Half of you should stop reading at the bottom of this page.
Most courses put the economics at the end, softly, after you've already bought. It's first here because I'd rather you quit now than refund later.
Every number below comes from my own KDP royalty report, pulled 15 July 2026, and my own expense ledger through the same date. Complete months run February 2025 through June 2026, seventeen of them. July 2026 is still in progress and is excluded from every claim I make about records or averages.
I'm not going to show you the catalogue. I'll explain why in a minute, and that reason is itself the first lesson.
One pen name. 549 titles that have earned anything. Seventeen months. $19,701.
Not per month. Total. Everything that catalogue has ever earned, from its first sale to the day I pulled the report.
| Copies sold (paid) | $11,954 |
| Kindle Unlimited pages read | $7,747 |
| Lifetime total | $19,701 |
| KU share of revenue | 39.3% |
| Pages read, lifetime | 1,959,835 |
| Paid units sold, lifetime | 4,679 |
| Revenue per paid unit | $2.55 |
Divide it out and you get $36 per title, lifetime. That's the average, and the average is a lie. The next section is why.
Complete months only. July 2026 in progress and excluded
The best month in the catalogue's history is June 2026, at $2,531, and it's the most recent complete month, so this is the operation at its best rather than a faded peak. For the twelve months ending June 2026 the catalogue earned $16,349. The last three complete months (April, May, June) are the three best months it has ever had, in order: $1,768, then $2,127, then $2,531.
That's the honest good news, and I want it on the record before the bad news arrives: the thing works, it is growing, and it is still small. If you came here because someone claimed $10k/month from AI books, close the tab. I have 549 earning titles and a fully automated pipeline, and my best month ever is a quarter of that.
Here's the distribution, and this is the part that matters:
| Best-earning title, lifetime | $1,229.83 (6.2% of everything) |
| Second-best | $492.90 |
| Median title, lifetime | $17.90 |
| Mean title, lifetime | $35.89 |
| Titles that earned under $10, ever | 161 (29%) |
| Titles that earned under $25, ever | 347 (63%) |
| Titles that earned under $100, ever | 505 (92%) |
| Top 10 titles' share of revenue | 20.1% |
| Top 50 titles' share of revenue | 44.3% |
Two thirds of my catalogue has earned less than $25 in its entire existence. Not per month, ever. Ninety-two percent has never cracked $100.
Cumulative share of revenue as you add titles, best first
Read that again, because it kills two ideas at once.
The first is the lottery framing: write enough books and one will hit. Sort of true, but look at the size of the hit. My number-one title, the best of 549, the one the whole catalogue is supposedly buying lottery tickets for, has earned $1,229 lifetime. It's 2.5 times the second-best, not 10 times, not 100 times. There is no jackpot in this data. There's a slightly-less-small number at the top of a great many very small numbers.
The second is the opposite error: the winners carry it, so the long tail is waste. Also wrong. The top 50 titles are only 44% of revenue. The other 499 books are the majority of the business. You can't prune your way to the good ones, and you can't pick them in advance. If I could, I'd write fifty books instead of five hundred.
This is a volume business with no jackpot. Both halves of that sentence are load-bearing.
The pitch you've heard is that a catalogue snowballs. Books cross-promote, the algorithm learns you, readers who find one book buy nine more, and revenue curves upward while your book count grows linearly. It's a good story. Here's seventeen months of my data:
The catalogue grew 4× over this window. This line did not move.
Each book earns about $3.25 a month and does not care how many siblings it has. Fifteen of those months had a catalogue big enough to mean anything. Across them, revenue per title bounced between $1.29 and $4.69 and went nowhere. It did not climb as the catalogue grew. It did not sink either. Flat.
I want to be precise here, because I got this wrong myself. An earlier draft of this module claimed per-title revenue was declining. What I had actually done was pick two months, one high and one low, and draw a line between them. When I pulled a fresh report and checked all fifteen months properly, the decline evaporated. It's flat, and June 2026 posted $4.52 per title, the second-best figure in the dataset. The catalogue is not decaying. It also isn't compounding. It's just adding.
