Self-publishing AI-written books on Amazon

The research was done by AI agents that open web pages. Some sites refuse them; where that happened we say so. The small numbers point to the source list at the end. Figures that are our own sums are marked “our estimate” or “our sum”. The first round of research on this page could not run web searches, only open pages whose addresses the agents already had. A second round on the same day ran 29 searches and opened the pages they found. Where it changed what an earlier draft said, the text says so. “What we did not verify” lists what is still left out.

The claim

These are examples of the claim, copied from our collection of 52 tweets about this way of making money¹. View counts are as collected on 6 October 2026. Dates are worked out from each tweet’s ID number. The page judges the method, not the people who tweeted.

“I’m convinced. Amazon KDP is the most beginner friendly business on Earth. I make $50,000/month doing it. • AI writes them. • AI designs them. • Amazon sells them.”

@TommiPedruzzi, 30 June 2026, 126,599 views¹

“Now don’t miss AI publishing in 2026. It’s boring… but if you start today, you could make $3,000 by the end of March.”

@Manu_Sisti, 13 February 2026, 578,546 views¹

“Create an Amazon KDP account, and sell AI eBooks. It’s boring… but if you start today, you can make $3,000 by the end of August.”

@TommiPedruzzi, 31 July 2026, 436,501 views¹

“I published 1,500 books on Amazon and made $3M”

@TommiPedruzzi, 10 October 2025, 197,494 views¹

These two accounts wrote 40 of the 52 tweets, which hold 94% of the views (5.25 million of 5.57 million; our count)¹. Both people are listed as co-founders of one company, Nespola, on its own site: “Tommi Pedruzzi Co-Founder Manu Sisti Co-Founder”⁴⁵. So most of what we collected comes from two co-founders of one company that sells a publishing programme (our reading; the tweets are on their personal accounts). Most of the tweets go on to offer a free guide or training. We quote only the part that states the income.

What the scheme is

Kindle Direct Publishing, or KDP, is Amazon’s service for publishing a book yourself. You upload a file, Amazon lists it in its shop, and you get a royalty (a share of the price) on each copy sold. Amazon describes KDP’s tools as “free and simple”⁴⁷. Paperbacks are printed one at a time when someone orders, and the printing cost comes out of the royalty, so there is no stock to buy⁵⁶.

In the version in the tweets, you do not write the book. An AI tool such as ChatGPT or Claude writes a short non-fiction book on a narrow subject, another tool makes the cover, and you upload it. One tweet describes a “10,000 words solution” packaged “in a 90-page eBook”¹. The promise is that many such books add up to a monthly income for about an hour’s work a day¹.

A book earns on Amazon in three ways:

• Ebook sales. A royalty on each Kindle copy sold².
• Paperback and hardcover sales. A royalty on each printed copy, after the printing cost⁵.
• Kindle Unlimited. A subscription where readers borrow books. Amazon shares a sum of money between publishers each month by the number of pages of their books that were read¹¹.

A related version is the “low-content book”: a journal, planner or logbook with little or no text inside¹⁷. Amazon allows these if they are flagged as low-content¹⁷. An earlier draft of this page said we had found no income evidence for them. The second round found three records checked by a business broker; they are in “The upside” below⁵¹. This page is still mainly about AI-written books with text.

The arithmetic

Income is copies sold, times the royalty on each copy, minus what you spend on adverts and on making the book. Amazon publishes the royalty side exactly. The tweet texts we collected give none of the other figures; we did not view their attached images, videos or threads.

Step 1: what one copy pays

Ebooks. Amazon pays 70% or 35% of the list price. The 70% rate comes with a delivery charge: “average delivery costs are $0.06 per unit sold, and vary by file size”². On Amazon.com the charge is $0.15 per megabyte of file size⁴. The 70% rate applies only inside a price band: “Effective July 7, 2026, the 70% royalty option price band is $2.99-$12.99 on Amazon.com, expanding from the previous range of $2.99-$9.99”³. Outside the band the rate is 35%³.

Our sums, taking the delivery charge off first: a $2.99 ebook pays 0.70 x (2.99 − 0.06) = about $2.05 a copy, and a $9.99 ebook pays 0.70 x (9.99 − 0.06) = about $6.95.

Paperbacks. Amazon’s formula is “(Royalty rate x list price) – printing costs = royalty”⁵. The rate on Amazon.com is 60% for a list price of $9.99 or more and 50% below that⁵. Printing a black-ink paperback costs a flat $2.30 for 24 to 110 pages, and $1.00 plus 1.2 cents a page for longer books⁶.

Our sums:

• 100 pages at $9.99: 0.60 x 9.99 − 2.30 = $3.69 a copy.
• 150 pages at $14.99: 0.60 x 14.99 − (1.00 + 150 x 0.012) = $6.19 a copy.
• 100 pages at $6.99: 0.50 x 6.99 − 2.30 = about $1.20 a copy.

