10 | AI SEO and Prompt Engineering with Tim Hanson

In this episode of the AI Marketer's Playbook, host Audrey Chia chats with Tim Hanson, founder of Penfriend, a leading AI-driven SEO content platform. Tim, with over ten years in SEO, dives into the challenges and opportunities of integrating AI into SEO practices. He shares his journey, the critical role of human elements in AI content, and his strategic approach to prompt engineering. Discover how to create high-quality, repeatable SEO content and the importance of a structured process in prompt writing. 

This episode is a must-listen for anyone looking to harness AI for SEO success.

Join my weekly Newsletter: https://lp.closewithcopy.co/welcome

Transcript

riverside_audrey_chia _ jun 18, 2024 001_tim_hanson podcast : 0:00

Hello and welcome back to the AI Marketer's Playbook, where we cover actionable frameworks to help you leverage AI in business. I am Audrey Chia, your host, and today I have with me Tim Henson, the founder of Penfriend, a cutting edge AI powered SEO content solution. Now Tim has over a decade of experience in SEO and in the agency world working with over 30 over clients as the head of SEO. Today, he runs one of the best SEO AI platforms in the market, and is also a well sought after prompt engineering expert. Welcome to the show Tim! Hey, thanks very much for having me on. I really appreciate the invite. Awesome. So I would love to know a bit more about your background and how did you even get started? Because I came from a copywriting background and I had to figure out the whole world of AI as well. Yeah. So, I mean, you must've done the thing where as marketers, we all get that email that tells us whether or not we still have a job and how difficult it's going to be. You must've got the one that we get in like January that tells you where, like how difficult it's going to be. So then. Got the thing, the email telling me that the robots are taking my job last year. And, I, I think the thing for me, like working in SEO is that's not scary because I get that three times a year. It's called a Google update. And I have to relearn my entire job again. So this one coming along of like, Oh no, the robots are going to take my job is, A little bit scarier than the majority of those like updates that we get, but it was It was enough that I Took it as a challenge more than anything was to really realize a lot of people are going to be very scared by this and when there's when there's fear in the marketplace as opportunities It is true. fun fact, right? So my client was the one who sent me the link of chat GPT and he sent me the email and he said to me, Audrey, you are going to be replaced. That was actually, yes. It really happened. That was my pivotal moment. I think he meant it as a joke, but sort of also sort of meant it. So yeah, that was pretty interesting. And how do you actually, Started picking up the skill sets for prompting. So we kind of, I was sort of playing around with ChatGPT. I want to say in like August, August, September time of 2022, it was just one of those things that I was kind of seeing and I was playing around with it a little bit in regards to like how I actually. Fully like, okay, cool. This is the thing I need to pick this skill up. I need to learn as much as I possibly can. was grabbing as many people's courses as I could, just absorbing as much of that knowledge as I could, like quickly as possible and just writing an enormous amount of problems, constantly writing. And the, the, the thing that I had more than anyone else, I think, was. Everyone was trying to write prompts for single use. I, I need to write this one page, blah, blah, blah, like all this kind of stuff. In SEO and content creation, we have so many processes that you do over and over and over again, that the thing that was most important to me was how do I write a prompt that gives me the same output every time? Yep. How do I be that specific with a prompt and The thing that I realized pretty early on was, Do you remember everyone was like, Oh my God, prompts don't work. Like it doesn't, it doesn't do its job. It's actually really bad at doing this thing. And I'm over here and I'm like, It's actually pretty good at doing the thing. But you're not asking it. You're asking it. one thing and you're expecting something else and never telling it the bridge in between, which is massively unfair. Like if you have a friend that you have a set of expectations for and you never tell them the expectations and you get mad at that friend, that's just not fair, right? So the thing that I realized pretty quickly was most people's prompts were just not in depth enough. They weren't, Covering specifics well enough and that comes from you write a prompt you get an answer and then you go Oh, that is a bad answer. I'm just gonna rewrite the prompt the prop. The problem is not ChatGPT. The problem is you. The problem is your prompt your prompt your prompt is bad I definitely see that as well, like even, cause I'm a conversion copywriter, right? So I have seen people try to get AI to create a landing page, but what they are not doing is they are not leading AI and guiding AI to get to that dream outcome. So that is what I have seen a difference between like a good prompt versus something that is like, do it for me instead of like, okay, let me help you to get there. Is that something you've also experienced? Okay. Yeah, so we did a really interesting thing with Penfriend