Fit a straight line through all fifteen months and that line rises by $0.03 per title per month. Take it at face value and it predicts a drift of about 45 cents per title across the whole fifteen months. Now compare that to the spread: the months themselves range over $3.40, from $1.29 at the bottom to $4.69 at the top. The supposed trend is about a tenth of the noise it is swimming in. That is not a trend. It is a flat line with weather on it.
It is also exactly why my earlier draft was wrong. Two months compared against each other can never tell you about a trend, because either one of them can be weather. Give me this same dataset and let me choose the pair, and I can show you a business tripling or a business collapsing. Both pictures would use real numbers. Both would be dishonest.
The honest limit of the claim: fifteen months is not enough to rule out a small real trend in either direction. What it does rule out is compounding, because compounding is not a subtle effect. If a catalogue snowballed, growing it to 549 titles would have bent this line visibly upward. Nothing bent.
That's the finding, and everything else in this course follows from it:
Revenue ≈ number of titles × ~$3.25/month. Adding books adds revenue. It does not multiply it.
If revenue is linear in book count, the entire game is cost per book and not losing the catalogue, because nothing you do to any one book matters much, and anything that threatens all of them matters enormously. That's why seven of the nine modules here are about pipelines and guardrails rather than about writing.
Now look at December 2025 on that chart: 378 titles live, revenue $488, a third of what the same catalogue earned four months earlier. Nothing broke. I did nothing wrong. That's the month-to-month variance of this business, and it's roughly when most people quit, having concluded the machine is broken. The machine wasn't broken. February came in at $2,059.
| Free promotional giveaways | 16,855 units (78%) |
| Paid units | 4,679 |
| KU pages read | 2,076,100 |
Seventy-eight percent of every copy I've ever moved, I gave away. Free runs drive ranking and push readers into Kindle Unlimited, where pages read become 39% of revenue. That's not a failure mode, it's the mechanism. But it means "units moved" is a vanity number, and anyone quoting one at you is selling something. My real sales figure is 4,679 copies in seventeen months, at $2.55 each.
From my expense ledger, 2026 year-to-date (1 January to 15 July 2026), everything attributable to this operation:
| Audiobook narration (human narrators) | $2,388 |
| AI: writing model API | $547 |
| AI: general assistant subscriptions | $570 |
| AI: image/cover generation | $471 |
| AI: legacy writing tool | $19 |
| Cloud hosting (3 author sites + newsletter) | $102 |
| Total | $4,097 |
Two things jump out.
Fifty-eight percent of my spend is audiobooks, a separate bet with separate economics that I'd tell a beginner to skip entirely for the first year. Strip those out and the entire machine (writing, editing, covers, formatting, hosting three author websites) costs $1,709 for six and a half months.
The marginal cost of one book is about $10. At roughly 40 books per month, direct API spend to write, edit, and cover them runs $400 to $450 a month.
Put the halves together:
A median book costs ~$10 to produce and earns $17.90 over its entire life. The median book in my catalogue nets about eight dollars. Lifetime.
That's the business. Not a money printer. It's an eight-dollar margin repeated five hundred times, which is roughly what the catalogue makes in a good month.
Note what is not in that ledger: my time. Charge yourself even minimum wage for the hours that built this and the operation is deep in the red and always has been. This is not passive income. It's a job with an automation layer, and the automation is the only reason the margin survives contact with reality. A human ghostwriter at $0.03 a word would cost $3,000 for one 100k-word book that will earn $17.90.
At an eight-dollar margin per book, nothing you do to any single book can hurt you much. What can hurt you is anything that hits all of them at once:
This is the asymmetry that defines the operation: the upside per book is eight dollars, and the downside is the entire catalogue. Every rule in this course exists to protect the catalogue, not to optimize a book. Anyone teaching you growth hacks for a single title has the math backwards.