Kindle Unlimited. Amazon says: “Author earnings are then determined by their share of total pages read.”¹¹ It does not publish a rate per page. Written Word Media, a book-marketing company, works one out each month. Its tracker gives $0.005170 a page for the United States in August 2026 and $0.004221 for July¹². That is about half a cent a page. A 90-page book read from start to finish would pay about 38 to 47 cents (our sum: 90 x 0.004221 and 90 x 0.005170). Amazon counts pages in its own way, which can differ from the printed count¹¹. If it counts a 10,000-word book as fewer than 90 pages, the payout is lower.

Step 2: the copies needed

Copies needed each month = income wanted ÷ royalty per copy. These are our sums from the per-copy figures in Step 1, before any advertising or production cost. “A day” is the monthly figure divided by 30.

So “$50,000/month” needs somewhere between about 7,000 and 24,000 ebook sales every month, depending on the price.

“$3,000 by the end of March” was posted on 13 February, 46 days before the end of March¹. Read as $3,000 in total by that date, it needs 432 to 1,463 ebook sales in six and a half weeks, about 9 to 32 a day from a standing start (our sum: 432 ÷ 46 and 1,463 ÷ 46). Read as a rate of $3,000 a month, it needs 14 to 49 a day. “$3,000 by the end of August” was posted on 31 July, so the total reading there is the same as the monthly one: 14 to 49 ebook sales a day in the first month¹. Amazon pays about 60 days after the month of sale¹⁴, so money from March sales would arrive in late May.

Step 3: how many books, and how fast

The tweets’ answer to the volume problem is many books. One says “I published 1,500 books on Amazon and made $3M”¹. Taken at face value that is $2,000 a book in total (our sum: 3,000,000 ÷ 1,500). An older tweet from the same account, from October 2023, says “I’ve published 100+ books on Amazon that generate $2,000/month each”¹. One is $2,000 a book in all, the other $2,000 a book every month. The tweets are two years apart and describe different numbers of books, so they need not conflict. We verified neither.

Amazon now limits how fast anyone can publish: “Creation limit: Up to 2 titles per book format each week. Submission limit: Up to 2 new titles per book format each week”⁷. The news podcast of the Alliance of Independent Authors, a trade body for self-publishers, says this limit took effect on 21 September 2026 and replaced a limit of 10 titles brought in during 2025⁶⁶. We could not confirm whether those 10 were per week or per day. A technology news site wrote in August 2026 that Amazon’s cap was “three per day”⁶³. For today’s rule we rely on Amazon’s own help page, which we read directly⁷.

This says nothing about whether the 1,500 figure is true. Those books were published under earlier, looser rules, and the count may include several formats of one book. It does mean a newcomer today cannot copy it quickly: two ebooks a week is at most about 104 a year. At that pace 1,500 ebooks would take about 14 years, or about 5 years if the ebook, paperback and hardcover of each book count as three titles (our sums: 1,500 ÷ 104 and 1,500 ÷ 312).

Step 4: what the main promoter’s own site says

Twenty of the 30 most-viewed tweets come from @TommiPedruzzi and six more from @Manu_Sisti (our count)¹. Their company, Nespola⁴⁵, has a blog post on how much KDP can pay, marked as updated in May 2026²². It gives figures that the tweet texts we collected do not.

The hours are close. The tweets we collected do not deny advert costs; their text does not give them. The timelines differ most: the site’s first six months are the tweets’ first month.

The site’s figures are royalties before costs. It defines “gross royalties” as “the total paid by Amazon before your expenses”²². On what is left, it says: “In months 1–4, aggressive advertising can mean net profit is near zero”, and from month 4 “net profit margins typically run 50–70% of gross royalties”²². It describes one student whose “first two months were losses — $400 and $600 respectively”²².

The post sets a target of $23 a day for each book, about $690 a month, before costs: “The $23/day target is not a per-book guarantee; it is what systematic execution in the right niche makes achievable.”²² At that target, $10,000 a month needs about 14 or 15 books all hitting it, and $50,000 a month needs about 72 (our sums: 10,000 ÷ 690 and 50,000 ÷ 690).

The company’s page for its larger programme works from a lower figure: “Six books generating $300 to $500 per month each is $21,600 to $36,000 per year.”⁷¹ The page does not say whether that is before advert costs⁷¹. At $300 to $500 a book, $3,000 a month needs 6 to 10 such books and $50,000 a month needs 100 to 167 (our sums)⁷¹.

We give more weight to the company’s low figures than its high ones, because a seller has no reason to understate early results. Nobody independent has checked either.

Step 5: the costs

Publishing through KDP is free⁴⁷. The money a newcomer can lose goes elsewhere.

