A (0:00)

GPT5 has been out for just a week and it's getting very mixed reviews. But one of the applications they told us it would be much better for is creative writing. And all of us marketers and people trying to grow our business would love to use AI to help us do more creative writing. Is GPT5 any good? We are going to put it through its paces on this show. But I'm going to show you a unique choice trick to turn GPT5 or any other AI assistant into a better writing tool for you. All of that and more on this episode of Martin against the Grain. Okay, so I'm in a hotel room. This is what it looks like to record a episode on the road. It has to happen sometimes, but we still want to come to you with a really great episode. And so I'm going to put GPT5 through its creative paces. Is it really a good writing tool? We're going to figure out how good it is in comparison to the other AI assistants by giving it three creative write in tasks. But to do it, I'm going to show you a quick trick to make these tools much better at crafting content. Now, first of all, let me go on a little rant. AI is a great assistant. It can help you do a lot of things. It can help you do research, it can help you do strategy, can help you do data insights. It can help you do kind of creating content. Now what I'm seeing is lots more people start to outsource their work to AI. So what are they doing? They are given AI a bunch of documentation or they've given a bunch of articles or given a content and then they're saying, cool, create some content out of this. And then they copy and they paste and they put the content live. And guess what? It looks like AI content. It reads like AI content. And everyone knows it's AI content. And it is just a bunch of words a lot of the times. And they are kind of outsourcing their actual work to AI. That is the lazy copy and paste marketeer. That is the lazy copy and paste knowledge worker. Think about it. What is your role if that is how you're using AI? Get data, put it into AI. Tell AI to do something with it. Paste output. You are a data entry specialist. You are not an important part of that workflow. That is not how you use AI. AI is an assistant. You are the one with the skills to turn the things that AI gives you into really great assets. And so on this episode, what we are going to show is AI as a creative tool that can help you think of great ideas, a great thought partner, but you never outsource the actual content creation. Final content creation to AI. That is a surefire way to have pretty poor results. Okay, so let's get into this quick thing that not many people do to start with that you should do if you want to use AI to create content for your audience. So let's say I have an audience which is VDP marketing. This is like too specific. But just as an example, VP marketing and mid market B2B SaaS. And so what do we have here? Okay, well what we have is a description of that audience. So if you're looking along at YouTube, which you should be, this show is much better in YouTube. But if you're in RSS and you can't make the YouTube, what I'm showing here is a pretty thorough description of that audience. It tells you about the core truth about this audience. This audience is perpetually defensive. They are smart marketers trapped in a constant cycle of proven their word to schedule executives who see marketing as a cost center. Darn, this is close to the bone. Pretty, right? Can you help me survive my next board meeting? Is something they think about. They're data obsessed but data overwhelmed. Oh boy, this is so good. What drives their engagement? We care about this because we want to create content that drives their engagement. Right? People do not think about this. You need to have a content ICP to create great content via AI. What drives their engagement? Fear based motivation, Time scarcity, peer validation, addiction. And it basically talks exactly how you bring these things to life. What is their content consumption mindset? They approach every piece of content with ruthless efficiency. They do scan them for immediate application. Yes, very, very important. This is actually really, really true. A lot of marketing leaders want to look smart and forward things to their CEO to say. I thought you might find this interesting. They're skeptical by default. They have psychological triggers. Urgency around credibility, FOMO around competitive advantage, professional survival. NeverBean has a more accurate description of marketing right now. How they actually make decisions. They don't buy features, they buy ammunition for internal blog, political. The battles. Every tool evaluation is really asking, will this help me defend my budget? It's too accurate. It's too accurate. Content that breaks through winners and losers. Right? It gives you the winners, losers and then it gives you the deep truth about this audience. Bottom line, they don't want marketing content. They want executive ammunition disguised as market insights. Brilliant. So this is an incredible description of an audience that we want to create. Content for. And I want to show you how we can use this for GPT5. But how did I end up here? You know, one of those movies. How did we end up here? Well, this is how we ended up here. We basically did a couple of cool things because you learn how to do the cool things on marketing against the grain. We created a content icp. And so what we did is we first start with the internal data. Now you're a new company, this is new to you, whatever it may be, that's fine. We're going to do internal data and then we're going to pair that with external data. If you don't have internal data, you can just use external data. The internal stuff is actually really simplistic. We're not creating an icp. An ICP is an ideal customer profile, who we are selling our products and services to. What we are creating here is, is a content icp. Who are we trying to create content for? That's very important distinction. So what this here is going to do is search for internal data and it does that through customer interviews, win loss notes, through CRM records, any kind of content performance you have, any kind of email campaigns. Now, I cannot build a prompt that's going to be applicable to all of you, but I'm showing this for illustrative purposes because the data you have internally may be different from the data