It’s 2025, and authors are at war over AI and whether or not it’s okay to use AI to write stories and design book covers. Relationships have been severed, battle lines have been drawn, and both sides are missing out on one of the most important elements of AI they could agree on.

You see, most authors started writing books because they love to write. Precious few started writing because they wanted to promote their books. And that’s where AI can help. AI can help improve your marketing while saving you time on marketing and giving you more time to write.

Who wouldn’t want to spend less time on book promotion and more time writing?

How can you use AI to help you sell more books? I asked Alexander Macris, author, game creator, and polymath who uses AI tools to market his popular tabletop games.

What AI tools are available for authors?

There are several different AI tools, and you can use them in similar ways, but you don’t have to choose only one tool.

Services such as Straico (affiliate and discount link), which I use, and Merlin (affiliate link) give you access to all of the AI models. With one monthly fee, you can access Claude, GPT 4, DeepSeek, Gemini, and Grok all in one place.  It allows me to compare the responses from each model. Sometimes, I find one model is better for some things than others, but in 90% of the cases, I’ve found OpenAI is the best. GPT is really solid.

However, I’ve found Gemini to be surprisingly good for questions requiring a broad knowledge of internet content. Even though it’s super woke, it’s very good because Google has read every webpage online, so Gemini has that ratcheted down. Grok is quite useful if you need a less censored tool.

Alexander: I agree with that assessment on all those tools. Gemini is the most woke. Chat GPT is generally the most useful, and Grok is where I go for uncensored material.

Thomas: DeepSeek is another source of uncensored AI, at least for Americans.  DeepSeek won’t talk to you about anything going on in China. It won’t talk about opposition to the Uyghurs or massacres in Tibet.

But it also doesn’t care as much about American politics, so it won’t censor much of that information. Merlin and Straico run local versions of DeepSeek so you can access the power of DeepSeek without sharing your interactions with the Chinese Communist Party, which is a big perk.

You don’t want to give the CCP info if you can avoid it.

Alexander: I’ve used a service called abacus.ai, which can tap into multiple engines.

Authors Can Use AI to Help Them Use AI

Thomas: I don’t have a strong recommendation, but I will say that Straico has a built-in tool to help rewrite and improve your prompts in a transparent way. You can tweak it and can toggle it. So, it’s helpful if you’re new to interacting with an AI assistant.

However, any AI can help you write prompts. With each one of these tools, you can say, “I want to accomplish [this goal]. What kind of prompt will help me accomplish that goal?”

For example, you might say, “I want help writing marketing copy for my website. How can I give you a good prompt to give me good marketing copy for my website?” And it will generate a prompt you can tweak to feed back into the AI.

You can use the AI to help you use the AI, which is pretty fun.

Alexander: It’s super meta. You can even use the AI for one step more by saying, “I would like to use you for marketing. What are the ways I could use you for marketing?” It will give you a few ideas, and then you’ll say, “What would be a prompt I could use to…” and then insert one of its suggestions.

Author Opinions and Judgements on AI

My friend group is divided into roughly three groups.

• Group one thinks AI is changing everything and is really leaning into it.
• Group two thinks AI is changing everything, so they’re moving to New Mexico to live in a bunker.
• Group three believes AI is just the new Pets.com, that it will change nothing whatsoever, and that it’s all hype.

I used to consider myself part of group three. But I’ve transitioned. I now think group three is wrong. AI has already become life-changing in terms of its ability to integrate into workflow. The more I dive in, the more convinced I am that I’m either in group one or two.

Thomas: Part of what perpetuates that group-three thinking is that many of them are only using the free versions of AI tools. It would be like saying, “I don’t think cars are that great. I think the horse and buggy is better. I tried driving a car once, and it wasn’t very good.”

But when you ask what kind of car they drove, you find out they drove a golf cart.

A golf cart is only useful in very specific situations, but a horse and buggy might be better in many cases. However, a Toyota Corolla is superior to a horse and buggy in most scenarios. If you’re looking for a romantic ride through Central Park, the horse and buggy is the better choice. But if you need to drive to New Jersey, the Corolla is hands down the better option.

The same applies to AI. If you’ve only tried the free versions, you should test one of the paid versions before making a judgment. The difference in quality between free and paid AI tools is significant.

Learning the Language of AI

Alexander: There’s also a learning curve that comes with using AI. When I was first learning how to use it, it felt like encountering an alien species that had learned to speak English but wasn’t a native speaker. I had to figure out how this alien being thought and how to communicate with it effectively.

But once I learned to translate my requests in a way it understood, I started getting exactly what I needed, almost like it was being beamed down to me from a spaceship. And the more I trained it, the more powerful it became.

Thomas: What really helped me learn to interact with AI was generating images. If you pay for GPT-4, you get access to DALL·E and can ask it to generate images for you.

After blogging about book marketing for 15 years, the temptation is for every blog post image to be a book, a computer, or a typewriter. And after a while, all the images start to look the same. The level of nuance we need for a good image isn’t something you can find on a stock photo site. For a blog image, paying a designer $500 to create something custom doesn’t make sense. That’s why many of the blog images for our Author Media blog have been AI-generated.