• Adverts. Amazon’s book adverts are paid by the click: “Sponsored Products are cost-per-click (CPC) which means you only pay when customers click on your ad.”¹⁸ A click costs money whether or not the visitor buys. Amazon sets no minimum spend: you choose a bid for each click and a daily budget⁶⁸. Its own worked examples use bids of $0.75 and $0.60 a click and a budget of $10 a day, which it says caps a 30-day month at $300⁶⁸. These are illustrations, not averages. At $0.60 to $0.75 a click, a paperback that pays $3.69 a copy has to sell to about one visitor in five or six just to cover the clicks (our sum: 3.69 ÷ 0.75 and 3.69 ÷ 0.60)⁵⁶⁶⁸. We found no measured average for what a click costs or how many clicks become sales. The one measured figure we have is a single seller’s: Amazon adverts costing 25% to 33% of the sales they brought, in a broker’s listing⁵⁵.
• Making the book. The company’s site gives about $150 a cover²². In a 2019 survey, authors who had earned over $60,000 in a year mostly reported spending “between $250 and $1,000 on editing services” and $100 to $500 on a cover for each book²⁹. In the 2025 survey, authors earning over $10,000 a month mostly paid $250 to $1,999 for editing and at least $250 for a cover; of those earning under $100 a month, more than half spent $0 to $100 on a cover²⁶. Reedsy, a marketplace that sells these services, puts a full-length book of 80,000 words at $2,940 to $5,660, with a cover averaging $930⁶⁹. A 10,000-word guide would cost far less to edit (our inference).
• Tools. ChatGPT lists a free plan and paid plans at $8, $20 and from $100 a month⁷⁰. Claude lists a free plan and a Pro plan at $20 a month, or $17 a month billed yearly⁷⁰. Atticus, a tool that lays out a book’s pages, is $147 once⁷⁰. Publisher Rocket, a keyword tool for KDP, lists a one-time price of $199, shown as reduced from $299³⁷. Together that is about $350 once plus about $20 a month (our sum)³⁷⁷⁰. Tools are the small cost. Adverts and covers are the large one.
• Waiting. Amazon pays “approximately 60 days after the end of the month in which the sale was reported”¹⁴. A sale in October is paid around the end of December. Adverts are paid for at once. Authors outside the United States also face “30% US tax withholding” on Amazon.com royalties unless a tax treaty lowers it¹⁹.

What the tweets leave out

Of the 30 most-viewed tweets, almost all state an income figure. In the text we collected, none gives a book price, a royalty per copy, a number of copies sold, an advertising budget or a profit after costs¹. Several texts are cut off in our copy, and we did not see the images or the guides. Three mention time, always about one hour a day¹. Twenty of the 30 are labelled in our collection as a giveaway post or a pitch for a guide, training or course¹.

What people who tried it report

An earlier draft of this page said we had found no data on what AI-written books earn as a group. The second round of research found one study, and it covers fiction ebooks only⁵⁶. There is still no figure from Amazon. The rest is surveys of self-published authors in general, a little from Amazon, a business broker’s records (set out under “The upside”), one regulator’s case, and individual accounts. The surveys describe people who mostly write their own fiction. They answer “what do committed self-publishers earn?”, not “what does an AI book earn?”.

A study of AI-written fiction. Four university researchers ran AI-detection software on the full text of 14,419 self-published genre-fiction ebooks released from January 2023 to March 2026, and matched them to daily sales figures held by one of the five largest publishers⁵⁶. Their findings:

• Books with substantial AI text (more than a quarter of the book) were 20% of the books but took 12.1% of sales and 11.3% of revenue⁵⁶. So an AI-heavy book sold about 60% as many copies as the average book in the sample (our sum: 12.1 ÷ 20)⁵⁶.
• Books with no AI text were 62.9% of the books and took 72.5% of revenue⁵⁶.
• The paper says most AI-heavy books sold very few copies, while a minority sold a great deal⁵⁶. Those are described under “The upside”.
• The market got more crowded. From early 2023 to early 2026 the number of books selling in a quarter grew 19.2-fold, while quarterly revenue grew 8.9-fold⁵⁶. Revenue per selling book fell in six of eight genres⁵⁶.

The limits matter. The study covers genre fiction, not the non-fiction in the tweets⁵⁶. It leaves out books whose sales rank never got better than 1.5 million, which means books that sold almost nothing are not counted⁵⁶. Its revenue is what readers spent, before Amazon’s cut, so authors received less⁵⁶. It is a working paper that other academics have not yet reviewed⁵⁶.

The tweets’ own genre. Originality.ai, a company that sells an AI detector, tested 844 new paperbacks in Amazon’s “Success” self-help category, published from 31 August to 28 November 2025, and judged 77% of them likely AI-written⁶⁴. Those books averaged 26 reviews, against 129 for the ones it judged human-written⁶⁴. It tested only the public text of each book (the summary, the author note and the sample pages), and reviews are a rough stand-in for sales, not sales⁶⁴.