when we were first writing all of the, when I was first putting all of the prompts together for the first version of Penfriend. And this is, this is like a year ago at this point, like, this isn't even the one that we originally launched with. But, Inga and I sat down and worked out how many decisions as a human do you need to make? to go from no content existing to published blog. And we worked out that there was at least 27 separate decisions that a human needs to make, like, between those two points. And for each one of those decisions, you're essentially wearing a different hat. You have a different, set of criteria that you need to take into account for. You have a different output that you want, you have different information that you're pulling in, let alone a different like persona that you need to embody to make the right decisions. And we looked at that and went, okay, cool. That's 25 separate prompts. Because I have to tell it a different thing and for it to take into account different, different ideas. And yet you have people on LinkedIn that are like, I can write a really good blog in one prompt. And it's like, no, you can't. We've just proven that I need like 25 prompts to write a good blog. And you're like, you can do it in one, it's like, you can't, it's not good. It looks okay, maybe, but if you actually have trained writers, you have people that care about it, like you've cherry picked a good output. You don't have something that's repeatable every single time. But then the difference there is like, I can write one prompt, and then the other one is like, I have 25 prompts to write. Yeah, it's true. And I don't think people actually see how much thought goes behind getting good quality copy with AI. So I've noticed that there are two camps, right? One camp is like, AI can't write. And then the other, the other camp is like AI writes everything, but there's no human in between. And then there are people sort of like us who are basically seeing that there's, how do we get it to sound more human, but also realizing that it's not as straightforward, as what most people think. and that there are actually multiple steps to get to the end result. Yeah. Yeah. And it's, I think the thing that really made The biggest difference to me in terms of how do I get good quality output from prompting is the things that I'm trying to get it to prompt are jobs that I have manually done myself several hundred times, if not a couple thousand times depending on what the specific job is. So, I am already approaching the prompt with an SOP, a standard operating procedure, like literally a checklist of how I do that job in mind before I write the prompt. So I have all of the rules on how I get a good output. All I'm doing is just transferring that into a prompt with a good persona, an expected output, a format of output, like those kind of things on the other side. And If I don't give it all of those things, one, I'm not going to get a good output, but two, if I've not already done that process several hundred times, how do I know what a good output even looks like? Yes. And the beautiful part that I always say is, even though writers of, you know, content writers, conversion copywriters, brand copywriters, they, they may be afraid of AI, but the thing is, actually writers are the best people to prompt For good copy, because you already know what good copy and good content looks like and you understand your own internal processes to get there. But I'm also curious to ask, right, for someone who is new to SEO, what kinds of steps do you actually take into consideration? Because for me, I'm not a SEO writer, so I have actually no idea what you need to consider during this process. So Let's dig a little bit more specifically from an SEO point of view as to what's the job that we're trying to We're trying to get away. Is it more so how do we get the content to rank or how do we get people to click from? The search results because from my side that to like to two different things So if it was a new business owner trying to go on SEO What would their priority be and what will you get them to focus on? Okay, cool. So the first thing that I would ask them to do is give me like what is one topic that they get asked about a lot and that they can answer with without really thinking about it because they're just an expert. They just know this thing really well. So let's for example take, I don't know why, but for whatever reason I've got in my head is like pool cleaning. Okay. You could take something like pool cleaning and something that they probably get asked about a lot is like, how do I look after it after it's been done? How do I, like what's the upkeep of that look like? They probably know that really well. They know the products to buy, they know if you need one of those cool little like underwater Roombas that can clean your pool, like if you need any covering, if there's any specific plants that ruin the filtration system, like they have a level of expertise that they know really well. But it's not as top of funnel as just like, clean my pool. How do I clean my pool? So it's further down, like there's, it's, it's a little bit more grounded in, in something that makes a, is more specific to them as an expert. And then what I would do is take that from them personally. What I would do is take some, have some kind of interview with them, do it through zoom, take the transcript and then get it to pull out all of the insights