With AI, I can say, “Generate an image of a book launching into space like a rocket in a 1920×1080 ratio.” Suddenly, I have a book launching into space like a rocket, which I can use to visually represent a book launch.

Authors use AI to Modify First Iterations

I can use that prompt multiple times and get multiple images. I can also click on the image to highlight a section and say, “Remove this or replace it with something else.” Occasionally, strange artifacts will appear on the image. You can talk to the image to fix it without redoing the whole image.

Alexander: That’s an important element people overlook when using AI. In the first generation of generative AIs, you got what you got, and if you didn’t like it, all you could do was hit refresh and hope for a new run. With these more advanced models, whether generating images or text, you work with the first iteration and target separate sections within the images.

For example, I was researching Roman agriculture and needed to compile a list of all the different Roman agricultural products. Partway through, I realized I had forgotten to include herbs and seasonings. So, I simply asked, “Please make a list of Roman herbs.”

Once I had that list, I said, “Now, please integrate that into the essay you just wrote.” And just like that, it seamlessly incorporated the new information into the essay for me. That’s incredibly powerful.

In my prior life, I used to have employees.

Very often, when you give an intern or a junior employee a task, their first attempt isn’t the final work product you need. It’s part of an iterative process. You sit down with them, review their work, provide feedback, and then they work on a second draft.

The same is true with AI.

The biggest realization that helped me use AI more effectively was understanding that it’s not about mindlessly clicking “retry” over and over, like a monkey, hoping for a better result.

Instead, it’s about figuring out why I didn’t like the initial output and refining my approach by adjusting the prompt in nuanced ways or using the iterative tools available to improve the response.

Thomas: It’s not like snapping your fingers and having an Ikea bookshelf instantly appear, fully assembled. Instead, it’s like using a power drill instead of a screwdriver to build the bookshelf. You still have to take each piece and put it in the right place, but instead of manually twisting each screw, you’re just pressing a button on the drill. The effort is reduced, but you’re still actively involved, pressing that button for every single screw.

AI doesn’t replace you; it just makes you more efficient. However, it might replace some of the tasks you would typically outsource to another person. In that sense, using AI helps insulate you from being replaced by AI.

It’s like farming. If there are four farmers in a town and three of them have tractors while one is still using an ox and a hoe, the ox-and-hoe farmer won’t be able to keep up. Over time, one of the tractor farmers will buy his land because the market has changed, and he just can’t afford to compete.

Authors use AI to Summarize

One of the hardest kinds of writing is sales copy, such as the back cover copy, the blurb, and the one-sentence high concept line. I have found that AI is incredibly good at that type of writing.

Dave Chesson has a free tool called the Amazon Book Description Generator that can turn your bad back cover copy into better back cover copy.

I’ve recently built a suite of AI tools to help authors generate back cover copy from scratch. The tool asks you questions based on my pitch session at the Novel Marketing Conference, where authors learned how to generate a book pitch.

In the session, I posed questions, and the attendees wrote their answers in their workbooks. Each pitch was a combination of all their answers. The process was similar to the formula I presented in my episode on The Nonfiction Pitch Recipe that Works for Fiction and Nonfiction.

Shortly after the conference, I realized I could put the question-and-answer process into an AI interface. So, I created an AI pitch generator that generated surprisingly good copy.

I emailed the conference attendees and had them feed their workbook answers to my different book description generators and give feedback on the results. The feedback has been overwhelmingly positive. Nearly everyone has found it useful, and some have been blown away by its usefulness. Others couldn’t use the result it generated, but it did provide a starting point for them.

The real challenge is that you know too much about your book. You’re standing inside the bottle, and you can’t read the label. You don’t even know where to get started. The four different pitch tools I created approach the pitch in slightly different ways.

I encourage everyone to use all four pitch formulas to see what will happen. The tool will generate three paragraphs of back cover copy, a one-paragraph blurb, and a one-sentence tagline-like pitch. That one sentence is often the best and most interesting bit because it’s often a punchy sentence you can immediately use to promote your book.

Alexander: I’ve been thinking a lot about why AI is so effective at marketing, and I think it’s because AI is trained on what we might call tropes, slogans, and catchphrases. It’s designed to predict the most likely next word based on patterns it has learned.

Marketing works in a similar way. The challenge is that if you’re too innovative, people won’t understand what you’re trying to say. AI naturally gravitates toward messaging that feels familiar and recognizable, making it surprisingly good at crafting marketing copy that resonates with people.

If your book description is filled with other-worldly names like “When the Gwala Magath attacks Kiza Nakamu…” the reader has no idea what’s going on. Your copy can’t be too unique or too predictable.

You need just the right amount of unpredictability and sizzle presented in a familiar way.

For example, if you go to an expensive restaurant, the food is presented in a particular way. Menu writers use descriptive adjectives, phrases like “drizzled with,” and other flourishes that create a sense of luxury.

If the menu didn’t follow that style, you might question whether the restaurant was actually high-end. At the same time, you still want that unique combination.

