Introduction
In a world buzzing with AI hype, can it really create the quality, nuanced long-form content we crave—without any human magic?
Join Mikael Svaal, the brain behind Muntanus, in this episode of The Content Universe as he tackles this burning question.
As companies jump on the AI train with tools like ChatGPT, Mikael breaks down the challenges and limits of using AI for complex, specialized topics.
He’ll share his secrets for striking the perfect balance between AI efficiency and that irreplaceable human touch—because let’s face it, good content needs both!
Ready to discover a proven method that merges AI with human creativity? Whether you’re all in on AI or still on the fence, this episode is packed with fresh insights that will up your content game.
Don’t miss your chance to supercharge your long-form content strategy—tune in now!
What You’ll Learn
1. Discover effective techniques for leveraging ChatGPT in long-form content creation.
2. Uncover the limitations of AI-generated content and when manual work is essential.
3. Learn the process of crafting engaging and tailored content with AI assistance.
4. Find out how to integrate AI and human input for high-quality content results.
5. Explore the role of interview transcriptions in enhancing blog post authenticity.
6. Understand the significance of refining AI drafts to achieve a conversational tone.
Table of Contents
00:00 Welcome and Introduction to Long-Form Content
01:12 The Role of AI in Content Creation
02:45 Common Misconceptions About AI Writing Capabilities
04:10 Importance of Unique Angles in Content
05:55 Collaboration with Sales and Marketing for Ideas
08:20 Transforming Ideas into Journalistic Angles
10:15 Conducting Expert Interviews for High-Quality Content
12:45 The Process of Transcription and Outline Creation
14:30 Utilizing ChatGPT as a Writing Assistant
17:05 Refining the Draft with Human Input
19:00 Final Editing for Coherence and Human Touch
21:15 The Value of Combining AI and Human Efforts
23:00 Closing Thoughts and Key Takeaways
Episode Transcript
Hello and welcome to The Content Universe. Today, I’ve been working a lot, and actually this past week, I’ve been working a lot with long-form content. And, I think long-form content, that be articles, that be blog posts, and the like, I think long-form content, it’s really an interesting thing to do, especially now that everyone in this field are so in love with AI. Now, what I want to talk about today is what you can do with AI, what you can’t do with AI, or at least in my experience, what I can do and what I can’t do with AI. And, that AI in particular is going to be ChatGPT because that is the one that I’m using myself for the most part. Now, I do use ChatGPT for both client work and work for myself. I do use ChatGPT to make drafts, even to make finished work sometimes. But, most often it’s going to be drafts and especially with this long-form content, I just really have an experience that it’s just one of the hard things to get ChatGPT to do really well. Now, when I look at Youtube videos and when I hear people talk about ChatGPT, there is this conception that, or there’s this perception that you can basically just plugin a title of whatever you want, whatever blog post you want to create and then voila, ChatGPT will spit out the entire article and you’re basically good to go. You can copy and paste it into your website, and just click post and you have a full article on whatever you do. Now this may be true for some businesses, this may be true for businesses that sell or that want to blog about really simple things, really trivial things, things that already are really well covered online. Now where it becomes a little trickier in my experience is when you want to write something for a company that needs to post or write about something a little bit more complex. Something maybe even a little bit more unique, not necessarily as in the content is unique, but maybe the tone of voice is unique, maybe they have a unique opinion, maybe they take a specific stance on whatever problem in the industry. Now when you want to do something that’s a little bit more specific, ChatUPT is really struggling especially if you give it nothing more than say target audience, and maybe the header of whatever you want to do. Now so what I want to do today is I want to kind of take you through what we do, what I do, and we do at my company in order to kind of use ChatUPT for what it’s really good at, but then still working, and I’m sorry about this because it’s gonna be a bit of manual work, i know it’s i know it’s terrible, but it’s a bit of manual work, and I want to go into how this manual work is, it’s basically or how the work is structured for you to use or to produce high quality blog posts, and it is quicker now than it was November two years ago, where ChatUPT just came out. It is quicker now, but it’s also just a completely different ball game, because you have so much more blog posts coming out because people use AI to write with. So that is a whole episode for itself, that discussion between quality and quantity, anyway I want to take you through our process. Okay let’s dive in. By the way, my name is Mikael Svaal and I am the owner of Muntanus, which is of course a content production company. We do podcasts, we do vlogs, we do video interviews, we do artists of all sorts and of course social media. Now with that aside, let’s dive in. So, first of all, it all starts with creating the angle of whatever that we need to blog about. We need to know what is, what are the stories that we need to tell? What are the points of the statements and what are the ideas that we need to convey to our audience? And that is actually something that takes place in collaboration with sales, it takes place in collaboration with of course marketing for say a client, and then it also takes place with the help of Jack Jay Baty in the ideation part. That I’m sure most of you already know how to do. Because basically you just type in what should I write about when my audience