How many books are published. Publishers Weekly, a trade magazine, reported a count by Bowker, the US agency for ISBNs (a book’s registration number): 3,529,980 self-published titles with an ISBN in 2025, up 38.7% from 2,545,885 in 2024⁶⁵. Kindle ebooks published without an ISBN are not in that count, so the real rise is larger (our inference).

Surveys of authors.

• Written Word Media surveyed 1,346 independent authors in 2025. It reports: “While 44% of authors earn $100 or less, the majority—56%—are earning over $100 monthly from their writing.”²⁶ About 13% reported over $5,000 a month and 8% over $10,000²⁶.
• In the same survey, income follows the number of books. “Authors with 1–3 books: About 80% fall into the under $100/month band.” Authors with 25 or more books reported a middle figure of about $3,000 a month²⁶.
• Those earning over $10,000 a month spent about $4,500 a month on marketing; those earning under $100 spent about $81²⁶.
• Romance is 21% of the authors surveyed and about 44% of those over $10,000 a month²⁶. General non-fiction, the kind in the tweets, is 6% of those surveyed and “Rare among high earners”²⁶. The survey gives a reason: “For many non-fiction writers, the book itself functions as a branding asset” for other paid work²⁶. So few of these authors live on non-fiction royalties, but the survey does not measure non-fiction paperbacks sold with adverts.
• The Authors Guild surveyed 5,699 published authors in 2023: “The median income of full-time self-published authors in 2022 was $12,800 from books and $15,000 total from all author-related activities.”²⁷ “Median” means the middle person: half earn more, half less. That is about $1,070 a month. Those publishing since at least 2018 “reported a median income of $24,000 compared to $13,700 in 2018, a 76 percent increase”²⁷.
• The Alliance of Independent Authors says its 2025 survey “found that median self-published author income was $13,500, growing at a healthy 6% year-on-year whereas the typical traditionally published author earned only $6k–$8k and was trending down”²⁸. That is the Alliance’s summary; we did not see the 2025 survey itself or who it asked. Its 2023 survey asked only authors who spent half or more of their working time on writing and publishing²⁸. The September 2025 announcement of the new survey sets the same kind of entry rule: at least one self-published book in English, and half of working time or half of income from writing and publishing⁷². So the $13,500 describes committed authors, not newcomers.
• The University of Glasgow’s CREATe research centre analysed that 2023 survey: 2,261 committed authors reporting their own 2022 revenue, before costs⁵⁸. The median was $12,749 in the Alliance’s release ($12,755 in the paper), and the release says almost a quarter reported $0 to $1,000⁵⁸. Authors who first published in 2020 to 2022 had a median of $2,000 for the year (437 people); those who started in 2010 to 2019 had a median of $22,256⁵⁸. This is the best figure we found for the first one to three years of someone working at it at least half-time.
• A 2019 Written Word Media survey of over 1,000 authors found that those who had ever earned over $100,000 in a year had a median of 28 books and wrote about 32 hours a week. Those who had never earned over $60,000 in a year had a median of 6 books²⁹.

All of these surveys are filled in by people who chose to answer, nobody checks the figures, and none is stated as income after marketing costs. Written Word Media says of its own: “The survey skews toward serious, prolific indie authors (that’s our audience!), and all data is self-reported.”²⁶ The publishing writer Jane Friedman calls it “a tortured exercise to try and run any kind of meaningful survey on what authors earn”³⁰. People who published one book, earned nothing and stopped rarely answer. So a newcomer’s chances are probably worse than these numbers (our inference).

What Amazon has said. An earlier draft of this page gave 2019 as Amazon’s newest count. The second round found one from May 2022: “Thousands of independent authors have already earned more than $50,000 in KDP royalties so far in 2022, and more than 2,000 of those authors have surpassed $100,000 in royalties.”⁵⁷ Amazon added that the number earning $50,000 or more had risen 40% in two years⁵⁷. For 2019 it had said: “Millions of independent authors have self-published millions of books through KDP since launching the service in 2007. Additionally, thousands of independent authors earned more than $50,000, with more than a thousand surpassing $100,000 in royalties in 2019 through KDP.”³¹ Its 2017 shareholder letter gave a similar count³². So high earners exist, and the top group about doubled between 2019 and 2022, on Amazon’s own unaudited word³¹⁵⁷. Two searches found no count from Amazon for 2023 to 2026. “Thousands” out of “millions” is a small share, though the “millions” counts everyone since 2007³¹. These figures are royalties before the author’s costs, and all are from before AI-written books were common.

In the press. Reuters reported in February 2023 on a first-time publisher whose children’s ebook, written with ChatGPT, had earned him “less than $100”, he said⁶². He was not described as selling anything on the subject⁶². We read a syndicated copy of the report.

In our own tweets. Three of the 52 tweets come from accounts labelled in our collection as having nothing on offer; we did not check the profiles. Two give numbers.