The way AI is trained makes it perfect for this kind of task. It’s as if it were designed for it, and I’ve found it incredibly powerful.

Thomas: While AI may not have read all the books on Amazon, it has read all the book descriptions. AI has read millions of book descriptions. It knows what a good book description looks like and can help you generate one.

You don’t even have to use my tools. If you use the paid version of an AI program, you can describe your book in your own language. Talk about your protagonist and what he wants. Explain the obstacles and the antagonist. Then prompt the AI to “Use this description to generate a book description that would make this book appealing for an Amazon customer.”

You’d be surprised how good the response will be. You might even start with a prompt like “I want to create back cover copy for my book. How do I do that?” The AI will suggest prompts that you can feed it, and it will walk you through the process of generating your back cover copy. It can hold your hand through that process, or you can become a patron of the Novel Marketing Podcast and use one of my tools.

Use AI to Brainstorm Titles

Alexander: AI is also incredibly useful for creating titles, especially for books in a series. For example, I’m working on a fantasy series with my friend Jonathan Oldenburg. We knew the first book’s title, but we weren’t sure what the others in the series should be called. We had a sense that there must be a method to it, some kind of pattern that successful series follow.

So, we turned to ChatGPT and asked it to analyze best-selling fantasy novels and series. We quickly discovered that there are certain commonalities. Series titles tend to follow a consistent format. Take George R. R. Martin’s books, for example, Game of Thrones, Clash of Kings, and Feast of Crows all follow a recognizable structure. There are multiple patterns like this used in fantasy series.

One that stood out to us was the use of iambic pentameter. So, we uploaded the novel to AI and asked it to generate 50 potential titles using iambic pentameter, drawing from the book’s themes and content. Suddenly, we had a whole list of amazing, cohesive titles that matched the patterns of bestsellers.

What would have taken an entire day was done in just two hours.

Thomas: We do the same thing with our episode titles. Historically, episode titles haven’t been that important for traditional podcasts. People subscribe because they like the content and automatically get the next episode, so titles don’t require much thought. That’s why podcast episodes were often numbered as Episode 123 or 124.

But that’s not the case on YouTube. Now that we’re publishing our content on YouTube, titles, and thumbnails are critical.

I’ve found that I can feed the entire transcript of an episode into ChatGPT and prompt it to generate potential YouTube titles based on the text.

I don’t always use its suggestions word for word, but the process helps me think in new ways and often leads me to a much better title that performs significantly better. Episodes with titles generated by my prompts and AI’s results get more views on YouTube than episodes with titles I came up with on my own.

I’m sure some people are naturally great at crafting YouTube titles without AI, but I’m not one of them. Even though I’ve studied blog titles for years, the combination of my own knowledge plus AI generates better results than me working alone.

Authors Use AI to Improve Book Covers

Thomas: Let’s talk about book covers. I’m not necessarily recommending using AI to design your cover, but you can use AI to create a better design brief for your human-designed book cover.

Years ago, I did an episode on How to Create a Design Brief for Your Book Cover. When I revisited it, I realized the process was very step-by-step. I figured I could input the questions and steps into AI and ask it to generate a design brief. I tried it, and it worked surprisingly well. AI can create a solid first draft of the brief, which you can then tweak before sending it to your designer.

This doesn’t replace human creativity, but an AI-assisted design brief makes the designer’s job easier and improves communication between the author and the designer. Many authors struggle to articulate what they want their cover to look like. Most authors aren’t fluent in the language of design. AI is an interpreter, helping translate your vision into terms your designer can understand.

Alexander: Absolutely. As you know, I create graphic novels, and one of my biggest challenges is translating the words I write into image concepts for an artist who lives across the globe and whom I’ve never met in person.

One of the hurdles is that artists have a very specific visual vocabulary that I wasn’t trained in. For example, terms like “medium three-quarters crowd shot” have a precise meaning to an artist, but when I first started, I had no idea what that meant. “Upshot” and “downshot” are terms that cinematic designers use. Even terminology related to color, like “dynamic lighting,” was unfamiliar to me.

I’ve learned that if I find a comic image I like, I can show it to ChatGPT and say, “Explain this image as I would describe it to a human artist.” AI translates it into the right artistic terminology, which has been an invaluable tool for helping me communicate more clearly. It has tutored me in the language of design so that my descriptions are accurate. I can even ask it to create a mockup.

I’ve also used AI to brainstorm cover ideas. Just as your blog images tend to be the same, fantasy book covers have the same problem. Every fantasy cover for men ends up being the hero standing nobly on an epic landscape, holding a weapon.

But with AI, I can say, “This is what my book is about. Give me 20 ideas for a cover.” I can even upload the entire book and ask it to generate cover concepts based on the story.

It’s incredible.

Thomas: With most of the paid AI tools, you can feed your entire book to the AI to give it context for all the questions you want to ask it. You simply upload your current draft of the entire book to GPT. This isn’t possible with the free version, but most paid AI tools can handle full book uploads. Once uploaded, the AI had the full context, which allows you to ask detailed questions about your book.