is X, and then Jack Jay Baty will give you a full list of that. So anyway, once you have that, what we do is we turn that into, we turn that, so to say, the demand for knowledge, we turn that into a journalistic angle. And in our case that is a working heading or working title of the blog post. And now here is where the magic happens, because some people think that we can now just take this or at least some people want you to believe that you can then just take that title and then plop it into ChatGPT and make it do all the work. And we found that that is not really possible. So what we want to do is now, we want to plug that title into ChatGPT and ask it to discover the topic, to present to us a question frame that would serve as a cultural orbiter for a good question frame in an interview situation concerning this specific topic. Once we have that, we refine that, and again I’m sorry but that is a manual process, you have to look at the questions, you can’t just take them point blank. You have to look at the questions, you have to qualify the questions, and you have to really work with them for them to be good. And now, as I said, here’s when the magic happens because then we actually have an interview. I know this is crazy, this is like, this is like two-thousand one, two-thousand twenty-one kind of knowledge, you do an interview, it’s so old school but what I find is that when you want to write about something that is high-tech, something that requires specific expert knowledge into some whatever field and the more niche that field is, the more So we need to find someone who actually knows what they’re talking about. And then conduct an interview. And that interview of course can then in turn, become a podcast episode, can become a video, blah blah blah, all this but that’s again for another episode. Now for the purpose right now we just did the interview and when we have conducted the interview we need a transcription. Alright so what do we have now? We have a transcription of an interview concerning the main topic that we want to address in the blog post. What happens now is we sit down with that interview and sometimes such an interview it can go a little bit sideways. Sometimes we elaborate things that are not in the original question frame, sometimes we follow segways down whatever path because it was interesting at that moment, so what we do is we go through the transcription, we look at what’s in the transcription, what is actually said, what quotes can we use, what is talked about, what knowledge comes to surface. And then we choose everything and then we revisit our working headline because like I said, sometimes the working headline for the interview needs to be tweaked a little bit because the interview took another turn than what we expected, so we revisit the the working headline, and sometimes we even divide it up into two or even three headlines for three different blog posts depending on the interview. And then again we go back to the interview, see what we have of material, and then we sit down and we create an outline, an outline of the blog post. What do we need to cover in this blogpost? And if you have a working headline that is say five things you need to know about X, well, a good structure for that would be finding five things, so having an Introduction, having five different Subheadings or H2s, and then have an Outro. That’ll be a pretty good blog post. You’ve seen them before, and it’s pretty good. Now what you can do, you can go to ChatGPT and ask. You just basically put in, pour in the entire transcription and say, chat, come on chat, please give me an outline for a possible blog post. That can work. It does work sometimes, but what I find mostly is that, when you do that, chat GPT tends to create such a wide outline for the article that you can’t really use it, because, for blog posts you want to just tell one single story. You don’t want to tell more than one story, because then people just, you know, the reader just gets confused, and it’s just not good. And that’s pretty much always what happens when you ask chat GPT to create the outline. And I think it has to do with chat GPT, we need to remember that it’s a sophisticated calculator, and not really much more than that. It’s a really sophisticated calculator that calculates the probability of the next word. And that means it can’t really discriminate what is high value knowledge, what is low value knowledge inside an interview. It can connect the title with points in the interview, it can it can connect that okay, these two topics, they’re probably related, they should probably be in the same article, blah, blah, blah. But it doesn’t really know, where’s the value. So that’s, again, where a human being comes in? I’m sorry, again, it’s going to take you a lot of labor. You have to look at it. And you have to also create an outline that you feel comfortable with and that you can see both is inside the interview, but you can also see will become a good blog post. Okay. Step three. Is it three? Two, three, whatever. Next step. You now have the transcription, you have the outline. And of course, you already know who your audience is. And now, this is where I use ChatGPT the most. This is the main step for me when I use large language models and I use ChatGPT. And that is I tell ChatGPT to be my writing assistant. And then I give it the entire interview or the, or the main part of it, the, the, the most relevant part of that interview. So I give it basically the transcription. And then I say, hold on, don’t do anything yet. And then I give it the, the, the outline that we’ve created with the working title and also a purpose of why is that article important, but also a what should the reader get out of this? So an aim of the article, so to say. With that, I still say, okay, hold up, hold up don’t start writing just yet because I also of course give it tone of voice and blah, blah, blah, all this, but I need it to understand that it needs to write in a coherent fashion. So it needs to