One, from October 2026, says its author published two books and that four months later the two had made $340 between them¹. That is about $42.50 a book a month (our sum: 340 ÷ 2 ÷ 4). The other describes a friend who wrote a series of four or five books, ran adverts, and then passed $1,000 in each of two months¹. We paraphrase both and do not name the accounts. Neither tweet says the books were written by AI; both describe the author’s own writing. Both are unverified, like the large claims.

@TommiPedruzzi gave his own timeline in a tweet from September 2024: “Month 1: $0 Month 3: $2,375 Month 6: $8,431 Month 18: $114,000”¹. We could not verify it. Taken at face value it describes 18 months, and the same account says it had more than 100 books by October 2023¹.

One course seller’s customers, and a regulator’s case. Publishing.com sold courses on publishing books with AI software. It does not appear in our tweets, we found nothing linking it to Nespola, and we did not establish how close its method was to the one tweeted. We include it because it is the one KDP course seller for which we found a regulator’s record. These cases concern the companies named in them only. We found no regulator action against any account quoted or named on this page, and we do not suggest any connection.

In July 2026 the US Federal Trade Commission (FTC), the consumer-protection regulator, approved a final order against it. The FTC’s summary: “Publishing.com LLC and its two principals will pay $1.5 million and be required to substantiate earnings claims in the future to settle Federal Trade Commission charges that the company and its operators misled consumers about how much money they were likely to earn using their products.”⁴² The complaint says the company claimed buyers could “earn substantial passive income by utilizing Respondents’ methods and Publishing.com’s artificial intelligence (“AI”) software to write popular online books”⁴³. It quotes one advert as promising “$1k to $3k a month while only working 1 hour a day”⁴³. It gives the usual prices as $1,995 for the course and $9,800 for an add-on programme, and alleges that “many purchasers have made no profits at all”⁴³. These are allegations settled by a consent order, an agreement that ends a case without a trial. We did not see whether the respondents admitted or denied them.

On Trustpilot, a review site, Publishing.com is rated 4.3 out of 5 from 2,383 reviews, 87% of them five-star³⁴. The FTC alleged that the company “unlawfully incentivized reviews, including by offering consumers free one-on-one coaching in exchange for giving Publishing.com a five-star review on Trustpilot”⁴³. Among the roughly 20 most recent reviews we could read, the ones that give both money spent and money earned report losses, and most of the money spent was course fees. One, dated 20 August 2025: “I bought the course thinking that I would be able to make a living, writing, and selling books. I’m over two years into it and I spent about $30,000 and I have received a little over $1000.”³⁴ The company replied publicly: “Success in publishing often takes time, and persistence”³⁴. Another, from October 2025: “I am still very angry that I lost 9,800 dollars from joining Publishing.com.”³⁴ A positive review says “I was able to create a retirement income”, with no figure³⁴. Reviews are unverified in both directions.

An earlier draft said we had found no income report from an AI-book publisher that was backed by payment records. That has changed in part. A business broker lists an AI-assisted fiction catalogue whose revenue the broker says it has checked, and a study built on sales data shows some AI-heavy fiction selling in large numbers⁴⁸⁵⁶. We did not see the payment records ourselves. Both are fiction. For the route in the tweets, AI-written non-fiction, the large income figures still come from the co-founders of a company that sells a programme for it. The small ones come from accounts labelled in our collection as having nothing on offer and from one press report¹⁶². All the loss figures we found come from one course seller’s reviewers and the complaints the FTC quoted.

What the rules allow now

AI-written books are allowed, if declared. Amazon says: “We require you to inform us of AI-generated content (text, images, or translations) when you publish a new book or make edits to and republish an existing book through KDP.”⁸ Editing does not change the label: “If you used an AI-based tool to create the actual content (whether text, images, or translations), it is considered “AI-generated,” even if you applied substantial edits afterwards.”⁸ Text you wrote yourself and only polished with AI need not be declared⁸. The declaration goes to Amazon. The Authors Guild wrote in 2023 that Amazon “has not yet agreed to add disclosures to all book listings for AI-generated books on its other channels outside of KDP”²¹; we did not check whether that has changed.

Mass-produced books can run into the quality rules. Amazon’s quality guide lists as not allowed: “Content that is not significantly different from content in another book available in the Kindle Store”, “Content that is too short”, and “Content that is excessively reused, recycled, or repeated within or across books”⁹. Whether a given book counts as “significantly different” is Amazon’s decision. We have not seen any of the books published by the people quoted here, and none of their tweets describes copying.

Keywords and reviews. Keywords must not “mislead or manipulate” shoppers¹⁵. A new book has no reviews, and paying for them is barred: “Offering anything other than a free or discounted copy of the book, including gift cards, will invalidate a review”¹⁶.

Kindle Unlimited has strings. To be in it, a book is enrolled in a programme called KDP Select: “you’re committing to make the digital format of that book available exclusively through KDP”¹⁰. Amazon’s terms also say that if “we determine that borrows or reading of your Digital Book originated from accounts attempting to manipulate our services, then we will not owe you Royalties” for that programme¹³.

