Leveraging AI

92 | Mastering AI-Driven Content: Step-by-Step Guide for Super Engaging Content with Charlie Hills

• Isar Meitis, Charlie Hills • Season 1 • Episode 92

Share episode

Join us for an engaging and insightful webinar where Charlie Hills reveals the secrets behind AI-driven content creation. Designed for everyone, this session will transform your approach to content.

Learn how to find the best sources using a specialized Perplexity prompt, and see how it all comes together with step-by-step guidance. Charlie will show you how to seamlessly integrate prompts into tools like Gemini and ChatGPT for flawless content.

We will also walk you through the editing and finetuning process with Gemini. You'll see firsthand how to craft compelling LinkedIn posts and original comments that resonate with your audience.

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Isar:

Hello everyone. And welcome to a live episode of the Leveraging AI podcast, a podcast that shares practical ethical ways to improve your business and advance your career with AI. I am really excited about today's show. a lot of, that you can use AI to create content. And some of you has already started doing this, but very few people actually know how to do it well. And when I say really know how to do it well is. Having an actual well structured process using very specific systems in a very specific way to achieve really good, solid results. When I say good, really good, solid results is getting to a situation that you have a piece of content that is consistently performing as far as driving engagement. This could be content on LinkedIn. This could be newsletters. This could be so on. And the way to do this correctly is to have AI systems and processes. For every step of the way, meaning you need to start with research, what kind of content we want to share and then aggregate data and then summarize the data and then put it into whatever form of content you want to put it in. And because every one of these steps, if you don't do it right, the quality of the content is not going to be as good as it needs to be. Our guest today, Charlie Hills is an expert in exactly that. And he made this work to a level of perfection. I haven't seen a lot of people do, meaning literally build a fully customized machine that knows how to go through every step of the process to generate amazing content that is valuable and engaging that a lot of people follow consistently. And if you follow him on LinkedIn, you would know what I mean. So recommendation number one, follow Charlie on LinkedIn. Recommendation number two, let's go through this episode and learn how exactly he does it. He's going to share with us every step of the way, like which tools he's using, how he's using them, what's the full process. And because this is really valuable to literally any company, any individual out there, I'm really excited to have him as a guest, Charlie, welcome to Leveraging AI.

Charlie Hills:

Thanks Azhar, it's great to be here and thank you for that intro. I really appreciate it. to be considered an expert in content AI marketing is, news to me, but hey, I'm that guy now,

Isar:

Yeah. Listen, I, we've been connecting to LinkedIn for a while. And when I see people that consistently get, dozens or hundreds of people follow their content and engage with their content, it means you're doing something right. It's when it's a one off. Yeah. Somebody gets a home run lucky. Yeah, it happens not a lot, but it happens, but to do that every day or several times a week, means that you figured it out. And I think that's something that is extremely valuable that probably everybody wants to figure out. Please share, what's your process? What is it that you're achieving? How you're achieving it? I know you prepare like a full flow. Let's do that and just follow the process.

Charlie Hills:

Sure. Yeah. So today I'll be showing my newsletter Megaprompt, which I put into Perplexity. So I've been writing my newsletter, Martech AI, Since October of last year, started out as a pure passion project, and I was writing it manually every week about the latest AI drops, and it was quite a time consuming task. As I've been ramping up on LinkedIn, posting more and more, my time has slowly but surely become less, and I've become more time poor. As it's gone on, but that's meant my mega prompts have got better. And my AI processes have to have had to improve to keep up with the pace. So that's why I made this mega prompt and it's probably cut around 80 percent of my writing time for my newsletter. So huge time saving there. And I guess I'll just jump straight in and share my screen.

Isar:

uh, For those of you who are not watching this either on YouTube or us live right now on LinkedIn or zoom. So a for next time join us so you can see it live while we're doing this. But B, if you are again, listening to the podcast, then don't worry about it. So we're going to share everything that's on the screen and tell you what we're watching. It's just a tool that helps us walk you through the process.