I have internally. Main thing is if you can get customer call transcripts, one lost notes. If you do sales from your CRM records and then basically any kind of content performance reports and put them in your Google Drive. And what this will extract is customer language, which is really important. We want to make sure that's in our content icp. How do you describe things? And then we go into kind of content performance, right? What worked, what didn't work, what helped influence our purchasing. So examples for HubSpot is in our customer transcripts. We can actually see people referencing AI ChatGPT specifically as a reason they came to HubSpot to buy HubSpot because ChatGPT recommended it. That's something we would want to know if we're creating content to attract that buyer towards HubSpot. We go through basic Persona info, right? Their job title, company size, all of those kind of things. And so we output it into this, which is their main challenges, why they buy. Language notes, content insights, data assessment recommendations. So this recommendation I think is really good. What I put here, this data assessment recommendation, you can choose to have this or not to have it, is if you're running this, it gives you a confidence level of how accurate this ICP would be based upon how much internal data you actually had based upon basically how much internal data did it find to craft this content ICP for you. And so what it will do is if it's not that confident that this is accurate because you couldn't find a lot of stuff, it will say it's very low. Right. Limited internal data. You should rely more heavily on the external research medium is find some stuff but still get the external data for research or for validation. And then high it found a lot of good internal data. So let us run this and then we are going to pair this with external data. Okay, so here's an example. You know, it goes through and finds a bunch of things. I'm not going to spend too much time on this because I purposely won't find a lot of things because I'm not trying to run it against a lot of internal HubSpot data. But you can see it's output in the way that we want it here. It's pretty good internal data. And so here the data assessment like it found good things here. So it has medium good foundation from interview templates and some actual customer interviews. Gives you the data sources, what you want and then a recommendation, usable insights, but external research validation and then it talks about the things that it needs. So you could basically take this and craft an external prompt to get the rest of the external data. But we're not going to do that. We're going to take this output and now we are going to use one of my new favorite tools, Perplexity. So what we're doing here is we will make sure we include these because it's hard to see this, but we give Perplexity the Persona we've created from that internal data and then we ask it to start to pair it with external data. So we ask it to do content consumption habits, their content preferences and format here, headlines and hooks that work, messaging and language guide topics and angles, trust and credibility markers, content structure right inside examples. So we do this whole thing to try to get external data and what you end up with because here's the trick, right, you're not going to just copy and paste. We never just copy and paste. We want to get a bunch of things here, which we do. It does an incredible job actually. It goes through here, has like a good TLDR of your content Persona. Then it will go through their consumption habits. It has like a lot of resources for these. It has their preferences, what kind of content they like. This is actually really good. If you're creating an entire content strategy, we're going to condense this down into just the thing that I showed you, which was an audience description. Headlines and hooks that work. Messaging, language guide topics and angles, trust and credibility marketers. So this is pretty good, right? Who do they trust? Hey, Dave Fender. The show's in there. Kyle Lacey. Why the hell am I not in there? Perplexity. I love you. But now you've kind of. I should be in here, right? We all agree with that content structure. Write in style examples. It always does. Like some pretty cool dashboards. Shows you their dashboard inside B2B content preference. This is very true. More and more people are gravitating towards short form video. Actually ebooks right down the bottom. White papers. Really people still waiting. White papers. I don't know about that. What else? Blogs. Blogs are still popular. Okay. And so what we've done is taken all of that and we started to work with perplexity to really summarize it. That's how we came up with this. So we don't want the entire content strategy. What I really worked on is like, tell me about this audience content preferences. That's the thing I truly care about. And we ended up with this. And so the takeaway here is you have the ability to take the internal data and craft a skeleton version of your ideal customer profile in terms of their content preferences and the kind of writing style that work for them. And then you compare that with perplexity to layer on the external research. And then you work with perplexity to condense that all down to this pretty great description of your audience. And so why do we do this? Well, for ChatGPT, what we do is we say, okay, we copy and paste this and then we say to ChatGPT, store this audience description in VP marketers. So this is my audience, right? Anytime I want to craft content for this audience, I just store this or inscription in in memory. Show you how this works. Give me the description of VP marketing. Awesome. Now we take our creative writing test to see how good ChatGPT is and we can even start a new chat because memories across chats and then we say, so we're going to generate some headlines. So product is pretty easy when we do HubSpot CRM and for in this case, what people don't do, we're going to give it the VP of marketing as our audience because now we have that stored. Now you can have multiple audiences stored, right? This is how good this is. And so what it's going to do is create 10 headline hooks for your product for this ICP. Now, these are creative writing tasks, so it's going to use a different device for each. Right? It's going to