The account is what you stand to lose. Amazon’s terms, last updated 27 September 2024, say:

• “We are entitled to terminate this Agreement and your access to your Program account or suspend your Program account at any time if we have concerns with your account, or activity relating to your account, or if you are in breach of applicable laws or this Agreement.” Amazon adds that it “will provide instructions regarding how to appeal”¹³.
• “If we terminate this Agreement because you have breached your representations and warranties or our Content Guidelines, you forfeit all Royalties not yet paid to you.”¹³
• “You may maintain only one account at a time. If we terminate your account, you will not establish a new account.”¹³

Because Amazon pays about 60 days after the month of sale¹⁴, two to three months of royalties are always unpaid and can be lost this way. That is our inference from the two rules.

Two broker listings mention trouble of this kind at first hand. One seller writes: “My KDP and ACX account was terminated when I first opened it in 2023”; the listing says it was restored⁵⁵. (ACX is Amazon’s audiobook service.) Another listing says: “One title was blocked by Amazon as part of a broader AI-driven compliance review that impacted multiple publishers.”⁴⁹ We found no count of how many accounts Amazon closes.

Selling an account. One tweet, from @NickDiFabio1 in March 2024, says: “You can start an Amazon KDP account for $0 & sell it for $70,000+.”¹ Amazon’s terms say: “You may not permit any third party to use the Program through your account”¹³. They also say neither side may hand over the agreement without the other’s written consent, “except that (a) Amazon may assign any of its rights and obligations under this Agreement without consent and (b) you may assign all of your rights and obligations under this Agreement to any corporation or other entity without consent in connection with the sale of all or substantially all of your assets”¹³. So a publisher can sell a whole publishing business to a company and pass the agreement on. Handing a login to a buyer is a different thing, and a buyer who takes one risks losing the account and the royalties in it. The tweet does not say how the sale was made, and we do not suggest it broke Amazon’s terms.

The second round of research found that whole KDP businesses are sold openly. Empire Flippers, a broker of online businesses, showed eight KDP-only businesses for sale when we looked, at asking prices from $88,990 to $1,190,151, which is about 1.7 to 2.5 times a year’s net profit⁵³. One listing says: “Buyers will have the option to either take over the full KDP account or merge the books into an existing account.”⁴⁹ At those multiples, a sale at “$70,000+” implies roughly $28,000 to $41,000 a year of proven profit (our sum: 70,000 ÷ 2.5 and 70,000 ÷ 1.7), so what is sold is the profit record of the books in the account, not the account itself⁵³. We did not check how a takeover of an account fits the terms quoted above.

Copyright. The US Copyright Office wrote in January 2025: “Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements.”²⁰ So a book where the AI wrote every word can be copied and resold by someone else, and the publisher may have little legal answer.

You carry the legal risk. Amazon holds you responsible for what the AI wrote, “including by complying with all applicable intellectual property rights”⁸. You also agree to “indemnify, defend and hold Amazon … harmless from and against any loss, claim, liability, damage, action or cause of action (including reasonable attorneys’ fees and costs) that arises from any breach or alleged breach of your representations, warranties or obligations”¹³. To indemnify means to cover the other side’s losses.

Who makes money from it

Amazon, on every sale. Amazon keeps 30% of an ebook’s price plus the delivery charge, or 65% on the lower rate². On a paperback it keeps 40% or 50% and the printing charge⁵. It is also paid for every advert click¹⁸. It earns whether or not the publisher makes a profit.

The people tweeting. In our 52 tweets, 45 come from four accounts whose tweets offer a course, training or guide. Two are the Nespola co-founders: @TommiPedruzzi (29 tweets) and @Manu_Sisti (11). The other two accounts wrote five tweets between them, four of them offering a free course or guide by direct message; we did not open their pages and do not know whether they sell anything¹. Twenty of the 52 are giveaway posts: like, comment and follow to be sent a free guide by direct message¹. Six say the free item usually costs $97 to $199; two more say they plan to or could charge $99 or $199¹. For example, from @TommiPedruzzi: “Usually, I’d charge $199 for my training, but for the next 48 hours it’s FREE.”¹ We did not see any of the guides. We looked for a sales page only on the company’s own sites and did not find one; that does not show the prices are untrue.

What Nespola’s own pages show. We read the company’s own pages; the two main accounts are listed there as its co-founders⁴⁵.