into specifics. That I would then put into some kind of prompt that gives me like 10 blogs that I should write on that thing. And I personally, because I own the tool, would put them all through Penfriend and then get all of those blogs out and put them on someone's site. 10 is probably not going to be enough. You probably want 15 to 20 or so. So I would look to see if there's any places where there are subtopics, bits where we can go a little bit deeper. I would then look to try and put as many of those insights that I've gotten from that interview, from that, that, that person into the content. So it pushes the human first element of it. The thing that I'm looking to do more than anything is establish what we would call topical authority, which is I have 30 blogs on topic A. And topic a alone, therefore, I'm a, I'm an expert in that space. Only an expert would have could talk about that topic for 30 blocks. Yeah. Right. And the problem that I see that most companies have, especially early on, is they treat their blog, they treat their content strategy as a magazine. So like a magazine will have like a summer issue and an autumn issue. And all of these, like, every issue is different. There's a different focus. That is a big problem when you're writing SEO articles, when you're writing, like, content to rank, because let's, for example, say the arbitrary number is you need ten articles to rank for topic A. Now, if you've got four or five topics, let's call it four for sake of argument, And you rotate between each, each one of those every week. So you're publishing four blogs a month, sorry, one every week. Topic A, B, C, D. Let's say you need ten for you to start ranking, for Google to start paying attention to you. You're gonna be at month ten before Google picks anything up. Or, you could be halfway through month three. If you do Topik A, Topik A, Topik A, Topik A, and just do that for 10 articles and get them all, all done. And the nice thing about that, and this is the thing that I've found so many times, is you, you have content that supports each other. Yeah. That is in a position where you can very easily direct somebody from one blog to another, to another, and they can, they can, If it grabs them and they want more information, you can send them to another article that you have. The hardest thing in the world is to get somebody to land on your blog. The last thing you want to do is introduce them to a topic, an idea, a concept that you then don't also answer and they go back to Google and they go to someone else's blog. You've lost them. They've gone. a great example that I had of this was I remember seeing a video where somebody had A pair of rollerblades with three wheels instead of four. I skated a ton as a kid. Like I've never seen a pair of rollerblades with three wheels. So I'm doing a search, like three wheeled versus four wheeled roller skates. And there's this big blog that's like telling me what all of the benefits are. Like the better over bumps, they're a lot more cruisier. And I'm like, that's the kind of, that's the kind of thing I want now. And then they didn't give me a link to the shop for me to go and buy them. They just lost a customer. So I went back to Google and bought them from a different website. Yeah, yeah. And that's definitely also the importance of having a clear call to action, like what I always tell my clients on any kind of landing page or even a blog page, right? But I also want to find out, right, because we've been talking about AI so much, but question, how much of the human element is important when it comes to AI generated SEO articles? So the way that we do things with PenFriend is you can publish articles directly from PenFriend and they do well, like they do really well. But the thing that I've found when it comes to like a conversion point of view is you need to go and add the human element to it because if you don't add the human element to it, what's stopping literally anybody else from buying the same tool as you. And putting in the same keyword and effectively getting the same article out from the other side. There's no differentiation. You're essentially just creating immediate competition for yourself. That human element is the thing that gets people to buy. It's the thing that gets people to resonate with the content a lot more. But mostly it's the thing that gives you unique insights and unique stats and figures. and connection from one person to another that nobody else has. And it's the thing that Google is rewarding really well in the search results, like at the minute. So if you don't add them in and you're stuck on page two, that's probably why you're stuck on page two. Interesting. And from a SEO perspective, right, what would you consider, you know, a human element in a blog? Because let's say for a Landing page, right? That human element could be the founder's story because that is really unique to them. What what happens for a long form article Ryan law put this really really well, which was AI will write you should Humans will write I I could no sorry humans will write I would people who know what they're talking about will write, I could do this. So, I should is like, just generic, goalless, I don't really care about your outcome, like, advice. I would, I, yeah, I, I would is, if I have this problem now, this is what I would do to solve it. And now you're showcasing, like, your own personal expertise, the things of, like, how you would do