Charlie Hills:

Awesome. All right, let's go. So I'll start by Showing my most recent newsletter, CTA number two, please subscribe to Martech AI if you're not already and essentially this is a newsletter that lives on LinkedIn. I've been running it for 30 weeks now, so over half a year and. It has a very straightforward process nothing Spectacular. yeah, it is a bit spectacular, but it's you know, it's a very simple bulletin of Just the latest ai news, you know listing the publisher the overview key points. And this is where my own niche comes in being an ai marketer I add my own marketers point of view with the help of ai so I won't bore you with this side of things for too long I'll just jump straight into the prompt here full screen. Okay, cool. So I'm start, so I've got two separate prompts for my newsletter creation. One is for sourcing the URLs and half this process of sourcing URLs or finding the news stories, half of it's manual. Half of it's done through this mega prompt, so I'm quite active on LinkedIn. I'll be scrolling through my feed. I'll see, ChatGPT drop this 4. 0 model. I see, Google's got some crazy updates and, I'll make a note of it in my phone. But, there's some stories that I miss. I'm only one person. I can't pick up on absolutely everything. So that's why I developed this source finding prompt, which I'll be using Perplexity, to find the sources. Let's just go quickly through the prompt itself, starting with a scenario setting. please act like a experienced research analyst specializing in AI trends with a particular focus on AI marketing developments and what I'm tasking the. the AI with it is to find additional trending, irrelevant marketing news articles from big industry players within this specific date range, which is, I cover my newsletter from Saturday to Saturday. So this was last the previous Saturday to last, and then ensure the articles are relevant to your major, relevant to major AI advancements and fact checked all information to ensure accuracy. This one's a big, Bit of the prompt that should be included because I found previously with my other iterations of this source finding prompt I was it was just spewing out a load of garbage to be honest. It was talking about chad gbt Three and gem, dali two. I was like what the hell, you know So that's why I bake so that's why I baked that in there And moving on to the objective itself, which is to report additional, trending and relevant AI marketing news from major industry players and announcements from the past week. And instructions are quite simple, identify those players. I've listed a few here, Google, OpenAI, Microsoft, and Facebook. I should say Meta really, but hey. Oh, Salesforce and Adobe too, as well as other top tier companies. And then search for relevant articles, list the relevant articles, fact check those articles, and then provide me with the title, publishing date, and a brief description. So if we're ready, I'll just jump into Perplexity. I have it. So I just want to,

Isar:

for those of you who don't know, Perplexity is a really great tool that is, basically if, Google search and ChatGPT had a baby together, that's what it is. So It's a search engine that is also a large language model. So you can do everything you can do with any other large language model and direct it and provide it prompts and explain what you want and what format you want it in and have a long conversation with it, but it also goes and fetches real time information. And the other cool thing is that it actually gives you the links to where the information comes from. So even if you don't. Act and add the fact check aspect to it You can actually factor check yourself because you can actually click on the links and see where the information comes from which makes it A really great research tool. So now with that let's go to see what the result is when you put in this prompt

Charlie Hills:

exactly Yeah, like you said the citations are super handy to know that what you're reporting is correct But one caveat I must add here is you have to enable the Pro search functionality to get the best results for this, part of the process. You only get five of those a day. So use them wisely. Luckily, I only have to do this once a week, so I don't pay for Perplexity Pro, the free version does the job. So yeah, I'm just gonna press enter. Let's see. Let's see what it comes up with. So with the process, it does take a little bit more time to like process and understand what the question is rather than just spitting out the answer immediately as perplexing usually does. So let's see what it's come up with. I must say sometimes it's not always 100 percent accurate, but it does help me understand additional stories I might be able to cover. So some, yeah, so for each, I'll run them through a quick Google search to make sure. Okay. I'll make sure that there, are what they say they are. So this was from the 7th of May, so therefore I can't report on it, but it's good for me to understand. potentially other stories I'm missing. It doesn't seem like any are too relevant here. It helps me just find additional stories, which I might've missed. But I'll move swiftly on to the main gamut here, which is the newsletter megaprompt itself. So this one's way more comprehensive. So please, get your popcorn and buckle up for it. So so quickly, as an experienced copywriter and journalist with a focus in marketing technology and a practical application of AI decade of experience as well, that's important to add into the prompts. just for additional context and scenario setting. And yet this is this part of the prompt is simply scenario setting. I call it. And then moving on to the objective is to compose the weekly edition of Martech AI, which is my newsletter. I've included the date range. but also some additional information here. Should you not only deliver information, but also captivating and engage the reader and then the instructions themselves. So this section is. Literally just for the title alone. I don't know why but every time I told Perplexity or chachi bt to write in Sentence casing or title case. I can't remember which one it never adhered to what? I told it to do so I had to build a four step, instructions just for the title alone, which is nuts to me. but it's quite comprehensive. I won't go into. let's just