1. A free community on Skool, a site for hosting groups: “PublishingOS (Free)”, with 16.8k members when we opened it. Its description reads: “We’ve already proven it—$3M in royalties managed in just 1-3 hours per day.”²⁵ It does not say whose royalties these are, and gives no figures for members.
2. A paid programme at nespola.io: “We partner with people to build profitable Amazon KDP book portfolios — from 1st book to compounding asset.”²³ The tiers are Velocity (3 books in 6 months) and Accelerator (6 books in 6 months)²³. No price is shown; you apply²⁴.
3. The home page shows named students with monthly royalties, for example “Earns $144,039 in monthly royalties”²³. These are the company’s figures, unverified, and royalties are before advert costs. Its blog’s case studies give 10 to 49 books for the students named²².
4. The company says these are exceptions. Its site: “These are real results from real people. They are not typical and we do not guarantee them.”⁴⁵ Its terms: “Testimonials and case studies represent exceptional results and are not typical.”²⁴
5. The two sites carry different refund terms, as read on 7 October 2026. One gives a 14-day refund only if no more than 20% of the content was used. The other, “operated by Nespola LTD” under UK law, gives a 30-day window with exceptions²⁴. A buyer should check which site they are paying through.
6. Nespola’s site points readers to Trustpilot⁴⁵. There it is rated 3.7 out of 5 from 22 reviews when we looked, most of them five-star⁴⁶. One, from February 2024, says “now I’m hitting $3,500 a month from my book sales”⁴⁶. One, from the start of April 2025 (shown as 1 or 2 April depending on how the page is read), says the reviewer paid “$9k for 3 ghostwritten books” and that Amazon later closed their KDP account. Nespola replied publicly that “the ban on your KDP account was not due to the books produced by Nespola, but appears to be connected to third-party involvement, outside of our scope”⁴⁶. We have no way to tell which account is right, and we do not suggest Nespola’s books caused the closure. Reviews are unverified in both directions.

Tools and platforms. Some products sold to KDP publishers pay commission to people who recommend them³⁶³⁷.

Sellers of services and brokers. Editors, cover designers and ghostwriters are paid for each book whether or not it sells⁶⁹. A broker that sells KDP businesses has its own interest in each sale going through⁵³. Two of the businesses it lists come with training attached: one sale includes “a 30+ hour professional training program and access to a private operator community”, and one seller of a sold catalogue also made KDP courses⁴⁹⁵⁰.

Regulators. The Publishing.com case above is the one regulator action we found about book-publishing courses⁴²⁴³⁴⁴. A search in the second round, across the FTC, state attorneys general and lawsuits, turned up no action against any other KDP course seller. The order was approved in July 2026; the company’s help page now says “Publishing.com has concluded operations”³³. Neither of its pages gives a reason, and we do not claim one caused the other. The FTC has also brought cases over a different product, AI-branded online shops sold as income opportunities³⁸³⁹⁴⁰. One of them ended in a final order in May 2025³⁹; we did not check how the others ended. These cases concern the companies named in them only. We found no regulator action against any account quoted or named on this page, and we do not suggest any connection.

One FTC rule gives a useful test. Where it applies, a seller of a business opportunity who makes an earnings claim must give a statement showing what share of buyers achieved it, and must hand over a disclosure document “seven days before the prospective buyer signs a contract or pays any money”⁴¹. Whether a KDP coaching programme falls under that rule is a legal question we did not settle. No seller’s page we opened gives the share of students who reach any income level. We do not say any of them is required to.

The upside

Self-publishing on Amazon pays real money. Amazon reports “Total KDP Select Author Earnings” of $69.4 million for August 2026 alone¹⁰. A third-party tracker puts the monthly Kindle Unlimited fund at $45.1 million in August 2022 and $66 million in August 2026¹². From 2020 to 2026 the rate per page stayed between about $0.0040 and $0.0052¹². So the pot has grown, though far more slowly than the number of books sharing it¹²⁵⁶.

How many reach the top. Amazon said in May 2022 that more than 2,000 authors had already passed $100,000 in royalties that year⁵⁷. That is Amazon’s own count, unaudited, and before the author’s costs. An older outside estimate, worked out from Amazon’s sales ranks in May 2016, counted more than 425 independent authors who had started within the previous five years and were earning at a rate of $100,000 or more a year⁶⁰. We read that estimate through a news report, not the original, and it is ten years old.

What the top tenth and the top hundredth earn. These figures are for committed authors, are self-reported, and are before costs.

• In the 2023 survey of 2,261 committed authors, the paper says the top 10% earned $190,000 in 2022 and took 71% of all the revenue reported; the top 1% earned $1.2 million and took 31%⁵⁸. We read these as the entry points to each group (our reading). The median in the same survey was about $12,750⁵⁸.
• In the 2025 Written Word Media survey, about 13% reported over $5,000 a month and 8% over $10,000²⁶.
• In the Authors Guild’s survey, full-time romance authors reported the highest median income from books, $31,725 in 2022²⁷.

