're just saying basically, ask any clarifying questions. You need to get the best response. Is that how you worded it? Is it that simple with step four?

Step four is really that simple. And I would still continue to add it onto this. It's a. I might say the interview request is a more advanced prompt, and maybe we were, you know, Mexico City is kind of a beginner prompt. The contract may be an intermediate, and the interview is probably more advanced style. But it's the same formula, it's the same four steps that we're going to follow in the framework. We're going to, again, tell it how we want it to act, tell it the context we need, ask it the question. And even if I'm asking it to interview me, I would still ask it to ask me clarifying questions. Because let's say, Michael, you and I were chatting, as we did prior to this call. There may be things that I want the AI model to know about the interview process that I forgot to tell it. And if while I'm prompting it, I'd say, you know, ask me those clarifying questions, it will likely, prior to the interview part, ask me for clarifying questions about the prompt itself or possibly even how I'd like it to set up the interview, things I'd like to cover, and things like that that will add more clarification to what it's doing.

Okay, I'm going to come back to the interview question in just a second. But if we ask it for clarifying questions and it specifically says, these are three questions I have for you, how ought we respond to those? Do we answer all three of them in one shoot? What's your thoughts on that?

Two things. One, you don't have to answer every question that it asks you, but by not answering, you will impact the quality of that response. If it asks you questions, I encourage you to answer them to the best of your ability. And if it asks you one question, answer one. If it asks you three, answer three. If it asks you seven, answer all seven, or answer those seven that you believe you can.

All in one shoot, though, do we just type it in as one big response or do we have to type it in as three separate responses? Like, any thoughts on that?

So generally, when it asks multiple questions, it will number them. So it'll say, I have some questions for you, number one, number two, number three, and on and on. When I respond, I then just give the same number. So I'll put a one dash and the answer to number one. Two dash, the answer to number two. Let's say I don't want to answer number three or I don't know how. One of the things that I found to be very effective is three. I'll just say, you advise me on this and I'll literally tell AI. I don't know, maybe I'm talking about audience identification, or maybe I'm talking about Mexico City, which I've not been to, and it asked me a question. I may say, I don't know, you tell me what's the best food in Mexico City type of thing, and then I'll continue until I've answered or completed all of them. If I decide not to answer, I can just say continue, and then it will proceed and fill in the blanks, make assumptions in order to give us that information that it doesn't have.

Love it. Okay, back to the interview stuff. The situation where you're telling it that, hey, you are this expert and all the other stuff we talked about, right? Like we started with the context and all this other stuff leading up to this interview. Interview me, when it interviews you, what is that experience like? Because most people have never been interviewed by the AI. What does that look like?

It's actually a great process in my mind. So I do. I add one step to it, and that's because again, I don't want to get a list of a hundred questions and have to remember all of them. So I ask it in that prompt to please ask me no more than three questions at a time, and then to repeat the process in the questions until it has everything answered that it needs. I have found in that process, it's a really simple, easy to follow process because you've got as much time as you want to respond. Unlike our podcast here, if you ask me a question where it seemed really weird if I stood here for two minutes and didn't respond, with AI, I can take as long as I want to respond to its question so I can ponder that, I can think about it, I can go grab a coffee break in the middle, whatever I want, and come back to that chat and it's still going to work just fine. So it's a very simple process. It's in my mind, it makes the entire information gathering process easier because now I don't have to prep upfront to think what questions should I answer or what information should I give to AI to get the best response. I can simply say, here's my goal and objective, and based upon that goal and objective, ask me whatever questions you need in order to give me the best response.

So when you're asking it to interview you, let's just say you'd want it to answer one question at a time. Is it going to ask you the question? You give the response and it's going to ask the next question. And how do you know when the interview is done? I mean, I'm curious about that.

So as a general rule, I find that AI models are going to ask between about five and 15 questions. And when you ask it to interview you, if you're not sure, I keep answering until it's done. I've never had a time that I just went on forever and kept asking me question after question after question. It always found a point at which it thought it had enough information. Again, generally in the first 15 questions. If you get frustrated though, if you're going through this process and you're like, okay, Jonathan, this was the worst advice ever. I don't want to sit here and answer these questions. Then just tell it to continue and it will do so without waiting for you to respond to anything else. And it'll take whatever you've given it and utilize that and ignore the rest.

I see. So you could just say, hey, stop interviewing me and continue with what I've.