One powerful use for this is generating a “book Bible.” If you need a reference guide for your series, you can ask AI to list all your characters, where they live, how tall they are, and other key details. It pulls this information directly from what you’ve already written and helps you maintain consistency across future books.

Another useful application is generating a book pitch. You can upload your entire book and ask AI to create a compelling pitch based on what it has read.

You could also request a book cover design brief. Most cover designers don’t read the books they design for. If a designer charges $100 per hour, and reading the book takes eight hours, that’s an extra $800, which most authors don’t want to pay. Plus, most designers don’t even offer that as a service. AI, however, can read your book and generate a detailed design brief in minutes.

Having an AI assistant that has read your book is incredibly useful. Most people in the publishing industry don’t have the time to read every book they work on, but AI can, and that makes it a valuable tool for authors.

Use AI to Create an Assistant

Alexander: It’s worth noting that when you pay for ChatGPT, you can create your own GPTs and Projects.

Create Your Own GPT

For instance, I created an ACKS GPT, which is my own GPT I trained on everything I wrote for my ACKS game. If I don’t remember my own rule, I can just ask my GPT for it.

Projects

You can also more easily create Projects. For example, you could upload your novel and create a project called “My Novel,” where all the information about your book stays in that GPT project. The Project can cross reference itself, and you can upload more knowledge to it. Setting it up takes about 30 seconds, but it’s not available on the free version of ChatGPT.

Authors Can Use AI for Emails

Thomas: When you upload your book, GPT  “reads” it, so you can ask it to generate content based on your book. You could then prompt it with the following:

• Generate an email announcing this book to my email list.
• Generate an email to my subscribers teasing the characters in this book.
• Write an email teasing the inciting moment.
• Write a book description for an Amazon sales page based on this book.

You won’t paste those AI-generated emails directly into your email marketing service. They’ll still need your human touch and tweaking. Remember, AI is the drill to help you quickly assemble a bookshelf; it’s not a magic wand that makes a bookshelf appear. It’s the drill that drives the screw and helps you assemble the bookshelf more efficiently.

For many authors, it’s easier to edit existing marketing copy than to create it from scratch.

Authors Can Use AI for Web Copy

Thomas: Just as you use AI for your marketing emails or onboarding campaigns, you can also use it for your website.

I had a course that originally started as part of Obscure No More, which is my big platform-building course. Within it, there was a section on marketing psychology that kept growing. Eventually, because of how Teachable works, it was easier to pull out the marketing psychology section and create a standalone course for it, which I called Marketing Psychology for Authors.

At first, I didn’t market or promote it because it was intended exclusively for Obscure No More students. But when Black Friday came around, I thought, “I should promote this.” The problem was that the course landing page had only a few paragraphs of terrible placeholder text that I had thrown together quickly.

However, I had transcripts from every session of the course. So, I realized I could copy and paste the course content into GPT. Then, I asked GPT, “In the style of Neville Medhora, write a landing page for this course.”

What it generated was surprisingly good. I ended up using about 90% of what it gave me and made only a few tweaks. Once the page was live, I started getting sales on a course that had never sold before because I had never promoted it. People were happy with the course.

Those sales came from a sales page copy that AI generated for me in just a few minutes. It wasn’t a long back-and-forth iteration process. My first prompt gave me something solid to work with. I made a few adjustments and suddenly had a decent landing page. Could I have refined it further? Absolutely, and I probably still should. But GPT is incredibly effective when it comes to getting from zero to something.

Authors Can Use AI for a Synopsis

Thomas: For those who need a plot synopsis as part of your proposal for a traditional publisher, you can upload your book to ChatGPT and ask it to generate a plot synopsis. Then, you can tweak it and send it to your publisher. It will save you days of hassle.

Authors Can Use AI to Find Their Market

Alexander: Another area where AI is incredibly useful in helping you determine which audience to market to or even how to write to market.

I took Ascendant: Star-Spangled Squadron and uploaded it to ChatGPT when I was creating my Ascendant project. I asked, “What is the plot of Ascendant: Star-Spangled Squadron?” it generated a one-page summary.

The summary was impressive. It perfectly captured all the key story arcs, theme, and tone. So next, I asked, “What audience would this book most likely appeal to?”

ChatGPT broke it down into several audience segments:

• Fans of superhero comics with military and political themes. It listed similar books like The Authority, The Ultimates, and Watchmen, which explore superheroes through the lens of government oversight and military intervention.
• Readers who enjoy team-based superhero action, similar to Marvel’s Secret Wars or DC’s Justice League.
• Libertarian and conservative-leaning readers who feel mainstream comics have shifted too heavily into progressive messaging. AI suggested they might find Ascendant a refreshing alternative.
• Fans of gritty, realistic superhero stories with dark consequences, such as The Boys, especially given how trauma triggers superhuman ascension in Ascendant.

It went on to identify multiple potential fan bases. Once I had that information, I could then ask, “Now create an ad campaign for this group of fans,” or “For this specific audience, which character should I focus on in my ad?”