Isar:

go through the steps so people understand what the steps are to just create the title. And I assume. What you feed it is the content that you found in step one, or you just do this and it's all in one step.

Charlie Hills:

Yes. So exactly that the content that I found in step one goes at the bottom here with all the URLs. So

Isar:

at the bottom, we got links to these different articles that you find in running searches on Perplexity based on the topics. And then in, and then the next The beginning of this prompt is just a title. So what are the steps of the title definition?

Charlie Hills:

Sure. so the title itself is to highlight all stories mentioned. Then the objective is to, it's to cover everything in a straightforward way. extracting the key points. Also the composition begin with the issue and series number followed by a concise enumeration of fancy word or the main stories or topics discussed. And yeah, it's honestly, I feel like a bit of this could be, a bit of fluff, but I like to keep it in just in case to cover all basis because I've had so much problems, so many problems with the title alone.

Isar:

So I want to touch on this for a second because the title obviously plays a very big role because that's going to be the decision for people, whether they're going to click or not. And people say, Oh, create a title for me. And it will create a title for you, but it may not be good enough to really attract people to open your email and then to actually read your email. And so it's worth investing and it, the way Charlie's doing this is by Breaking this into several different steps, including summarizing the information, extracting key points, and only then creating a composition that will be a hook that will be engaging enough for people to actually click. So it's worth understanding how the AI addresses this in order to give it the right instructions in order for it to give you an outcome that makes sense. And some people are like, it's only writing a subject that can write the subject myself. I'm like, yes, but not. You'll then you have to do the whole thing on your own every week versus investing two hours, three hours, once to figure out the prompt that actually makes it work. And then you don't have to do it again. And you get prompts that work and you get titles that work every single time.

Charlie Hills:

No, exactly. I've spent, half an hour just umming and ahhing over a title or a hook. And the fact that this does it in a routine fashion is yeah. Like you say, a massive time saver and yeah. And then The final part of the title section itself, clarity and, conciseness as well as comprehensiveness might sound a bit contradictory, those two points, getting to as close to the!Equilibrium between them as possible. Awesome. Moving on to the next section. So this is all, it's a little bit more, succinct, let's say. So the intro after the title, begin with a dynamic introduction of the latest developments in AI marketing, emphasizing potential impact on the industry. No more than 25 words. That's very important. Otherwise the AI will just spit out maybe a hundred words and then you'll spend five, 10 minutes, maybe sifting through it and deciding on the 25 that you'd like to include.

Isar:

Question about this, you're running this prompt through Perplexity. Oh, yeah. Because every time I try to run word count on either ChatGPT or Claude, it fails miserably. So I'm actually, happily surprised to know that one of them actually knows how to count words. Like when I go to ChatGPT and say, oh, make it, 30 words. It may make it anything between eight and 50, like the spread is pretty big. It's not like I'm telling you 30, it's going to land somewhere between 28 and 32. So the fact that it's doing it is amazing.

Charlie Hills:

