Chapter 3 of The Recursive Grift mentions the book was reworked from a single AI prompt. This page has the original draft. There are 100% fewer footnotes here than in the final version.
The Recursive Grift: A Short, Honest Guide to Writing AI Books About Writing AI Books
(An AI helped write this. You’re about to learn why that sentence is both the whole business model and the whole joke.)
A Note on What This Is
You are holding — or scrolling — a book about how to make a small amount of money writing books with AI, written by an AI, for a human who wants to write books about writing books with AI, to sell to other humans who want to write books with AI.
Nobody in this chain is lying to anybody. That’s the interesting part. Every layer is disclosed. And it’s still a little absurd. Let’s sit with that for a page before we get practical, because understanding the absurdity is what will keep you from writing a bad, dishonest version of this book — the kind with a fake income screenshot on page one and an URGENT!!! BONUS CHAPTER at the end that’s just an affiliate link.
This book will not have a fake income screenshot. It will tell you, plainly, that this is a modest hustle, not a moonshot. If you came here for a Lamborghini, please close this file and go do literally anything else with your time — ideally something with better unit economics, like flipping thrift store furniture.
If you came here because the idea amused you and you’d like to make a few honest dollars while learning something real about self-publishing, keep reading.
Chapter 1: Why You’re Actually Doing This
Let’s rule out the wrong reasons first, because the wrong reasons produce bad books, and bad books produce refunds, and refunds produce Amazon flagging your account, and flagged accounts produce nothing at all.
Wrong reason: “I heard you can make $10k/month passive income.”
You heard that from someone selling a course about making $10k/month, which is the actual $10k/month. This is the ouroboros you’re about to join. At least know you’re joining it.
Wrong reason: “AI does all the work, I just click publish.”
This is true, and it is exactly why it doesn’t pay well. Anything a human can produce by clicking a button, ten thousand other humans are also producing by clicking the same button, this week, in your exact category. You are not making a product. You are making noise in a category that is already mostly noise.
Right reason: “I want to understand self-publishing mechanics — covers, categories, keywords, pricing, ASINs, the whole KDP machine — and this is a low-stakes way to learn it by doing, with a topic I find genuinely funny.”
Right reason: “I have actual expertise or a genuine point of view about AI writing tools, and I want to package it well, disclose my process honestly, and sell it to people who’d rather pay a few dollars for organized, checked writing than spend four hours reading forum posts.”
Right reason: “I want a modest, honest side project, not a business.”
If your reasons are in the second group, you’re in the right place. If they’re in the first, no judgment — but recalibrate your expectations before you spend a weekend on this, or you’ll be disappointed in a way that makes you write worse, more desperate marketing copy, which readers can smell from across the store.
Chapter 2: The Niche Is the Only Decision That Matters
Everything downstream of your niche choice — cover style, price point, keyword strategy, whether anyone buys the thing — is basically fixed once you pick it. So spend real time here, not on your prompt engineering.
Three viable lanes for a book like this:
1. The straight instructional guide. “How to Use AI to Draft Your First Novel.” Competes directly with a hundred similar titles. Wins, if it wins, on being more specific and more honest than the competition, which is a low bar — most entries in this category are vague, hedge everything, and were clearly written in an afternoon by someone who has published exactly one AI book (this one).
2. The narrow expert angle. “AI-Assisted Worldbuilding for Cozy Fantasy Writers” beats “How to Write Books with AI” because it’s a smaller pond with less competition and a reader who’s easier to identify and market to. Specificity is underrated as a strategy across all of self-publishing, not just this niche.
3. The meta/satirical angle — what you’re reading now. Smaller total audience, but near-zero competition, because most people playing this game don’t want to admit what game they’re playing. The honesty itself becomes the differentiator. This lane has a ceiling — it will never be your day job — but it’s the most fun to write and the least likely to embarrass you later.
Pick one lane. Do not try to write “the definitive guide” that covers everyone — definitive guides are what people write when they don’t know who they’re talking to, and readers can tell.