Asked for at any point in time. Yeah, and it'll never get offended at you. So it's. You're perfectly safe in doing that.

I want to talk a little bit about input audio versus the written word. Talk to me a little bit about how maybe it could interview you and you could speak to it versus typing. I mean, have you experimented with that?

I have. So I live in a state where they have hands free driving laws, so. But I've used my phone many times and what I do is I just start a voice conversation with it where I'm literally speaking to Chachi PT and it's then responding in audio form back to me. And I find it to be an incredible way to use drive time because if I've got a brainstorm or something, I want to plan. Just yesterday on the way in, I was, I was planning a new challenge for my Facebook group and I wanted to brainstorm about it. I've got about a 12 minute drive and when I was leaving the driveway, I pressed the chat button and I explained to it, hey, I want to develop a new challenge for my Facebook group and because it has some memory, it knew what I was talking about. And we literally had a back and forth conversation for probably close to 20 minutes because I sat in the driveway here at my office for probably five to 10 minutes after I arrived finishing that conversation. But the neat thing, like you said, it was so easy because I didn't have to type. I literally could converse with it. Just the way you and I are conversing here. And it would then literally speak back to me, not like my 17 year old, but like a consultant giving me relevant information and then letting me respond. And we truly had a conversation.

This is where it gets kind of interesting for anybody who has an iPhone, you know, you might have the hey, Siri feature activated and I apologize if I just triggered anybody's iPhone. But with ChatGPT, oh, my. My phone just said, sorry, I don't understand. Hopefully that didn't pick up. But with ChatGPT, it doesn't have a keyword trigger kind of thing. Right. So you have to kind of start it before you drive. I mean, give any advice on how you start the conversation. And does it stop at a certain point and then you have to wait till you're at a stoplight to reactivate it. Do you understand what I'm asking?

I think I do. So, yes, in order to activate it, you do need to press a button on the bottom of the screen. I'm using my mobile phone now. There's a headphone type button and you literally press that and it goes into audio mode and then you start speaking. They have a number of advancements coming out. There's one. I don't want to say flaw, but one slightly frustrating part to it right now. It waits for silence when you're talking. So if you pause too long between, it's going to just take off and start running. It's going to go, oh, Jonathan, must have been done. And it's going to take off and run. The good news, if it does that when it's done, giving you the response, you can always go, okay, let's go back to that. I didn't finish what I was saying.

Can you train it to say, hey, wait till you hear the words I'm done, and then pause. Have you tried that yet?

I've never tried that. I'm going to have to do that today because I've never thought of that. That's a great idea.

Yeah.

I've been on the other end. OpenAI has announced they have an updated version coming out. In fact, they demonstrated it about 60 days ago, where you can now interrupt it. Just like I could interrupt. You might be rude, but it's not considered rude to AI. So you can actually begin to interrupt like, no, I wasn't done. And it will say, oh, I'm sorry. They have not released that though, yet. They did demo it though. It looks amazingly cool and I can't wait. But I'm going to try that to tell it to maybe give me a keyword to proceed or something like that before it does. I've never thought of that, but it should work just fine.

Jonathan, thank you so much for sharing your insights on prompting. If people want to discover more about you, what's your preferred social platform for them to connect with you? And if they want to learn more about your business, where do you want to send them?

Well, it's probably one in the same if they just go to Jonathan Mast.com that's M A S T.com forward/linktree I've got links to all my social platforms there ways to email me and all that. It's the simplest way to get in touch with me. I probably spend the most time on Facebook. But you're welcome to connect with me on any of the major platforms on all of them at least daily.

Jonathan, thank you so much for sharing your insights with us today.

Well, thank you for having me and for giving me the opportunity. I hope this added some value to some people and helps them get some better responses.

Hey, if you missed anything, we took all the notes for you over@socialmediaexaminer.com a15. Also, be sure to follow this show on your favorite podcast app. For example, with the Apple podcast there's a little arrow button that you have to click on and then you'll get all the latest episodes of the podcast. Also, be sure to share this show with a friend. If you want, you can tag me on Stelzner on Facebook and ikestelsner on X. And do check out our other shows, the Social Media Marketing Podcast and the Social Media Marketing Talk Show. This brings us to the end of the AI Explored Podcast. I'm your host Michael Stelzner. I'll be back with you next week. I hope you make the best out of your day and may AI help you become 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 Business World Live today.