You know, when you look at the creator economy right now, it really feels like the only way to win is to be the loudest person in the room. Oh, for sure. It's the attention economy. Everyone is chasing viral fame, trying to be the next big thing on TikTok, you know, filming themselves in public, just desperate for that attention. If you aren't visible, you don't exist. That's kind of the rule. That's the common wisdom. Yeah.
But then you read a document like the one we're covering today and you realize there's this. This whole other group of people doing the exact opposite. The quiet ones. They're anonymous. They aren't winning literary awards. They're not doing book tours. But they are quietly, you know, sitting in the shadows, generating somewhere between $800, $1 ,100 a day. Which is just a staggering amount of income for a writer. It's
absolutely wild. Right. And they're not doing it by writing the next, you know, great American novel. They're doing it with these extremely specific niche fiction books, things like Age Gap Werewolf Romance. And they're using a workflow that looks less like writing and, I mean, more like industrial manufacturing. It really is. It's the ultimate example of systems engineering applied to creativity. It's not art in the traditional sense. It's supply chain management. Welcome
back to the Deep Dive. Today we are unpacking a fascinating guide called How to Write and Publish an AI Novel on Amazon. By Maxa. And I want to be clear right off the top, we're not here to debate the ethics of AI art today. That's a whole other deep dive. We're here to look at the system. Because what really struck me about this source is that it doesn't read like a creative writing class. It reads like a technical manual for fulfilling ultra -specific demand. That is the perfect way
to frame it. Most people think AI novel means you just, you know, push a button and a book pops out. Right. But this guy argues that if you do that and just get garbage, the workflow we're going to look at is a seven step process. It involves market research with like data science tools, architectural plotting so the AI doesn't get confused and a very specific distribution strategy. So let's make this out from it. We're
going to look at the economics first. Why? These tiny little micro niches are printing money while the big publishers just ignore them. Yeah. Then we'll get into what they call the surgical workflow. Using perplexity for research, Claude for drafting, and a tool with the truly hilarious name for the covers. Nano Banana Pro. I still can't quite believe it's a real tool, but we'll get there. And then finally, we'll talk about the context cage and how to stop the AI from hallucinating
and, of course, the Amazon SEO game. Sounds like a plan. Okay, so let's start with the money. The source makes this distinction that I think really explains everything. It contrasts the traditional publishing model, you know, your random house, your penguin, with this new AI indie model. And it all just boils down to the size of the audience you need to survive. It's all about overhead. I mean, if you're a traditional publisher, you've got offices, editors, marketing
teams, legal departments. Right. So for a book to even be worth your time, it needs to sell tens of thousands of copies. You need mass appeal. You need the next Harry Potter. But mass appeal is incredibly expensive to market because you're competing with Netflix, Fortnite, everything. And this model just flips that on its head. Completely. This model relies on what the source calls niche hunger. You don't need millions of readers. You need a very small group of absolutely obsessed
readers. Obsessed. Yeah. The source lists these examples that, I mean, they sound like jokes to the average person. Mafia romance with neurodiverse protagonists or cozy mystery specifically for 45 to 60 year olds. I genuinely laughed when I read. Age gap werewolf romance. But the argument is that these readers are like starved for content. They are. I mean, if you love that specific subgenre, you can't just walk into a Barnes and Noble and find a whole shelf for it. You might find one
book a year if you're lucky. So when you go on Amazon and you find an author who produces exactly that flavor, you don't just buy one book. You buy their entire back catalog. You subscribe to their newsletter. The loyalty is just off the charts compared to general fiction. It's the long tail theory. Kind of turbocharged. And the math in the document is pretty compelling. They break down the unit economics. You price a book at $2 .99. Which is the impulse buy price.
It's basically the price of a coffee. No one thinks twice about it. Exactly. So in $2 .99, you only need to sell about 50 copies a day to generate, what, $5 ,000 to $7 ,000 a month? And that's just one book. The strategy here isn't to write one masterpiece. The source explicitly talks about the four book series strategy. If you launch four books in a series, they all act as a funnel for each other. You can hit that 50 sales a day mark across the whole ecosystem.