This is the kind of work I used to do as a marketing consultant in the video game industry, and companies paid me tens of thousands of dollars. My entire business revolved around identifying audiences and tailoring marketing strategies for video games. And now, with AI, you can generate insights at the snap of your fingers.

Thomas: Let’s say you want to run Facebook ads but aren’t sure who to target. Going through this process helps you determine your ideal audience and clarifies which aspects of your book to highlight in your pitch.

Different elements of a story appeal to different types of readers. With AI, you can create highly tailored messaging for each audience segment. No author could afford to hire a consulting firm for $30,000 to build out these detailed reader personas. Books simply don’t have high enough margins to justify that level of investment. Even New York Times bestselling authors don’t get that kind of marketing support.

But now, for as little as $10 or $20 a month, whatever you’re paying for your AI tools, you’re getting the equivalent of tens of thousands of dollars in marketing consulting. All it takes is uploading your book and asking the right questions.

Authors Can Use AI to Write to Market

Alexander: Another area where AI can be incredibly useful is in writing to market. Traditional marketing focuses on promoting what you’ve already written, but writing to market flips that approach; you write a book tailored to an existing audience. When done well, the book practically sells itself because it aligns so well with what readers are already looking for.

You can prompt ChatGPT by saying, “This is the audience I want to reach. What kind of book should I write to appeal to them?” Or you can take a more personal approach and say, “Here’s my background, the topics I know a lot about, and my life experiences. What audience would I be most likely to connect with in a fiction book inspired by these elements? What genre would best fit my experiences?”

AI can generate detailed insights to help guide your decision-making. This level of market research used to require hiring high-end marketing consultants from Madison Avenue, costing thousands of dollars. Now, AI can do it for a fraction of the price.

And yes, sometimes AI makes mistakes. But you know what else makes mistakes? The human you hire.

Coach AI to Help it Help You

Thomas: When you correct someone’s mistakes, you risk hurting their feelings. As a boss, you’re acting as a coach, helping people improve and perform better. It’s not enough to simply point out what they did wrong; you must provide feedback in an encouraging and uplifting way. People don’t like feeling like failures, and some may even leave their jobs if they don’t feel like they’re winning.

While you don’t need to navigate the emotional side of things with AI, you still need to approach coaching it in a similar way. Don’t fire the AI (or the human employee) the first time it makes a mistake. Relationships are built over time, and unless there’s serious misconduct like embezzlement, give your employees more chances and coaching. Many mistakes are simply due to poor communication.

Assign AI a Specific Role

When you’re prompting AI initially, it’s easy to forget to provide enough context. The AI only knows what you tell it, so if you say, “Give me back cover copy for my book,” it might ask, “What is your book?”

As you interact with the AI, you learn to add additional context because things that seem obvious to you aren’t obvious to the AI. One useful tip is to specify the role you want the AI to fill when creating a prompt. For example, say, “Act as a marketing consultant and write copy for this book.” You want help with marketing, but without specifying the role, AI might think you’re asking for a literary analysis from a professor.

The AI could approach your question in many ways, so clearly stating the role you want it to play makes a big difference. This strategy dramatically increases the quality of responses. I now try to include the role I want it to play in almost every prompt. For example, if you’re writing a medical scene, you might say, “Act as a doctor and give me your professional opinion on this medical scene.”

Prompting AI

Alexander: I have been reduced to screaming frustration in the past because I failed to understand that the AI considers everything in its context. I don’t always realize how it’s appraising some details. For example, I was trying to mock up an image of a warrior without a weapon in his hands. It drew a picture I liked, but the warrior had a sword in his hand.

So, I adjusted the prompt to remove the sword, instructing that “the warrior has empty hands.” Yet, when I ran it again, it put a sword in the guy’s hand. This cycle repeated until I was red in the face, practically screaming at the AI as if it had a personal agenda against me.

Then my wife looked at the image and said, “If the guy doesn’t have a sword, why does he have an empty scabbard hanging at his belt?” I looked at her and thought, “Oh.” So, I told the AI, “The man doesn’t have a scabbard.” When I ran it again, the warrior was drawn with empty hands. In the AI’s mind, if there’s an empty scabbard, the sword is in the man’s hand.

That sort of thing happens all the time. Often, you get the wrong results because of one little thing you said wrong earlier in your prompt. If you just tweak it, it works.

Thomas: I think prompting with the same prompt over and over again can lead to frustration. Occasionally, you’re just trying to use that randomizer to get you what you want, and you get close, and it can work. But when possible, it’s better to revisit your prompt and look for ways to improve it.

When Straico rewrites your prompt, it guesses what role you want it to fill (which is where I learned that tactic). Sometimes, it’s useful because you realize you don’t want it to play a certain role. For instance, you might realize you don’t want it to approach the text as a professor, so you say, “Act as an advertising or PR person.” By fiddling with the prompt, I get much better responses from the AI.

When in doubt, revisit your prompt. The revision may yield much better results.

Speaking of prompts, I want to share an amazing prompt I got from Shauna, who does the blog post versions of all these episodes. I had a great interview with her about how she builds blogs from podcast episodes. If you’re a podcaster wanting to turn your podcast into a blog, I encourage you to listen to that episode.