it. Again, I could actually is now showcasing a comparison of different ideas. I could do this if this was the situation. I could also do this if the situation was different. And the point there that is really only experts and people who know what they're doing would have different outputs depending on what the input is. So if you're writing like a how to if you're writing that something that is like I'm gonna show you how to solve this specific problem Yeah, and you you you can take into account. This is where it's gone wrong with for me before This is how different people have come to me with this problem. And this is how I would solve it You could do this you could do this you could do this. You're gonna have much more nuanced And human forward content that is human to human because you're taking into account their different situations versus just you should go and do this. This is how you solve X problem and it sits in a box and it's in a vacuum and nothing else will ever, ever impact that problem that you're trying to solve. Interesting. And here's a question, right? Would we be able to prompt for that writing style because technically you could provide a couple of different options by prompting for it. I have no idea. Yeah, that's really interesting because I think you could. I think there is a point where you could prompt for it. Yeah, I'm doing the thing. I think you could. I think the, the problem that you would have is that is a prompt that at least initially Is somewhat single use, right? Because I'm trying to solve problem A under circumstances A, B, C, right? One, two, three. You would have to put in what is the problem that you're trying to solve and what are the circumstances for each one of those. And at that point, it's like, just write the thing by yourself. Because this really isn't a prompt that I feel like you could use Multiple times. So at least for me, I wouldn't, I wouldn't want to write that prompt because it's not actually saving any time for me. If, on the other hand, you had author bio and expertise and some way that you could work out what are the different situations that somebody might be trying to solve that problem in and how they are different. And you could, you put those things in with some kind of like vector storage, some kind of like on the fly prompt building, then yes, I think you could. But then like, I'm looking at it from a point of view, which is how do I make it a repeatable prompt? Not how do I just do this once? Yes. And yeah, that repeatability is a big thing. Very important when it comes to building a system that works right and speaking of prompt, right? I think it would be amazing if you could share maybe a couple of prompts with us or the way you would structure Certain prompts so that our audience also can learn a little something from your amazing brain Yeah, totally, totally down to do this. Let me just pull up. So I've done a couple of workshops on how I build prompts before. So I'm just going to walk through it, like exactly how I build like massive prompts and then go through the like iteration process of it. Yeah, that's amazing. So. For those of you who cannot see what's on screen, Tim can walk us through, and then you can just talk through the different prompts and how you actually think about it. Because I think a lot of people see prompts out there, but they don't understand the thought process behind crafting the prompt. And maybe you could, you could share a little bit of like perspective on how you build prompts and what is your thought process when you're doing so. Yeah, totally. More than happy to. Okay, cool. So. I came across this, like, whole term of mega prompt. Feels like it must have been, like, over a year and a half ago at this point. Of people writing these ridiculously long prompts, and not really explaining, kind of, how they do it, and why they do it in the way that they do it. So, I tried as much as I could to break it apart into something that was replicatable, And if it wasn't 100 percent replicatable, I could at least understand which part of the prompt wasn't doing its job. Because each part of the prompt kind of has a very specific thing that it's supposed to do. So you'll have seen kind of everywhere, let me make this really big, everybody always says like, you have to give it, like, who is it pretending to be? at the time when I was writing this, the way that I would write this would have been like, You are a A. I. persona, blah, blah, blah. The way that I do this now is this bit here is, would be more along the lines of something like this, where it's adopt the role of A. And, if you've got a couple of roles that you feel like, You want that overlapping point of a Venn diagram, then it's you, you just change that to roles. And then you give like whatever the role two would be. So in this instance, instead of it being like bottom of funnel software, like list pages, it could be bottom of funnel software list pages and the role of a, like product specialist or something along those lines. So you're combining two different roles and giving it like this funnel. additional knowledge from both roles, right? Yeah, yeah. So, and then the rest of this is just what do I expect it to do? what kind of, give it an outline of what that person's job is. This can be super in depth or you can literally just keep it to this first sentence of these are the roles that I want you to embody. really depends on how specific and how repeatable