The document claims this has an annual potential of over $240 ,000. That is a top 1 % income for being essentially a Pulp Fiction writer. And
just look at the cost. book can cost you anywhere from a thousand to five thousand dollars just to get it ready for print right editing cover design all that all of it here your cost is the subscription fee for claude and maybe mid -journey 20 maybe 40 bucks a month the barrier to entry has just completely collapsed so it's not about writing the next great gatsby at all it's about feeding a very specific hunger exactly you're moving from mass appeal to obsessed reliability
okay so if the barrier is zero Why isn't everyone doing this? Or I guess, why is most AI fiction so bad? Because most people skip the engineering part. They just open up ChatGPT and say, write me a book about a werewolf. And the AI writes something generic and boring and just full of plot holes. This workflow, it's all about constraints. Let's get into that workflow then. So step one is research. And this isn't just, you know, brainstorming
ideas. They're using data. to find what the source calls content gaps yeah they're acting like data scientists the guide recommends using perplexity specifically in its pro search mode you are not looking for what's popular okay if you search popular romance books you'll just get crushed by the competition you're looking for complaints complaints you mean like bad reviews Sort of. The prompt strategy is to search Reddit threads,
Goodreads reviews, forums, anywhere. Readers are saying things like, I'm so tired of reading X or why are there no books about Y? You're looking for frustration. Frustration equals demand. That is brilliant. You're literally finding the market failure. Precisely. And once you find a potential niche, let's just stick with our monster romance example, you have to validate it. You go to Amazon, find the top books in that weird little category and check their bestseller rank. Or BSR. What
does that tell you? If those books are selling 50 plus copies a day, you've got a green light. If they're selling zero, the niche is dead. Move on. OK, so we found a hungry crowd. Now we have to actually plot the story. And this is where the guide brings in a concept they call reverse thinking. This was one of the biggest takeaways for me. It's so key. The rule is you must plot the ending first before you write a single word of chapter one. Why is the ending so critical
for an AI? because of something called prompt drift it's the technical term for what happens when an ai writes a long story it starts out strong but by say chapter four it forgets the main character's motivation by chapter 10 it completely forgets a subplot you started in chapter two the story just It wanders. It creates those infamous hallucinations and plot holes. Massive
ones. So by deciding the ending for, I mean, literally asking the AI to generate 10 satisfying endings and then you just pick one, you create a destination, you turn that ending into a PDF. The source calls it ending reference dot PDF. Now, every single time you ask the AI to write a chapter, you are referencing that PDF. It acts as an anchor for the whole project. It's like a GPS. If you don't put in a destination, the
GPS just drives around aimlessly. But if you lock in the destination, every turn is calculated to get you there. That's a great analogy. The AI is the car, but that PDF is the satellite lock. So why is that PDF step so critical for the AI? It acts as an anchor so the AI never, ever forgets the destination. So we have the ending. But we still need the middle, obviously. The source calls this next phase the Bible. Yeah, this is where you build your constraints. You
need two more PDFs. The first one is the character profiles. Okay. And the source notes something really specific here. Ask for detailed physical descriptions. If you don't lock down that the hero has a scar on his left cheek, the AI will forget it or move it to the right cheek or just remove it entirely three chapters later. Consistency is really the enemy of large language models. It really is. So you lock the characters in a PDF, then you build the structure, and the source
swears by the save the cat method. Which is that famous Hollywood screenwriting formula, right? Exactly. Blake Snyder's beat sheet, it's got 15 beats. Opening image, theme stated, catalyst. Dark Night of the Soul, Finale, all of it. Why save the cat specifically? I mean, why not just a standard three act structure? Because AI naturally wants to resolve conflict. Have you ever noticed that if you ask an AI to write a story, everyone tends to get along a little too quickly? It tries
to be helpful and nice. Yes. Oh my god, yes. It hates tension. It just wants everyone to go have some tea and resolve their differences calmly. Save the Cat forces tension. It forces a dark night of the soul where the hero has to lose everything. By forcing the AI to follow this beat sheet, you stop the story from becoming this boring sequence of nice events. So you upload your ending PDF and your character PDF to Claude, and you just say, create a chapter -by -chapter