She uses AI as part of her process and does a great job using it to make the raw audio transcripts clear and readable. Occasionally, she takes sections of the transcript and prompts the AI to “rewrite this text with clarity and use most of the original wording wherever possible.”

That phrase “use the original wording whenever possible” is like magic. I don’t want my words to get lost in the blog version; I want it to be a written blog, not just a transcript. Transcripts can be annoyingly hard to read and very difficult to skim. However, the blog post still needs to sound like Thomas, so including that phrase in your prompt is essential when converting rambling dialogue into a written blog.

Transcription services might insert periods and paragraph breaks but won’t fix a rambling sentence. ChatGPT can fix it. This approach saves Shauna a lot of time generating the blog versions and allows her to focus on the aspects of creating blog posts that only a human can do.

Use AI for Podcast Questions

My team uses AI to generate questions for podcast interviews. Often, Laurie, our producer, generates the initial list of questions. I’ve found that having a bank of questions for a podcast interview is quite helpful, but the most important question you can ask in an interview is a follow-up question that only you, the human, can ask.

So, I try not to be too dependent on my question bank because follow-up questions are more useful for the interview. Still, having some good backup questions is very beneficial.

Alexander: I agree. Although, as AIs become more and more powerful, they will be able to generate dynamic follow-up questions.

Have you played with Google’s NotebookLM podcast at all?

Thomas: I haven’t.

Alexander: It’s pretty wild. You can upload something, and it will generate a podcast of two people discussing whatever you uploaded. For example, I uploaded my Adventurer Conqueror King System book for my role-playing game to NotebookLM, and it generated a 15-minute podcast of two people discussing my game.

However, since Adventurer Conqueror King System is hilariously complex, the AIs are getting it all wrong. Everything from my name to the game name. They recently added a new feature where you can dial into the podcast as a caller and talk to the AIs about the product.

I dialed in and said, “I’m the creator of the game, and you’re pronouncing it wrong.” Without missing a beat, the AI podcast hosts said, “Wow, that’s cool! We’ve got the creator of the book here on with us!” It occurred to me that an author could use this tool to practice doing podcast interviews.

Many people get intimidated by podcast interviews, so they turn down opportunities, but NotebookLM offers a way to practice.

Thomas: Another technique you can use is to give ChatGPT a prompt such as, “I’m going to be a guest on the Novel Marketing podcast with Thomas Umstattd, Jr. talking about such and such topic. What questions is he likely going to ask me?” The AI will then generate some questions you can practice on.

We never give the questions ahead of time because I’ve found that when I’m interviewing an author, if I provide the questions beforehand, the author writes out their answers and then reads them to me, and it no longer feels like a conversation.

I like this technique because those questions may be similar to the ones I’ll ask, which will allow you to practice and get more comfortable. Depending on how popular the podcast is and how well-trained the AI is on that particular podcast, it may nail it.

I recently built a tool called AI Thomas, which contains the blog versions of all 450 episodes of Novel Marketing and over 100 episodes of The Christian Publishing Show. We loaded all the blog posts into a model that allows users to ask AI Thomas questions. It then generates answers based on my episodes and cites its sources.

For each paragraph, AI Thomas provides a citation linking back to the specific episodes it pulled from so you can click the link and see the full context. So far, it has never hallucinated. I’m not saying it can’t hallucinate, but the fact that it cites its sources keeps it on track. This means it genuinely answers questions based on my insights using either my exact words or the blog versions of them.

I’ve found it incredibly helpful, especially for topics I haven’t dedicated a full episode to. Even if I haven’t recorded an episode on a topic, I may have mentioned it repeatedly across different episodes. For example, we recently had a question during a Patreon or mastermind call about KingSumo (affiliate link), a tool I reference all the time. Someone asked, “Thomas, what tips do you have on KingSumo?” At first, I thought I need to do an entire episode on this. But then I suggested they ask AI Thomas. When they did, it compiled all my tips from various episodes and produced a 1,200-word guide on how authors can use KingSumo. It was amazing!

Another great use case is book recommendations. I’ve never done an entire episode dedicated to recommending books, but I frequently mention books on different topics. AI Thomas can pull together all my various book recommendations and present them in a neatly organized list.

Watching people use it is also fun because I can see all the interactions. If I notice that certain queries aren’t generating good answers, I realize I might need to do an episode on that topic to strengthen the AI’s responses.

Building AI Thomas was expensive. The AI tool wasn’t, but generating all those blog posts took thousands of hours of work from Shauna and me. Creating those blog posts was a massive effort. In some ways, AI Thomas is even more useful than the blog because it synthesizes all that content and delivers the answer you need for your question.

AI for Book Cover Analysis

I also built a book cover analyzer. You upload your book cover, and it analyzes the design, predicting what genre it belongs to, identifying the major design elements, and describing the book based on its cover. It also suggests who it thinks your target audience is and provides tips on how to improve the cover.