You'll notice I haven't named the pen name, the titles, or the genre.
That's not modesty, and it isn't because I'm hiding a bad result. You just read the results. It's because the pen name is an asset and exposure is a cost. These personas are built to read as real, separate authors. A course that doxes them converts a working business into a marketing prop. I'm not doing that, and if you build this properly, you won't either.
The niche is the same story. I won't hand you my categories, for two reasons. Telling ten thousand readers exactly where I sell is how a workable niche stops working. And more usefully to you, my niche is a 2025 answer to a 2025 market. It'd be stale by the time you read it. So instead of the list, the paid handoff includes the niche research generator: the method for finding a live niche in your market, now. The method keeps working after the list goes stale.
That's the trade. Less of my specifics, more of what produced them.
You're going to ask, so I'll answer before you do.
The objection is that machine-written books flood a market human authors are trying to survive in. It's the first thing people say and it deserves a straight answer rather than a dodge.
Mine is: nobody in this transaction is deceived.
That's the whole claim. Not that it's noble, but that it's honest at the point of sale, and the person spending the money knows what they're spending it on and comes back.
Now the strongest version of the objection, because a weak version isn't worth answering: even if no individual reader is deceived, you're competing for the same shelf as people writing one book a year, and you're doing it forty times a month.
True. I won't pretend otherwise, and I'd rather say it plainly than argue it away.
What I'd say back is that genre erotica was a high-volume commodity market long before I arrived. It was written fast, priced low, read once, largely by writers using pen names and formulas for exactly the reasons I do. AI changed the cost of production. It didn't invent the category or its economics. And the numbers in this module are what that competition actually looks like from the inside: a median book earning $17.90 and netting eight dollars. Nobody is getting rich crowding anybody out. That's less a defense of the practice than a deflation of it.
If you find that insufficient, that's a legitimate place to land, and this is a good moment to stop. No hard feelings and no refund necessary, because you haven't paid me anything yet. I'd rather lose you here than have you build something you'll feel bad about.
Stop reading if:
Keep going if:
The rest of this course is the machine: the personas, the lanes, the pipeline, the covers, the KDP mechanics, the guardrails, the sites, the survival rules. Every one is written the same way. Here's what broke, here's what it cost, here's the rule that came out of it.
I start with the failures because the failures are the only part you can't get from a YouTube video.
Method, so you can check my work rather than trust me:
I used approximate conversion rates, not the exact rate Amazon applied on the day each royalty was booked. Rates move daily, each marketplace books in its own currency, and reconstructing seventeen months of that precisely would change nothing and take a week. So I did not, and I would rather tell you that than imply a precision I do not have.
The reason the error stays small: those three named markets are 94% of sales royalty. The other five currencies split the remaining 6% between them, so even a bad conversion on one of them moves the lifetime total by a rounding error. The 3% is dominated by the dollar figure, which needs no conversion at all.
So if you are hunting for the number in this module most likely to be wrong, it is not the total, and it is not any of the per-title figures. It would be a figure quoted for a single small marketplace, and I have not quoted one.
KDP gives you two royalty figures per title: the total royalty, and the royalty from paid sales. It does not break the Kindle Unlimited money out as its own line. So subtract the second from the first. What is left is the KU money. Divide that by pages read and you have the rate.
That gives $0.00395 per page, blended, and blended is the load-bearing word. It is one number standing in for eight marketplaces that pay different rates, weighted by where my pages actually came from. 67% of my pages are US, so the blend leans US, and the other 33% drag it below the US rate. That is the entire reason my figure looks low next to the numbers people quote at each other in forums. They are quoting the US rate in a good month. I am reporting an average over everything, which is the only figure I can actually defend.
Do not port my rate into your own forecast. Yours depends on which marketplaces your readers sit in, and on a rate Amazon resets every month.