outline using the Save the Cat structure. That's the prompt. I love the analogy the source uses here. It compares this to... To building a blueprint before you lay a single brick. Most people just want to start writing. But this workflow spends a huge amount of time just building the constraints. That's the key word right there. Constraints. AI is creative, sure, but it's chaotic. You have to fence it in. So we aren't writing yet. We are just building constraints. Right. We're building
a context cage so the AI stays on track. I love that term, context cage. It really implies the
AI is this wild. animal that needs to be penned in so okay the cage is built now we actually have to generate the pros now the fun starts the source recommends using claude 4 .5 opus in project mode let's walk through the actual drafting phase because this is where the industrial part really kicks in yeah this is the assembly line you open a fresh chat you upload your three holy grail pdfs and you start with chapter one But, and this is really crucial nuance, you don't
just say write chapter two. You have to reprime the pump every single time. How so? Your prompt has to be something like write chapter two, aligning closely with the save the cat structure and the character profiles in the attached files. You have to constantly remind it of the constraints. If you stop reminding it, it starts to hallucinate. And even with all of that, the source mentions a very specific technical hurdle. The context
window limit. This is the bottleneck. Even the best AI models, like Claude Opus, have a limit on how much text they can hold in their active memory. For a novel, you usually hit that wall around chapter 15. And the AI starts to get dementia, is how they put it. It forgets what happened in chapter one. Right. I think we've all seen that in long chat threads. The bot starts repeating itself or contradicting itself. The source offers a workaround for this that feels very... Manual.
It is manual, but it's effective. It's a hard refresh. Once you hit that limit, say chapter 15, you take everything you've written so far, compile it into a new PDF called chapters 1 to 15, and you start a brand new chat session. Oh, wow. And you upload that new PDF as its history. You're manually giving it a long -term memory. You're giving it a summary of the past so it can continue the future. It's like you're clearing
its cache. Yeah. And the speed difference. I mean, the comparison between the free plan and the pro plan was startling. The source says on the pro plan, doing all this, you can finish a full novel draft in a single afternoon. A single afternoon. That is, I mean, honestly, it's hard to wrap my head around that kind of volume. It's industrial scale. It is. But I have to be vulnerable here for a second. I still wrestle with prompt drift myself, even with all these PDFs and context
cages. Sometimes the AI just... It goes off the rails. It takes a lot of active management. You aren't just watching Netflix while it writes. You are reading every output, checking for drift, regenerating scenes. It's not writing, but it is intense editing. It sounds like memory management is the real skill here. Yes. You're managing the AI's short -term memory to maintain a long
-term narrative. We're going to take a quick break, but when we come back, we need to talk about the final polish, how to make sure this doesn't sound like a robot wrote it, and the tool called Nano Banana Pro that solves the biggest problem with AI art. Stick around. Welcome back. We're deep diving into the workflow of high -volume AI authors. We've researched a niche using data. We've plotted the ending first to avoid drift. And we've generated a draft using a context cage
of PDFs. Right. But now we have a raw manuscript. And let's be honest, raw AI text is usually pretty dry. It's competent, but it's soulless. It tends to overuse certain words like shiver, tapestry, and delve. It really, really loves the word delve. So how do we fix that? The source outlines a two -phase editing process. Yeah. And phase one is actually using AI against itself. Right. You don't start by reading it yourself. You upload the whole draft back into Cloud and you ask it
to look for logic errors. You say, identify continuity errors. Did I say the door was locked and then she walked through it? Did I change the car from a Ford to a s***? Chevy. And AI is good at that. Incredibly good. It spots those logical breaks because it just treats them as data points. But phase two. Phase two has to be human. Phase two is the human Polish. This is where you actually earn your money. You have to look for tone, voice, and emotional rhythm. And the source suggests
a simple, brutal test. Read the first chapter and the last chapter out loud. Out loud. Yep. Your ear catches things, your eye misses. If you trip over the words or if the dialogue sounds like a robot trying to act human, the reader's going to hate it. You have to make sure the emotional beats are actually landing. If the dark night of the soul doesn't make you feel something, you have to rewrite it manually. Okay, let's move on to the cover. We've all seen AI art.