While developing it, I struggled with getting it to provide truly useful feedback. We tested it on both good and bad book covers, but it was too polite on terrible covers. I didn’t want to give people false hope, so I created two versions: one that gives general feedback and another that’s a bit harsher. People tend to prefer the friendlier version, but I think the harsher one is probably more helpful.

My entire suite of AI tools for authors is available exclusively to Novel Marketing patrons.

The Book Cover Analyzer is surprisingly good. I was shocked at how accurately it could detect the genre of the book based on its cover.

Alexander: If it can’t detect the genre from the cover, you’ve gone really wrong.

Thomas: Our hope was that it would get the answer wrong most of the time, prompting people to realize, Oh, wait, this isn’t a thriller, it’s a mystery. This cover is far too exciting. However, it has been getting the answer right more often, which is less helpful. That said, it’s also possible that people are simply uploading well-designed book covers that accurately reflect their genres.

AI for Data Analysis

Sometimes, as a marketer, you receive data in the form of clear charts and graphs, but other times, the data is messy and overwhelming. This is where AI is incredibly powerful. It excels at finding the needle in the haystack. That’s why DOGE is uncovering so many inefficiencies; they can sift through massive amounts of data and find data that stands out. They’re using AI-powered tools to analyze treasury data and finding all kinds of stuff.

It’s similar to what happened a couple of hundred years ago when businesses switched from single-entry to double-entry bookkeeping. Suddenly, accounting fraud that had been hidden was being exposed.

Data Analysis: A Case Study

After the Novel Marketing Conference, we surveyed all attendees using a Google Form. The survey included objective questions, like rating various aspects of the event on a numbered scale, but it also had short answer fields where people could write detailed responses. Since writers love to write, we ended up with hundreds of pages of feedback about what people liked and didn’t like and suggestions for improvement.

One of our first questions was, “Were you glad you came?” Every respondent said “yes.” That was encouraging but not very actionable. That’s why we asked the other questions. I exported the survey results as a CSV file, which contained responses from over 50 attendees. Then, I uploaded it into Straico and tested multiple AI tools to see which would provide the best analysis.

They all did an excellent job. Each tool identified the key areas for improvement, highlighted what attendees loved and what shouldn’t be changed, and generated a detailed report to help us improve next year’s conference. The insights were incredibly valuable. I think GPT provided the best analysis, but the other tools weren’t far behind.

Going through spreadsheets and extracting meaningful takeaways is tedious and time-consuming for a human. When 50 people say the same thing and one person phrases it differently, it’s easy to miss those nuances. AI, however, can quickly scan and pull out the most valuable insights with remarkable accuracy.

Anytime you find yourself overwhelmed by a spreadsheet, it’s an opportunity to let AI step in. You can simply tell it, “This spreadsheet is overwhelming me. I want to sell more books. How can I use this data to do that?” Even a simple prompt like that can lead to valuable insights. AI can analyze sales data from your email list, ad campaigns, or reader surveys.

Speaking of surveys, I have an entire episode on reader surveys. Sending a survey to your readers can be a game-changer for your career. Today, you can upload your survey data to GPT for analysis. For quantitative responses, Google Forms can generate charts, but when it comes to analyzing open-ended, essay-style answers, AI shines.

Coming Soon: AI Agents?

Alexander: I think the next big breakthrough will be AI agents. When using AI, you have to consciously decide, “I’m going to do this project, and I’ll use AI to help me.” With AI agents, we’ll be able to set up virtual personal assistants that handle ongoing tasks automatically, without needing to be prompted, almost as if they have a level of operational agency.

Right now, you can set up Google Alerts for certain keywords to stay updated on relevant news. But soon, you’ll be able to create an AI agent that scans the best blogs, YouTube channels, and other sources on book marketing or your genre of interest. It will compile a daily list of the most important content you should pay attention to. It will highlight emerging trends and explain their implications for your work.

I can ask AI to “write me three tweets to promote myself.” But before long, I’ll be able to tell an AI agent, “Generate three tweets, pick the best one, and tweet it daily.” It will handle the entire process, and I can simply check in to see how my feed is performing.

This shift is going to be profound.

Thomas: Some of this is just making more work for AI on both ends of the process. For example, you can ask AI to rewrite something simple and make it more flowery and fancy. On the other hand, you can also use AI to simplify flowery language. You might prompt AI to “write an email based on this summary,” and when someone receives that email, they might then prompt AI to “summarize this email.”

It can be both comical and fun, and I think that’s the right way to approach it. Too many people have watched 1980s sci-fi films and assume AI will rise and destroy humanity. But AI isn’t going to come and kill you; it has no will of its own.

The real concern is how other humans will use AI. The best way to protect yourself from bad actors using AI is to become fluent in AI yourself. The more you understand how it works and the more experience you have interacting with it, the better equipped you’ll be to protect yourself from bad people and from well-intentioned people who use AI in ways that affect you.

The Scariest Form of AI

The scariest form of AI isn’t the tool you use to analyze spreadsheets or improve your marketing efficiency. It’s the AI that companies like Meta use to decide what you see on social media. It’s TikTok’s algorithm, which can shape your worldview simply by choosing which video appears next in your feed.