understand the outline. It needs to know that section one shouldn’t contain points from section five. It needs to know that you can refer back to section one when you’re writing section five. It needs to know that it needs to work, it’s called cohesion so coherent, so it needs to know that it’s needs to work coherence into the text, okay. And then what I tell it is I say okay fair enough you go on and you write section number one and that’s always the introduction and then I write section number one, usually I would tell it to write you know two paragraphs or three paragraphs. I would… sometimes, I would have to repeat the tone of voice so I would say keep it in conversational style, you know be be conversational with the reader. So say you, not they or all those kind of small tricks that you probably already know. So I make it write just the first section and then I tweak that section and make it right in the fashion that I like or that I know that we need for this specific client. And then once that section is is complete or is finished, I go on and say okay thank you for that. Now, go on and write section two. And when it’s done, done with that. I go and say okay go on write section three. So I actually write them like every section for itself and there’s a reason for this and that reason is that then I can it’s easier for me to control the number of sections or the number of paragraphs in each section and also I avoid it just you know writing 100 words for each section which is going to be completely pointless to me. So I’d rather have a lot of output than a little output because it’s easier to delete than it is to write yourself. At least at this point. What happens now is I take all of this output from chaptivity I put it over into my Word document and then for now I’m closing down the AI section of producing an article. Okay so what I’ve been working on today is I’ve been actually working on refining that first draft that I got out of Chatgalopathy for one of my clients and what that means is that I go through the entire article first with the eyes of an editor who looks for coherence who looks for is it the right story, is it are we are we answering the the heading that we set out to answer in the first place are we answering that working title or not. If the answer is no I would have to go all the way back refine my outline and then go through the entire ChatGPT work again to get a new one okay so that’s that’s how I do it and once I once I’m satisfied with this okay so okay we are actually answering whatever outline that we set out to to answer. Now comes a work that puts life into the text and that work is very typical text editor work so that means that you can plot in small conjunctions you can plot in small sentences that couple paragraphs together you would go in and oftentimes I feel like the tone of voice of Chachapiti is a little bit monotonous so to say so it tends to write in sentences varying between 10, 12 maybe 15 words and when you have a text that does that it’s just really hard to read and becomes really boring so I would also go through it and now I would basically shorten some sentences lengthen some other sentences you know connect together some sentences to make them really long and and you know write in really small you know author commentary kind of style things into the text and really basically work the text through and oftentimes This part of editing the article takes at least a couple of hours because you need it to sound human and it’s not because if it doesn’t sound human you know google won’t find it or something it’s because if it doesn’t sound like a human being wrote it no one’s really gonna care reading it because first of all it’s gonna be super boring and super bad you know to read it’s just it’s just boring stuff and also the other thing is that it it might come across as something an AI wrote and if someone if i’m reading something and i’d say oh god an AI wrote this you know i would basically go out. So it’s it’s not for the search engines that we need to make it more human. It’s for the humans we need to make it more human. Okay so what then happens once all this work is done and as you can hear this is quite far from from just plotting in a heading into chat GBT and then copy pasting it directly into your website you know. So it’s what happens after this is of course that it goes to proofreading and and then once we’re really satisfied with the entire article we formatted right. We formatted so it has you know the sub headings it needs it has a title it maybe have it has a subtitle uh you know it has all those kinds of small things and tips and tricks that you would put plot into a normal article, it could be captions for images, all this. Once that is done what you do is you we send it off to the client because now it is actually finished and then what most often happens is that my clients are and myself included we find actually this method of writing with the assistant of chat gpt to make the first draft we actually find this method of writing it produces higher quality posts it produces higher quality articles than if we just of course if we just used ai to do the whole thing it’s way way higher quality but actually it’s also higher quality than if we’d had a human write the entire piece themselves and and you know thinking about that for just a brief moment before closing down today i think that has to do with you know if i’m writing a piece that is 2000 words long you know the first thousand words are just going to be better than the last thousand words simply because i am more energized so i get tired as a human being i get more tired along the way and that means that if i can do all the heavy lifting with an ai i can reserve all my energy for making the text stand out and making it really an easy read for the audience all right that was actually a quite a long a long little vlog here or whatever we call this but that is it for for today’s content universe i really do hope that you enjoyed it and see you around so