It can be incredible, but it has a notorious weakness. Hands are getting better. But text. Oh, text usually looks like alien hieroglyphics. It's a mess. It's like a soup of letters. A disaster. But this guide recommends a tool I had never even heard of. Nano Banana Pro. I know, right? It sounds like a mobile game for toddlers. Nano Banana Pro. It's actually a specific tool you can access through the Gemini ecosystem. And the source recommends it over mid -journey or
daily. Purely for text rendering. It can actually spell the title correctly. It can spell. If you're selling a book called The Werewolf Secret, you cannot have the cover say The Werewolf Scrit. Midjourney really struggles with that. Nano Banana Pro handles typography inside the image generation much better. And it handles the aspect ratio, the book cover dimensions? Yes. The guide is
specific. 2560 by 1600. pixels the claim is that you can generate a professional looking cover with legible title text in about 15 minutes and critically it's free that really is a moment of wonder for me i remember when getting a cover design cost 500 and took three weeks of back and forth emails with a designer and now it's 15 minutes and zero dollars it's just wild it's the democratization of the entire supply chain it unlocks the ability for anyone to look professional
So why use Gemini or Nano Banana specifically for the cover? It's the only one that doesn't mess up the text on the book title. OK, so we have the book. We have the cover. Now we have to sell it. And the source says this is where most authors fail. Great book plus bad SEO equals no sales. This is where we have to change how we think about Amazon. We tend to think of it as a bookstore. You know, we imagine brising shelves. But the source is very clear. Amazon
is a search engine. It's Google. But for products, if you aren't optimizing for the algorithm, you are invisible. So how did these authors optimize? They use AI again. The workflow uses ChatGPT specifically in its thinking mode to analyze the market. You feed it your novel summary and ask it to analyze successful book descriptions in your specific niche. And then ask it to generate titles. SEO optimized titles and subtitles. You're not trying to be clever with your title. Right.
You know, The Moon's Glow is a poetic title. Alpha Wolf's Forbidden Mate and Age Gap Romance is an SEO title. You want the keywords that people are actually typing into the search bar to be right there in your title and subtitle. And then there's the pricing strategy. The source is very, very specific about the sweet spot. The golden zone. It's $2 .99 to $9 .99. This is an Amazon rule. If you price your book within that range, Amazon gives you a 70 % royalty rate. If you
go below $2 .99, your royalty drops to 35%. And if you go above $9 .99, it drops to 35%. Wow, that's a massive cliff. It really forces your hand. It forces the entire market into that pocket. So almost every indie author sits right at $2 .99 to maximize their volume while keeping that 70 % cut. It's fascinating how the platform dictates the economics, which then dictates the content. Completely. You format it with Kindle Create, which is a free tool from Amazon. You upload
it to KDP. And within one to three days, you're live on the biggest bookstore on earth. So Amazon is treated more like a search engine than a bookstore. Precisely. You are optimizing for keywords, not for browsing. It's incredible to see it all laid out like this. When you zoom out and you look at the research, the context cages, the nano banana covers, it's... not really about writing in the romantic sense of the word, is it? No, not at all. If I had to summarize this whole
deep dive, I'd say this. It is not about art. It's about pipeline management. It's a seven -step industrial workflow. You have research with perplexity to find the demand. You have structure with Save the Cat to tame the chaos. You have context management with those PDF cages in Claude. And you have distribution via Amazon SEO. It's just systems engineering applied to imagination. Systems engineering applied to imagination.
That's a really powerful way to put it. The source wraps up with a bottom line that I think is important. It says, this isn't magic. It works for people who think in systems and value volume over perfection. It's for the person who's willing to put in the reps. That's the truth of it. The barrier to entry is low. Anyone can buy a subscription. But the barrier to success is consistency. You have to treat it like a job, not a hobby. As we wrap up, I just want to leave you with a thought.
We talked about niche hunger today. Those readers who are just desperate for something very specific that they can't find. My question to you is, what is the micro niche that you know better than anyone else? What's the weird specific topic that you and your friends complain there isn't enough content about? because chances are there are a few thousand other people thinking the same thing, just waiting for someone to build
the system to feed them. That's the million -dollar question, or at least the $240 ,000 question. Thanks for listening to The Deep Dive. We'll see you next time.