Your use of AI to improve marketing, analyze CSV data, or streamline spreadsheet analysis isn’t going to end the world. But it will help you spend more time writing your book, and ultimately, that’s the goal.

Patterns of Human Technological Development

Alexander: I think of it very similarly to the invention of firearms and what a small personal firearm meant for a warrior. Imagine you’ve spent your whole life training to fight with a sword and shield, and then you run into someone with a six-gun who can kill you in an instant. Suddenly, he’s more lethal than you are.

But do you just throw up your hands and say, “Well, I guess none of my training matters anymore”? No, because a fit, elite warrior with a six-gun will still be a better warrior than Joe Schmo, who just picks one up and shoots blindly.

It’s another tool in the arsenal. And now, you can defend yourself in situations where before you couldn’t. It can also serve as an equalizer, covering for certain weaknesses you might have had before.

Thomas: Although it’s painful to have just spent a year’s wages on a fancy suit of armor only to find out that a peasant with some musket can poke a hole through that armor.

Alexander: That’s the history of human technological development. I was reflecting on that while preparing for this podcast. Here I am, typing on a keyboard instead of writing by hand. When I print books, they’re automatically printed, but there was once an entire industry of people whose job was to handwrite books. Scribes dedicated their lives to copying texts, and illuminators had a profession just for decorating the edges of manuscripts. Today, we don’t even think about those jobs. Our lives are built on an edifice of replaced professions that once existed but are now obsolete.

Another thought that came to mind is how, twice in my career, I’ve witnessed new technology emerge that caused controversy and outrage. One example that stands out is the introduction of pay-to-play micropayments in video games. When they first appeared, the backlash was intense. People claimed it was ruining the gaming industry, destroying gameplay, and that anyone who designed such games was terrible. Some even said, “You’re dead to me if you create a game like this.”

Fast forward 20 years, and those who pioneered those games are now billionaires. Everyone plays and buys these games, including many of the same people who were complaining about them two decades ago.

Thomas: People still complain about pay-to-play micropayments, but the fact is, they worked. I remember hearing a developer at Blizzard talk about how he and his team spent years working on StarCraft II. It was a masterpiece, a critically acclaimed game that made hundreds of millions of dollars. Yet, just a few months after its release, Blizzard made the same amount of money from a World of Warcraft microtransaction that only took a few development hours to create. That was the moment he realized the entire gaming industry had changed.

You can try to fight against change, and in some cases, there’s value in that. There’s an important conversation to be had between those trying to preserve tradition and those pushing new developments. That dialogue is necessary. But it needs to be a conversation. If the response to disagreement is to attack, cancel, or silence people, that isn’t productive or helpful.

AI is here, and there’s no going back. You can’t put the horses back in the barn or the toothpaste back in the tube. And the reality is that AI already has good and practical uses that most people already take advantage of. Social media runs on AI. Most people are using, or quickly switching to, AI-powered spellcheckers. The idea that you can ban AI entirely isn’t realistic, and part of the reason these so-called honor codes against AI don’t work is that the people writing them don’t fully understand what is AI and what isn’t.

If you’ve ever posted about your book on social media, you’ve used AI. The irony is that the same people condemning AI on Facebook are using AI-powered algorithms just by being on the platform. AI itself isn’t evil; people are. As a true Texan, my approach to technology has always been this: it’s not about good or bad technologies; it’s about good or bad people.

What stops a bad person with AI? A good person with AI. So be that good person. Learn how to use AI. Share your AI marketing strategies in the comments. People are constantly discovering new techniques and better prompts.

We even have a dedicated space for discussing AI on AuthorMedia.social, where people can share their findings and ask questions in a supportive environment. We don’t have AI extremists trying to burn people at the stake for using it. If you’re curious and want to explore more, check out AuthorMedia.Social.

And most importantly, have an attitude of play. Experiment with AI, try things, and have fun with it. You might find that learning and interacting with AI is useful and pretty enjoyable.

Connect with Alexander Macris

• Substack: Tree of Woe
• ACKS Game
• Website: AlexanderMacris.com (Coming soon!)

Related Episodes

• AI for Authors in 2024 – The Good, Bad, and Ugly
• Time-Saving Tools for Authors
• AI Book Covers & ChatGPT
• How to Work with an AI Writing Assistant
• AI Tools for Authors

Sponsor

B.D. Lawrence, author of Chilled to the Bone: A Jake Sledge Mystery (affiliate link)

In River City, trouble isn’t hard to find, and private eye Jake Sledge knows where to find it when a murdered lawyer drags him into a web of political intrigue, neo-Nazi thugs, and bloody showdowns. With sharp wit and hard-hitting action, Jake tackles scumbags the only way he knows how—head-on. 

Fast-paced, gritty, and laced with dark humor, Chilled to the Bone delivers a gripping mystery that will keep you on the edge of your seat until the very last page.                  

B.D. Lawrence, thank you for being a patron! Without patrons like you, there would be no Novel Marketing. 

You can become a Novel Marketing Patron here and get access to my patrons-only AI Tools for Authors.