Yeah. I think Perplexity is the only one that I've found can adhere to the word count that you provide it. I was doing a test between Gemini 1. 5 and GPT 4o. And I was just asking it to write a simple blog for me. Gemini always went under the word count I asked for, and ChatGPT always went over the word count. So pick your LLM on that basis, I guess. so yeah, moving swiftly onto the TLDR. that is just, to summarize. It's each of the sections, including an arrow symbol, just to make it a bit more engaging rather than bullet points. And then the section headline itself. So device a compelling headline for each article of the newsletter, ensuring between six to 10 words, you sentence casing only in the first. Yeah. So this is what I had to do to be very clear to the AI. Instead of just saying you sentence casing, I had to then also add a bracket saying only the first letter of the title was capitalized except for proper nouns. That. changed my prompt engineering game after including that. And then, and it grabs the attention of your professional audience. Also provide the URL because on LinkedIn, I'll embed that into the newsletter itself, and as well as the publisher in UK dates, if anyone's confused why they're the wrong way around, overview. So for each section, one sentence that summarized the key aspects of the content, no longer than 16 words each. And highlight the importance, Of my niche, obviously being marketing Key points, identify and elaborate on the three most crucial insights or developments discussing each section. I decided to pick three, because any more, and it becomes a bit overwhelming for the reader. And then the market's POV in a single well constructed paragraph, no more than 25 words, synthesize the insights and the overview and the key points, and yeah, provide a strategic analysis from a marketer's perspective on how these developments could impact the market. I

Isar:

want to pause you and go back for a second, if you can scroll back so we can see the steps. This process, yeah, from the intro and down. So you're right there. The interesting thing about this, like this process that goes from basically, Charlie very specifically defines the format of the outcome that he's looking for. So it includes an intro TLDR, section headline, the URL where it's coming from, who published that information, an overview of what the information is, the key points in that section, and a point of view of something specific in this particular case, marketing, because that's the content that he's sharing, but the beauty about this, if you think of it, this can be used, a similar process can be used for any content you're trying to summarize for your company. This could be market research. This could be reviewing our customer reviews in a structured way. This could be, employee, references to like whatever information you have that you want to summarize for some kind of a report that could be monthly, could be quarterly, could be whatever. You can do in the same exact process. And it's again, worth investing the time in developing and defining the exact format that you want it in. But then you can do it again and again, just by feeding it the data that exists. and in this particular case, it just comes from research and Perplexity, but in your case, it could come from anywhere else. So I really love the process as far as the process more than the specific thing that you're using it for.

Charlie Hills:

Thanks. Thanks Isar. Yeah, totally. I, yeah, to be clear, I had this structure before I started using Perplexity and this mega prompt. So if you have a preexisting newsletter or process, yeah, just, try to reverse prompting engineering and yeah, let AI do the heavy lifting for you as we like to do. So moving on to cross referencing and depth of analysis. So the objective here is to cross reference the information from each article with additional credible sources to validate accuracy and enhance depth of analysis. this helps in providing a well rounded perspective in each story and the method for each key point extracted from the main article, seek out additional reputable sources that discuss the same topic, compare and contrast the view points or data presented to ensure a comprehensive understanding of the subject. So this section. I've included to make it a more richer and enjoyable read for the reader because you know providing one url Or link you're limited to that publisher and how they've highlighted the story this section synthesizes and it pulls all the information from additional resources and just makes it a greater read. I think anyway and then yeah clearly cite all additional sources used in cross reference cross referencing And yeah, this is

Isar:

basically just doing proper research, right? So instead of just providing one article, it's looking for other sources of the same information to see if there's other information that's not mentioned in one article, but maybe mentioned in another in order to summarize one concise thing. And again, I think it's absolutely brilliant.

Charlie Hills:

Thanks, I appreciate that. No, definitely. I added this a bit later on as I yeah I slowly but surely realized that it wasn't really working just providing one Link and that's another thing i'd like to add is, this didn't happen overnight this mega prompt I added it bit by bit and eventually got to the place i'm at now. I just yeah Small, small gains make big wins. So conclusion. My conclusion is always the same. That's all for this week Did you find it useful? It's repeated there. Very strange. I don't know. Let me get rid of that. Subscribe to this weekly newsletter to stay updated on marketing and AI. That's all for this week. Stay tuned for more next week. Warm regards, Charlie. And then take a deep breath and work on this problem step by step. I think you could probably remove that. I'm not sure if it makes a massive difference, but it's just nice to have. Definitely. It doesn't hurt

Isar:

research. Research is showing that it does work. So it's worth including.

Charlie Hills:

Good to know. Good to know. Yeah. Treat AI like a human, I guess. okay, cool. All right. I will, I know I'm copying and pasting. It's actually thinking about it. I've already got it prepared over here, so you don't need the pro search for this aspect. Which is, which is good news. And if we're ready, I'll press enter. For those of

Isar:

you who are not watching this, we're now literally just pasting this whole very long prompt with all its different steps into Perplexity to see what it's going to do.