do contrast curiosity gap metaphors. You'll see what it does here. It's pretty good. It says no buzzwords. Always put no buzzwords. I cannot emphasize this enough. The amount of people who copy and paste AI and just allow it to have lots of buzzwords. You don't want to do that. All right, so here's 10 that it's created for. VP of marketing, right? And so what's it got? The report your CFO wishes you read. Remember, the VP of Marketing really cares about proving their worth. Turn your marketing chaos into a single Source of Truth. Seven boardroom questions you'll answer in 60 minutes. What did we say in the audience description? The marketer really cared about what their board thought about. Mini story. Last quarter a VP kept their budget thanks to this. What did we say in the story in the audience description? They want to protect their budget. Here's one for paradox. Grow faster by tracking fewer metrics. That's a pretty great one. Taboo. Stop faking roi, start showing it. That's a pretty great one. Here's one for villain. Your spreadsheet is quietly killing your pipeline. Social proof. Have 43% more CMOs prove their worth this year. Future cast again. Board meeting. Next board meeting, you walk in holding the win again. Remember, for this audience, that was a big thing that was important to them. So this is how you start to get it really exact for your audience. And what we put here is a very simple thing, right? No buzzwords. Look how much better it is when it doesn't do like buzzword lingo. Let's do another one. This one is pretty cool. We're going to explain growing revenue with less metrics to our icp, which is VP of Marketing, using an analogy from a surprising but fitting domain you choose. This is for the AI. Your domain must be unrelated to market and business, yet to make the concept feel more tangible and memorable. So again, this is a really cool creative writing task. So let's see what it comes up with here. We call this the cross domain analogy test. So growing revenue with fewer metrics is like running a Michelin star kitchen with a trim down menu. Instead of offering 50 dishes, you focus on eight each perfected through taste, time and presentation. The chef isn't checking a dozen timers or juggling endless recipes or listening to the sizzle in the pan, smelling the butter breath. What's this mean? This is actually pretty good in terms of writing. I just don't know. It's a very good marketing piece of text. Clarity frees the team to execute flawlessly. Blah, blah, blah blah. Cta, stop counting every number. Perfect. The fear that feed your growth. That's pretty good. So this is like a fun one, but again you can see it's crafted for that audience. Let's try this one which is one of my favorite. This is going to actually create an ad using VO3 and 8 second clips. So I'm actually pretty excited to use this one. So basically what it does is for HubSpot CRM, we're going to put in again our audience because that helps us to make this more tailored to them. And then it's basically going to start to create an ad for them. It's a 52 second ad because we're using VR3. It's going to break it into seven second sequential clips. And it basically talks about it being paranoid to pragmatic to playful. And then each clip should have a visually memorable hook and be a self contained enough to stand alone if seen out of context. Right. So basically let's see how good it is at creating this 52 second ad. So it's given the variable, but it has definitely tailored it for this audience. Hotspot turns your marketing chaos into boardroom because again, the board really knows that the marketers care about the boredom, proof, revenue, clarity. So why this works for VP Marketing gets their pain. They're drowning in fragmented metrics under constant pressure to justify budget. Again, where did we get all this from? The audience walk into any board meeting with indisputable proof of marketing's revenue impact. What's unique advantage? And it talks about HubSpot's unique advantage. Let's see the ad here. Okay, clip one is our paranoid clip. There's a visual dimly lit office. A VP scrolling through 12 different dashboards at 1am, eyes darting and I will build this ad and I will do it for the next episode because I'm doing a VO3 episode. Your board meeting is in six hours. Wow, they've really picked up in the board meeting. Which number will you trust? Paranoid overhead shot of sticky notes, spreadsheets and tangled wires sprawling across a desk. Every tool says something different and the CFO is already suspicious. Okay, this is pretty good. This is true visual, pragmatic cam, moon and light. One clean HubSpot dashboard loading instantly on a screen. What if every deal, every dollar lined up perfectly pragmatic, side by side messy spreadsheet versus HubSpot's clean attribution report. Hotspot connects your data so the story writes itself. Accurate and undeniable. Playful. The VP is striding into the boardroom, dropping a single printed report on the table. One number. All eyes and you. All smiles from the cfo. Playful close up of champagne, cork popping in slow motion. You didn't just survive the meeting, you owned it. Okay, it's a little cheesy at the end. I think I would make the playful much more playful. But. But for a first version, I think this is pretty good. I do think GPT5 is doing good creative work here. Gets you a first version that you could really iterate on. And so I do like that. And so I will work on this and I will show a clip. But again, the core takeaway here is to show how good this is with the audience part. So this is three use cases. I think GPT5 is much more of a creative writing model. Can you copy and paste what it gives you and put it out there? No. But the main takeaway here is always make sure you load your audience variable into whatever content you're creating. And so whatever you're creating content with AI, you want to give it the audience that you've crafted. Whether you're using these tactics that I've given you or you just have one already, you have like a description, load it into memory now. Then every time you want to create content for different audiences, reference that audience and the content would be much more tailored for them. You saw it in every single result. Works really well. That's how you go and turn AI into a better content creative assistant. See you on the next show. SA.