you need a prompt to be. But this is the bit that I found. Once you've got everything else good in a really good position that you like playing around with the persona is usually the bit that helps you to take an output that you have and see how it might differ from a different point of view. Interesting. So your persona is basically the lever that you move in order to change the output to a bigger extent. Yeah, it's like, imagine it like this. I have a problem. I have a task I'm trying to solve. The persona is the angle and the lens that I'm using to look at that problem. So I might want to do this in regards to like, how, how different would it be if I changed this to top a funnel? How different would I, would it be if instead of it being software list pages, it was camping gear pages. And I don't know, like, it's, it's, I've found that this is the bit that you can do the smallest changes and have the biggest output, like different, because it's, it's as if it's a completely different person performing the same task. Got it. And I, I wanted to ask, so like, based on your experience prompting, right, do you feel like a longer persona with more extensive details work better than a shorter one? that's a good question. So sometimes some of the personas that we've got for some of the pen friend prompts are so specific. It's like that's not even that's nobody's human job ever. And it's just me trying to get like as tight as I possibly can onto a very specific, very, very specific repeatable output. some of them are pretty good. Pretty, pretty loose in all fairness. The thing that I've found is if you want a crazy repeatable output, have a more specific persona, but you are doing it at the expense of how big, how much of that output is that you can get out of it. So if you've got a really specific persona, You will get a repeatable output, but you might only get a couple of sentences that are exactly perfect that you want every time. If you have less of a persona, you're kind of giving up the exact amount that is repeatable for a bigger output on the other side. So that's one of the reasons why with Penfriend, when we write blog articles, we do it heading by heading. We don't write the entire blog in one go. We, we write from an H2 to an H2 as one single prompt, and then each H2 is effectively then pulled through its own specific series of prompts for it to write the output because I needed it to be loose enough to write good copy, but tight enough for it to be stuff that we can then sell and is a good output on the other side. Yeah, yeah, definitely. It's always like that balance, right? and there are so many different ways to prompt, but even for myself, when I prompt, sometimes I. intentionally break it down into super small steps. so that I can get really high quality output. Whereas when you combine everything into one prompt and you realize that the output is totally different from what you expected. So, yeah, yeah, true, true. So then after we've done a persona, I then give it just a top level view of what the task is. Like, what are we just trying to achieve? And then from there steps. Of what the task is now this you can literally break this down into like step one step two step three step four for this it's It's kind of a combination of what's going on in the persona and the steps here So it works pretty well for this prompt specifically and then constraints so I think the thing that a lot of people do is they get mad because oh it went off the rails and it did this And it did this and it did this and it's like you never told it not to Yep You'd never said, don't, don't do this. Don't use these words, like keep within this box. So I like to have some level of constraints, context kind of going on. And then the goal. So the goal isn't something I think you have to put in every time, but I find that it's nice to give it a goal because this is kind of bringing it back to the persona. Like, what is this person that you've invented here? What is that purpose? Yeah. Really. Like, how would you give them a purpose? What is the thing that they're actually trying to achieve? And I found that, like, having some level of a goal, their purpose, gets you a lot closer to what is me as the person writing the prompt, what is my expectation on the other side? Because if I don't tell it what my expectation is and you write something, now we're misaligned and I'm going to be mad, but that's unfair because I never told you what my goal was. So, Telling you what the goal is, is usually a good thing to do. And then formatting, like how do you want it to deliver you all of that information? I think what most people forget is also the formatting part. Like, even when I ask for add copy, for example, the formatting makes a huge difference. for example, if I ask for a certain output with specific character constraint, Then when I get the copy for my ads, it's a lot tighter. whereas if you don't do that, often what ChatGPT does, it gives you a paragraph of content for ad copy, which you cannot use. So that's also something I've noticed. Yeah. yeah, the formatting is the bit that I think is also a lot of fun to play around with because sometimes you see it, you see what you get out of ChatGPT, Anthopic, whatever you're using. And you're like, oh, it would be better if it looked like this, but you don't know that until you kind of see how it presented it in the first place. And that's always just a nice thing for you to then like, maybe