Charlie Hills:

Exactly. Let's see the AI magic happen before our very eyes.

Isar:

Voila. So let's read. At least just the top section and then one of the examples, so people who are listening can understand what the outcome is.

Charlie Hills:

Sure. Okay. So the title is started by listing the edition Martech AI30 and then the TLDR. Google unveiled VO and image gen three open. I launch is new free chat, GBT, Google updates, Gemma Gemini and Gemma open. I partners with Reddit and anthropic launches in Europe. The intro AI continues to reshape marketing with groundbreaking advancements on tech giants. This week, TLDR is the same as the. title I just read, so I won't repeat. And moving on to the first section, Google unveils VO and ImageGen 3. So publisher being Google and the date is indeed the correct date. Overview, Google introduces VO for video generation and ImageGen 3 for improved image creation. Key points, VO launches AI to generate high quality videos from text prompts, which, I don't agree that is not accurate. Imagine 3 enhances image generation capabilities with improved coherence and detail. Both model prioritize safety and responsible AI developments. Marketers POV, these advancements now open new avenues for dynamic visual content creation, enabling marketers to craft compelling campaigns. F. Efficiently. So looking at this section alone, I wouldn't say it's the best output from Perplexity, but that's okay. Cause you know, I've built an additional part of the process or an additional prompt, which I input into Perplexity to improve the final output. So here it is, it's called my fine tuning. Copy prompts or copy prompter. And what I'll do is I will put that in there, put that into Perplexity. Sorry for the people listening. And then I will re copy and paste that article section. sorry, maybe I should briefly go over the prompt. Yeah. Let's go briefly

Isar:

on what that prompt says.

Charlie Hills:

Sure. So write an engaging yet concise informative headline for this article subsection. So it essentially goes back over. the mega prompts, but just with a different, just with a few different commands and then just focusing solely on that subsection alone. Cause as we're probably all aware at this point, doing mass content creation with AI doesn't always lead to the best results and the more kind of granular you can get with it, the better the outcome. So again, Scenario setting, act like a savvy digital marketer with over a decade experience specializing in AI tools. Improve the following concise overview of this news story, including the most important details in one to three sentences. Summarize the key points of the story in three bullet points, focus on the most relevant information for marketers. In a single well constructed paragraph, Of no more than 25 words synthesize insights from the overview and key points provide a strategic analysis from a market perspective on how these developments could influence various marketing dimensions. So I will just press enter here in Perplexity.

Isar:

And you add just one article out of all your articles every time and you just do it again and again. Yes.

Charlie Hills:

If I'm unhappy with that section, most of the time, the ones that are

Isar:

not written great, you just add them to fine tune how it's written.

Charlie Hills:

Yeah, exactly. most of the time, to 80 percent of the time, they're usually pretty good to go and I can just copy and paste them straight into chat. It's like a GPT, obviously with a bit of human fine tuning. But this one, it was wrong. It was inaccurate. I must say. so Google doubles down on generated, AI with the, and imagine gen three. And so I quite like here how has expanded the overview because then me, the human, I can come in and be like, cool. I like that first bit, but then. The rest get rid of, and yeah, it does. It does seem to be more accurate. The key point here being VO leverages cutting edge AI to generate realistic hybrid, high fidelity videos from text descriptions. That's pretty spot, spot on if you ask me and the rest seems to be quite accurate, but then there is an additional prompt just to fine tune everything, which Celeste actually, I found out through Celeste, which she was on the, The podcast that I believe yeah a few weeks back and it's great. I love this prompt one One cent shorter and cut the fluff

Isar:

So I want to say two very important things on the overall process and not specifically about what you're doing, and I think it will be important for people to understand as far as stuff they can implement on their end. One is the concept of a prompt library, right? So I talk about this a lot, but this is what it is, right? You need to work. And rework your prompts again and again, until you land on something that works. And even then you probably want to keep on making small changes to just to test it out, but your main work is going to be that one prompt, which you want to save somewhere. And it doesn't matter whether that somewhere is notion or a Google doc or, SharePoint or magical Chrome extension or whatever it is that you want to use, but it's somewhere that you can save the prompts that you and your team can reuse again and again. So that's. Concept number one, concept number two is that you can refine more and more steps of your process. So think about what Charlie's doing here. He started with research, but then he went to writing the actual content, but then he went to refining the content still as the writer. And then he went to play the editor and saying, this is good, but it's not perfect. And I want to edit this and that in order to make it shorter and more concise. And when you go through that mindset, you can break every process you're doing. It doesn't matter whether it's content creation or anything else, do the same thing, what are the steps of the process that I'm doing? How can I build a prompt to replace and or shorten or, and, or improve that particular step of the process that I'm doing? And this is exactly what Charlie's doing here. And again, it's just. Works

Charlie Hills:

No 100% and that first point having a prompt library. I probably only started putting together my prompt library a few months ago and since Just oh what a time saver instead of just typing away being like, please could you make this like this? It is you know, yeah massive if you don't want to have one i'd highly recommend considering it for sure. Yeah, not you

Isar:

Yeah, me too. it's a good, it's a good recommendation for everyone.

Charlie Hills:

a hundred percent. And yeah, I don't always use one central to cut the fluff. Usually I prefer having more in the LinkedIn newsletter and then just using my human expertise there, to cut it and make sure that it's accurate and relevant and engaging ultimately. But yeah, If you'd I can show, everyone the previous process of my custom GPT and, why I moved away from that and more towards prompts, mega prompts and prompt library.

Isar:

So let's maybe not show, but let's talk about. So first of all, those of you don't know what GPTs are these mini automations that you can build into ChatGPT. It's basically for a very similar process of what Charlie's doing. It allows you to create instructions that you want to do again and again, and you can build it as a GPT, which again, is this mini automation built within the ChatGPT universe, tailored to do something very specific. So why. So I'm going to talk a little bit about how did you stop and why do you think this current process is better? I think it would be interesting to them.

Charlie Hills:

Of course. Very happy to share this. So custom GPTs, when they came out, I thought they were all the rage. it makes so much sense. You just upload the prompt and it's going to continue producing results day in, day out. It's not the case. I didn't know this, but GPTs degrade over time and get worse. You need to, I guess like you said, is all with the prompt library. You need to keep updating it. And also the fact that you're stuck in the chat GBT ecosystem means that you're limited from using tools like Perplexity, Gemini, Claude, whatever you fancy really. And yeah, it was honestly causing me more. Struggle having to go back into the GPT and change this instruction, change that. I just wasn't effective for me. It was a nightmare, by the end of it, to be honest. So I think the store is great, but custom GPTs. I'm not so proficient in that. So

Isar:

got it. Yeah. I'm actually a big fan. We do more and more in, in Multiplai my company and for my clients with GPTs. But that being said. Ed, I think there are two big disadvantages. one you already mentioned, which is you're stuck with. If you get better results with Claude or get better results with Perplexity or Gemini or whatever tool you're using, then you can't use them in those other universes, which is a big disadvantage because there are clear use cases. the other tools are doing a better job in doing something. So that's number one. Number two, that I think that's interesting. It's limited. I think it's 8, 000 words and then it stops. So if you want to write a really long instruction and it happened to be twice, trying to create GPTs for something, it just wouldn't do it because I ran out of words or characters or whatever the limit was. And then you start to make compromises because you want to fit the whole thing into a single instruction. Into that one. And so there are two solutions to that. One is like you're doing, which is just creating a prompt library and running it on a third party tool. If you don't have a problem with the GPT. With the ChatGPT universe, what you can do is you can run a regular chat in ChatGPT and calling different GPTs in the process. And I know that sounds a little confusing, so I'll try to re explain what I said. There are two worlds. There's the regular chat within ChatGPT that most people are using. And then there's the GPTs, which are these little mini, automation. And in a regular chat, you can use the at symbol to call a custom GPT that you've created. Created which means you can call several of them in a single chat So if I go back to the process that you did you can start the conversation a regular chat again versus the custom gpt I think that would be the best wording I can use to explain this So in a regular chat you can start the first step by using an ad symbol and calling your research step And doing that and then once this is done Still in the regular chat, you can use the ad symbol to call the next one that summarizes it. And then in the third step, you can call the custom GPT that refines it and so on. So that's a way to do this still within the ChatGPT universe. But I agree with you that there's things that Perplexity and Claude and Gemini just do better. So just giving you options, depending on what you're trying to achieve.