Chapter 3: The Actual Pipeline
Here is the boring, load-bearing part of the book. This is the part worth the price of admission, because it’s the part most guides in this category gesture at vaguely instead of actually describing.
Step 1: Outline it yourself, by hand, badly, first. Not with AI. You, a blank document, and a bulleted list of what each chapter needs to accomplish and in what order. This takes twenty minutes and saves you from the single most common failure mode of AI-assisted books: structure that reads like a Wikipedia article, because that’s the shape AI defaults to without a strong human skeleton to fill in.
Step 2: Draft chapter by chapter, not book by book. Never ask for the whole manuscript in one prompt. You’ll get bland, repetitive, structurally identical chapters that all sound like they’re circling the same three points. Feed the model your outline for one chapter at a time, with specific instructions about what makes this chapter different from the last one.
Step 3: Read every sentence like you mean it. This sounds like advice from your seventh-grade English teacher because it is, and it’s still correct. AI drafts contain confident, fluent, wrong things — invented statistics, plausible-sounding claims with no basis, advice that’s outdated or just made up. You are the fact-checker. This step is non-negotiable and it’s the step people skip, which is why so many AI books have that particular flavor of hollow confidence you can spot from the first paragraph.
Step 4: Cut the AI’s favorite words. Every model has verbal tics — right now, “delve,” “tapestry,” “in today’s fast-paced world,” and sentences that arrive in suspicious threes. Find and remove yours. A five-minute pass here does more for the book’s voice than any amount of prompt tweaking.
Step 5: Rewrite the introduction and conclusion entirely yourself. These are what get read in the “look inside” preview and what get quoted if anyone reviews the book. They should sound like a person, because they’re the part a person is most likely to actually read closely.
Step 6: Disclose your process, specifically, not vaguely. Not a buried line on the copyright page. A short, plain paragraph — up front or in an afterword — that says what you did: which tool you used to draft, what you edited and fact-checked, what that means for the reader. This is the ethical minimum and also, not coincidentally, good marketing, because “here’s exactly how this was made” reads as more trustworthy than silence.
Chapter 4: Covers, Blurbs, and the Machinery of Being Found
Nobody buys a book. They buy a thumbnail, a title, and the first two lines of a blurb, in that order, in about four seconds. This is true whether AI wrote a word of it or not, and it’s the part most first-time self-publishers underinvest in relative to how much time they spent on the actual manuscript.
Cover: Spend real money here, or real time, or both. A template that looks like every other book in your category is invisible. A template that looks slightly better than every other book in your category is the whole game — you don’t need to look expensive, you need to be the one someone’s eye catches while scrolling.
Title: Front-load the thing people search for, and put your actual angle in the subtitle. A plain, specific title does more work than a clever one, because the clever title doesn’t show up in search.
Blurb: Lead with the specific problem you solve, not a list of features. Naming what makes you different from the noise around you is a better opening line than a generic promise of secrets or success, because it signals — correctly — that the book inside isn’t the same as the others.
Categories and keywords: This is unglamorous and matters more than people think. Research what’s actually working in your specific sub-category before you pick — not what you assume is working.
Chapter 5: The Ethics, Stated Plainly
Disclosed and original clears the floor of “not deceptive.” It doesn’t automatically clear the bar of “good.” What determines that:
- Is there a real human point of view in here, or is it filler? A book that’s mostly AI output with a light human polish, sold as expertise, is a thinner product than its packaging suggests — even with a disclosure line.
- Does it deliver honestly on what it promises? No fake screenshots. No implausible income claims unless they’re true and documented. The income-hype genre this book lives adjacent to has a real cost: it takes money from people who can least afford to lose it, chasing a promise that was never realistic.
- Are you adding a signal, or contributing to the noise? Self-publishing categories are genuinely harder to navigate than they used to be, partly because of low-effort AI volume plays. A well-made, honestly disclosed, actually useful book is a defensible thing to add to that pile. A templated rehash isn’t, even if nobody catches you.