copy and paste your entire prompts and have a play around with this and see what your, your output's going to look like in a different thread. play around with those things. The thing that I, I always kind of want it to do on the other side is like once, once I've done, done this, this kind of step, once it's gone, it's gone lie, like I've put it all, I've put it all through is I will always tweak and I will always tweak as to kind of what, what I, what I want from the other side of it. It's never going to do what you want straight away. Yeah. But then. I found that these kind of prompts are the kind of prompts that you want for when you know you've got a very specific task for it to do. The other way that I love doing prompts is to get, get your LLM to almost play the role of a consultant. This works really well when you know what you want. Sometimes you have no idea what you want and you need it to ask you a load of questions. So one of the things that I. Where are we? One of these. Okay, cool. This one. So one of the things that I love doing with a lot of prompts is again, Persona this is kind of what I have I've given it like I need, I need some kind of output that I'm specifically looking for, but what I'm telling it is, you're going to ask me a lot of questions, because I don't really know what to do yet. Yeah. I just know what I'm, I know what I want to do. I kind of think I know where I am at the minute. I have no idea what that bridge in between looks like, but I also know that there's probably a load of data, a load of points that I'm holding onto that I've not told you that if, again, it's back to that whole, like, if I ask you to give me an output now, you're not going to give me a good output because I have a load of expectations I've not told you yet. So the thing that I love to do is you're going to ask me a lot of questions. This is where I am at the minute. After every question that you ask me, you're going to give me a score out of 100 as to how close you are to giving me a good answer. Only when you're above 90 can you start trying to give, like, give me an answer. That is a very interesting prompt. Yeah. And it's like, okay, cool. Let's, let's ask you a load of questions. What are your goals? What are your, like, what do your data points exactly mean? Like, this is where I am at the beginning. And the funny thing here is, like, it's asking me goals, which is literally one of the things that I was, I, I have in this one, because I know what the goal of this prompt is specifically, like, the, the, where I'm trying to work out, like, this is what my output is. It's literally just like, what do you want to do? What specifically are you trying to do? So it can sort of formulate all of those different bits. So it asks me the questions. I give it some base, like rough, rough answer. These are all perfectly fine. It's like, cool, here's some more stuff we're going to go through. This is. me trying to prioritize a load of keywords, but it's more the idea of like, I'm now working with it and it's giving me, this is how I'm thinking about these things. This is where I, where we are at the moment. Am I going on in the right direction? So the really interesting thing here that it's done is it's added this relevant score that I've told that I don't have the data for, like, this isn't something that I have. So I'm telling it here. Like, I don't have a relevant score. How do we figure that out? Is that something you want me to figure out? Or is this something that, like, can we move on without this score? So it's like, okay, cool. This is how I would try and work out what a relevant score might look like. We could potentially use CPC data. We could use a proximity to a core keyword. We could do some level of manual sampling. I think if I remember correctly, I'm like, I'm not going to do any of these things. Can we just not do it? But it's, it's like, okay, cool. Here's, here's 75 out of 100. But what I've noticed is it's not asked me any more questions. So I'm telling it like, ask me more questions. Give me, give me more questions. So then we go through, eventually we get to a point where it gives me this nice, like, formula. I tell it what I have as a spreadsheet. And then it's like, cool, here's, here's that entire formula as a spreadsheet, go test it. Like, as a, like, for your, for like Google Sheets, go and test it. And then there's just a load more like back and forth. This entire thing goes on for, for quite a while. But it's like, cool, I, I knew what my output was at the beginning. I knew what I wanted. I didn't know how to get there. I had no idea what the bridge looked like. I just know I've got this data and I want to get to somewhere. The problem that people will have is they have all of these. Unexpressed biases and expectations and they never express them and never give chat GPT or whatever LLM you're using an opportunity to work those things out to help you solve the problem. And I think that's probably one of the biggest things when it comes to like, you want to just be good at prompting, give it everything, you know, and if you don't know, you know, it, give it the opportunity to ask you a question for you to give it. Wow. I. I don't think I have seen anyone prompt in this manner where you are working with it and asking it to guide you to get to that final outcome, which is amazing because that is something that I personally have not tested so far. What I've always done is give it my