Charlie Hills:

Yeah. I find ChatGPT for web searching, especially when you provide it with lots of URLs. It's horrendous. which is why I moved away from custom GPTs because I'm providing five, six URLs at a time. And I found it searched the first one and then gave up.

Isar:

It was like, great. Yeah, no, I definitely for stuff like that Perplexity is still the best tool to use. And it will be, I'm wondering now with the new updates to Gemini, did you try it over there? Because in theory, it's supposed to do something very similar.

Charlie Hills:

Sure. Yeah. So I did actually do a mini test before coming on here using. The same newsletter mega prompt and chat gbt since the 4. 0 update is definitely a massive improvement I used to rank gemini Higher than chat gbt. I've been a massive gemini fan. I can't recommend it enough for human sounding engaging That's its strength, but ChatGPT seems to be, yeah, it seems to have upped its game. So if we're ready, I'll put the same prompts that I put into Perplexity for the newsletter creation into charge UBT and we'll compare let's go.

Isar:

I'll say one thing for those of you wondering what we're not mentioning, Claude, Claude is. I find as far as writing content in a human sounding voice, Claude does the best work, but for a reason that still escapes me, Claude does not have access to the internet. So you cannot give it links to summarize. Otherwise it would have been probably on this list as well.

Charlie Hills:

Yeah. I've never actually used Opus myself, but Claude Opus that is being its most advanced model. but yeah, I'm hearing so like lots and lots of good things. And the fact that you're, A bit of a Claude fan. It's definitely, yeah, I'm turning my head towards it, but there we go. So the newsletter has been created. And the annoying thing about Gemini and chatGBT is that it loves to put subsection headings in the key points. And no matter how much I prompt it not to, it still does it every single time. So just a small caveat, but generally I'm it's comparable to Perplexity. Yeah. But I think even what you

Isar:

said, if you now have to go in and edit it every single section when in Perplexity you don't, then for this particular use case, it's go with Perplexity if it's going to save you that extra edit.

Charlie Hills:

Yeah. you could do the same process of, putting those fine tuning prompts in there. Yeah. the same as Perplexity, but obviously Perplexity being an AI search engine. it's a no brainer if you're doing the news bulletin style newsletter, like I am for sure. And then I guess we'll move on to my old faithful Gemini here. So I'm subscribed to Gemini advanced, which costs me 20 a month, but you do get access to Gemini 1. 5 with 1 million. Context window, which is quite something. it's like a very good marketing or very powerful Ai, And yeah, if we're ready i'll just press enter on this and let's compare so a little bit slower with the initial. Thinking. Oh, that's not good. It started with a dear name here. oh my Sorry, if it's writing an email to somebody versus a newsletter, right? i'm like i'm a bit. You Yeah, befuddled here. I, I tested it before and it did actually come out with the full newsletter and the fact that it's just writing a, an email here. Sure. Here's the marketing email from Google. God,

Isar:

which is, I think this is a, an amazing comparison, right? We literally took the same exact problem, run it on three different tools, and we see very different results, right? Perplexity, that is your tool of choice for obvious, for an obvious reason, actually did the work you intended. the latest model by open AI. So GPT 4. 0 did an okay. Job, but not perfect. And Gemini, you tried it once, twice, once worked okay. And the second time completely misunderstood what it needs to do. So love it. It actually shows what works better. And what it actually shows more is for different use cases. You have to use different tools because some of them are going to perform better than others.