None of this requires you to be a saint. It requires you to not lie, not overpromise, and put in the editorial work that makes the thing worth the reader’s few dollars and forty minutes.
Chapter 6: What You Will Actually Make
Here is the number most books in this category won’t give you, because the real number doesn’t sell courses.
A well-executed short ebook in a reasonably specific niche, priced modestly, with a decent cover and honest positioning, will typically bring in a low, one-time-feeling sum over its lifetime. Some do better if they catch a wave of interest or get picked up by an algorithm at the right moment. Most don’t. This is true whether or not AI was involved in writing it — it’s just the base rate for a single, unmarketed, self-published nonfiction ebook in a competitive niche.
The meta angle you’re reading right now has a lower ceiling than a straightforward instructional guide, because the total audience for “a book that’s honest about the genre it’s in” is smaller than the audience for “a book that promises to make you money.” That’s a fair trade if you’d rather sleep well than sell aggressively.
If you want the number to be bigger, the lever isn’t the book — it’s what you attach to the book: a template pack, a short paid course, a service you sell to other people using the book as a portfolio piece. The book itself is closer to a business card than a business.
Chapter 7: The Upsell, Because You Asked
If you want to make more than a modest, honest sum, here’s where the actual money in this space lives, roughly in order of how much trust it deserves:
- A companion template pack — outline structures, prompt sequences, a keyword-research checklist — sold alongside the book for a bit more. Low effort to make once the book exists, and it converts because the buyer already trusts you enough to have bought the first thing.
- A short paid course, if you actually enjoy teaching and have more to say than fits in a book. This only works if it’s genuinely more valuable than the book — video, feedback, structure — not the same content re-packaged at a higher price, which readers notice and resent.
- Freelance or service work, using the book as a credibility artifact when you pitch small businesses on AI-assisted content help. This is the most durable income of the three, and the least passive, which is exactly why it’s the most trustworthy one.
What I won’t recommend, because it’s the part of this genre that actually deserves its bad reputation: stacking affiliate links to AI tools and publishing services throughout the book itself, dressed up as “resources.” If you do this, disclose it clearly and keep it to an appendix. Don’t let it bleed into the advice chapters, where a reader can’t tell if you’re recommending a tool because it’s good or because you get paid when they click it.
Chapter 8: When to Stop
You will be tempted, if this book does even modestly well, to spin up several more just like it — different niches, same skeleton, minimal new thought. This is the volume-play strategy from Chapter 1, the one we ruled out. It’s not that it can’t work; it’s that it stops being the thing you said you wanted, back when your reason was to understand this and make something honest. Somewhere a few books into a templated series, you’re not writing anymore. You’re running a very small, very low-margin content operation, and you’ll produce worse work while telling yourself it’s a business.
If you find yourself there, that’s fine — some people want that. Just notice the moment you cross from “project” into “grind,” and decide on purpose, rather than sliding into it because the first book did okay and you assumed the next one would do better.
It usually won’t. Niches saturate. Novelty fades. The honest math rarely scales the way the course sellers imply.
Afterword: The Part Where I Admit the Joke
You now own, or are reading a sample of, a book about writing AI books that was itself written by an AI, at the direction of a person who wanted, and I quote, “modest sales.” I think we’ve delivered on the modesty, at minimum.
If there’s a real lesson buried under the recursion, it’s this: the honesty is the product. Not the AI. Not the niche. Not the cover template. Every layer of this small industry runs on withheld information — hidden income numbers, buried disclosures, unstated affiliate relationships, promises calibrated to what sells rather than what’s true. The most differentiated thing you can do in a crowded, low-trust category is simply not do that.
It won’t make you rich. It might make you a few honest dollars, and a book you’re not embarrassed to have your name on.
That’s a fine afternoon’s work. Go publish it.
Disclosure: This manuscript was drafted by Claude (Anthropic) at the direction of a human author, based on direction and an outline the author provided in conversation. The author is responsible for final review, fact-checking, and any claims made under their name before publication.