mental model and my mental framework, but it's definitely something I would love to try to see if we can like work together to get to the end, end goal, especially if you don't know how to get there. This seems to be it. A great solution. Yeah, it's something that I've, I've used a lot, especially when I've got, so in, in content marketing, in SEO, you, you have a lot of data. There's a lot of places you can pull a ridiculous amount of data from, and it can be really hard to know what is worth my time, what isn't worth my time. How can I use it best for my specific situation? And a lot of the time when you look online, people have solutions to problems that feel like they answer your problem. This is how you prioritize a load of keywords, but in reality it's, it doesn't take into account what are your goals. Are your goals to prioritize the keywords that have got the highest sale value? Are the goals to Is the goal to get as much traffic as possible is the goal to try and rank quickly for keywords that are really easy for you to rank for and don't have a lot of competition, like they're all very different goals, but the problem that you have is can be labeled the same as I'm trying to prioritize keyword research. Yeah. So, but you can't talk to a blog and say, actually, my goal is different. Give me a different formula. You can hear. And if you wanted to, with all of these, you can plug in all of those blogs. You could go and find blogs about keyword research and, and prioritization and plug them all in and go, this is how people have done it before. Use these mental models. This is my situation adapted. Ask me whatever questions you need to know so that you understand my situation as best as possible. But when it comes to answering the problem, use these proven proven methods. What questions do you have? Yeah. And it's so cool because you're actually tapping onto the power of AI for higher level strategic thinking and instead of just pure execution, which is the beauty of it. And there's so much to unlock. And I'm so curious, you know, for because as a copywriter, I've tested different LLMs for conversion copywriting, so I know that different LLMs are great for different use cases. Yeah. What is your current favorite for long form SEO content? Good question. So we have, all of the writing on PenFriend is done with Claude, is done with Opus, but the logic on how we provide it with the right data is done through through, GPT. And then we do a thing where each section, we then pull it all out and run it through perplexity to fact check and cite all of the different information that kind of comes, comes from it. So it's really difficult to give you just a straight answer as to like what is the best for long form creation. If it's just the writing, Claude, I think Opus, Opus does an insanely good job at their writing. You still have to prompt it well, but I don't think you need to be as tight for some of your prompts and it can write just more. It's a lot more fluidity to it, if that makes sense. Yes, there has been what I observed as well. Interestingly for me, I also use all three tools for a copywriting flow. So if I'm looking to write a high conversion landing page, right? perplexity for, you know, your numbers, your, data, your analysis, ChatGPT for strategic positioning. It's really good at strategy. yes, it's still the best so far. And then plot for the final copy and content. So it's interesting to see that we are both using the same kind of pool of tools for, you know, different types of writing. And maybe one last question I would love to ask you is do you have any advice for someone who is in your field or Trying to use AI, but still trying to navigate, through this insane world. Understand the job that you're trying to solve for first, the problem. And the thing that I found super helpful was get on Loom or Zoom or whatever, and record yourself doing that job and say out loud how you do that job. Take the transcript, put it into ChatGPT, and tell it, give me a checklist on how I'm solving this problem. One, nice thing, now you have an SOP, that's always a good thing to have. Two, now you know all of the individual things that you can look to try and automate or do better, but the point that I'm making here is You're not hiding anything. So much. So I genuinely don't feel like I can iterate this enough. The reason why you write bad prompts is because you're not giving it all of the information. So if you can get all of that information as a good checklist of the job that you're trying to replicate as a prompt, then now you have a really good starting point and you know that the job isn't the problem. Now you just need to refine your prompt. So work on the input. That you're giving it fast. I absolutely love that. Yeah, I think it for me also. It's one last tip on my end is You have to understand your own workflows. Like this is the one thing that has really helped me figure out from thing by me Understanding how I come up with a high conversion ad copy and it actually is for example a five to six step process and then when you break it down into this tiny parts Prompting for it and get to Really high quality output. And with that, thank you so much for your time, Tim. And thank you for joining us. It has been such a pleasure having you on board and thank you guys for tuning in. So don't forget to subscribe to the podcast and hit the bell for more actionable AI marketing insights. We'll see you next time.