Charlie Hills:

A hundred percent. And that's where my typical process is Perplexity for the main bulk of it, and then I'll use Gemini maybe just to fine tune certain areas because I really like it for its human sounding copy. It's so engaging and I'm, I'm a big Gemini fan, so I'm very happy to share maybe one or two Gemini prompts if you've got time for it. sure. let's go through one or two. sure. So this one itself, so I posted this, about the new church, UBT updates, it was on Sunday and it's actually one of my best performing posts to date, it was very simple. It was just a brief overview of, how to access GPT 4. 0. Yeah, it was so simple. It wasn't anything spectacular. and yeah, all I did was write very loose instructions on how to access the new model. And I provided Gemini with the carousel attached and all I said was improve. That's it. And then if I press enter,

Isar:

don't let me down. So in this particular case, you're using Gemini as an editor to improve the way your post is written based on a post you've already created.

Charlie Hills:

Yeah, exactly. I've done the hard work and I've put together the carousel. And then I'll just throw in very simple instructions into Gemini and it'll give me my post I literally used the output Gemini gave me just gave it with some nice formatting and it was good to go that's a bit of a departure from the newsletter, but the newsletter side side of things is very simple I'll just put these same prompts in here that I used in Perplexity, but I'll break them down into subsections. So let's say that I wasn't happy with, a marketer's POV in the newsletter, that, the output that Perplexity came up with here. So like this one, for example, yeah, I'm, I actually don't like that marketer's POV, those innovations unlock new, instantly It's not good. So going back to the newsletter, let's test and Oh, much, much better. I think personally, Google's VO and imagine three and power marketers to revolution. I visit visual storytelling and. Enabling the creation of engaging high quality video and image content that resonates with audiences while streaming streamlining production processes, which, if we compare to the original output that Perplexity gave me, those innovations unlock new possibilities for dynamic visual storytelling. It's not awful, but I prefer the Gemini output.

Isar:

So again, going back to what we said before, which I think is a good summary, is that you're not locked into any of these components or the, any of these tools, and you can use different tools for different steps of the process that do that step of the process better. I'll say something a little more extreme for people who are a little more advanced. You can use tools like. Zapier or make to actually automate this entire process across these different tools, meaning you can drop the first thing into, into Perplexity. And then have Zapier grab the output from Perplexity and drop it into Gemini to write the second component and then put that into wherever you want to put it. let's say a Google doc so you can then edit that or whatever. And then you, all you have to do is go through that step. So you don't even Once you perfect the steps, once you figure out which step needs to go, where you can even go through the automation of all of that. So you end up with a draft somewhere that you can read once and fix versus just go back and forth. I think doing what Charlie's doing in the initial steps, when you're still trying to figure it out is very important. Just go through different tools and try it out. there are several different tools. One of them is called chat hub, which is a, Chrome extension that allows you to run all of these all at once. So you can actually run the, your prompts on Gemini and Claude and ChatGPT and, whatever other tools you want, even some open source ones. And just see in parallel, which one is giving you the best results. And then you can just pick and choose and mix and match or whatever you want to do. So that's just another way to approach it.

Charlie Hills:

Wow. You're giving me some homework to do here. yeah. this one's awesome.

Isar:

You're going to, for what you're doing, you're going to absolutely love this. it's a Chrome extension. It literally just opens a bunch of them in parallel, and then you just run the prompts through them and you continue the conversation with all of them all at once. And it's really, Charlie, this was. really awesome. I think your process is amazing. I think it's very well thought after. I think the fact that you keep on experimenting is what really drives the success that you're seeing, both experimenting and finessing the prompts as well as experimenting in the different tools. And the combination shows, like you said, you saved 80 percent of the time and you're actually getting better results. Like you're getting higher read readerships and a growing, successful newsletter. As I mentioned before, if you're not following Charlie, look for Charlie Hills on LinkedIn. He's sharing really great, valuable content. if people want to follow you or know about you or work with you beyond, LinkedIn, what are other ways to do that?

Charlie Hills:

Oh, that's a very good question, which is, LinkedIn is my main focus at the minute, but, keep an eye out cause you don't know what the future holds.

Isar:

Okay. This was really great. thank you so much. I really appreciate you taking the time and sharing this information with us.

Charlie Hills:

Thank you for having me. It's been an absolute blast sharing it with you. And I hope your audience enjoyed it as much as I did sharing. So yeah, cheers.

Isar:

Thank you. Have an awesome rest of your day, everybody. Thank you for joining us.

People on this episode

Head Shot

Isar Meitis

Host