Businesses a broker has checked. Empire Flippers is a broker that sells online businesses, including KDP catalogues (a catalogue is the set of books a publisher has on sale). It says: “All businesses are fully reviewed and vetted by our team of advisors before being listed on the marketplace”, and lists checks of the seller, the revenue, the traffic and the expenses⁵³. This is the nearest thing to payment-checked income we found. It has four limits:

• The broker has an interest in the sale. Its check is not an audit.
• The proof sits behind a buyer’s login. We did not see it.
• These are survivors: businesses good enough to put up for sale, not a sample of everyone who published.
• Most of the listing pages are undated. The profit figures appear to be for one year, because the asking price divided by the stated multiple gives the same number (our reading).

What the listings show:

• Non-fiction catalogues that sold. Eleven businesses marked as sold, created from 2017 to 2020, showed net profit from $14,964 to $117,205 a year, or about $15,000 to $117,000⁵⁰. “Net profit” is what is left after adverts and other costs. Most were outsourced. In seven of the eleven, ghostwriters (paid writers whose names do not appear) or freelancers wrote the books. In three the seller wrote most or all of them, and in one the seller wrote 10 of 26 non-fiction books and a ghostwriter wrote the earlier ones (our count)⁵⁰. They were sold under pen names (invented author names). Nine mention advert campaigns; two list paid adverts only as something a buyer could add (our count)⁵⁰. An earlier version of this page said all eleven were ghostwritten and sold with adverts; that was wrong. The pages give no sale dates. From the creation dates and details in the listings we estimate the businesses were two to five years old when sold (our estimate)⁵⁰. The books were created before AI writing tools were common⁵⁰. The highest, with 21 exam-preparation books, says: “The two top-selling books make up ~90% of revenue.”⁵⁰ Eight of the eleven netted under $58,000 a year⁵⁰. One of the sellers also made KDP courses⁵⁰.
• The closest match to the tweets. A business launched in 2023 with about 40 non-fiction titles (health and fitness, relationship gift books) shows net profit of $152,780 a year on revenue of $294,234⁴⁹. It uses Amazon adverts, Meta adverts, several TikTok accounts, an email list and “custom GPTs”, which are tailored versions of ChatGPT⁴⁹. Its top title has more than 900 reviews and brings about a third of sales⁴⁹. It is for sale, not sold. The listing says it was “Built and scaled directly by the Seller alongside a specialized team”, and the sale includes “a 30+ hour professional training program and access to a private operator community”⁴⁹. So it is not one person’s result at an hour a day, and the seller appears to be linked to a training business. We do not know whose.
• A small recent one. Twelve ghostwritten titles for children and teens, launched in December 2024, show net profit of $39,551 a year on revenue of $112,459⁵⁵. That is about $3,300 a month, with 35% of revenue kept (our sums)⁵⁵. The owner reports about three hours a week once it was built⁵⁵. It is for sale, not sold.
• An AI-assisted fiction catalogue. A business that first earned money in May 2025 shows net profit of $517,457 a year on revenue of $525,581⁴⁸. It has more than 500 books under five main pen names, earns heavily from Kindle Unlimited, and spends under 1% of revenue on adverts⁴⁸. The listing describes “a repeatable custom AI-assisted production system” that let it “publish approximately 10 new books per week”⁴⁸. It is for sale at $1,190,151, not sold, and was listed on 3 September 2026⁴⁸. Amazon’s limit of two titles per format a week began on 21 September, so that pace can no longer be copied on one account⁷⁶⁶. It is fiction, not the short non-fiction in the tweets. No course is mentioned, though the sale includes 30 days of training⁴⁸.
• Low-content books. A sold business with 147 low-content books netted $19,411 a year on revenue of $28,552; “the top two selling books account for 50% of sales”⁵¹. One for sale, with template-based books under about 67 pen names, shows $144,460 a year, is strongly seasonal, and notes “past account-level issues”⁵¹. A third, with 50 low-content books, shows $60,899 a year, with over 70% of profit from one title that has more than 2,500 reviews⁵¹.
• Romance and other fiction. Eight sold romance businesses netted from $35,172 to $269,568 a year⁵². The highest had 32 ebooks under one pen name and was created in 2020⁵².
• Everything on sale now. The eight KDP-only businesses listed when we looked showed net profit from $39,551 to $517,457 a year⁵³.

AI-heavy books that sold. In the study of 14,419 fiction ebooks, the best-selling book with substantial AI text took $643,000 from readers on 80,431 copies in its first 12 months⁵⁶. The 15 best each took more than $220,000⁵⁶. The top pen name took $1.7 million across 8 titles in 12 months⁵⁶. These come from a sales panel, not from the authors, and the authors are not named⁵⁶. They are what readers paid, before Amazon’s cut, so the authors received less⁵⁶. All are fiction.

People who have told their own story. Each of these is the person’s own statement.

• “Coral Hart”, a pen name, told the New York Times in February 2026 that she had published more than 200 AI-written romance novels under 21 pen names in a year, sold about 50,000 copies and earned six figures⁶¹. That is about 250 copies a book⁶¹. She sells courses and a tool for AI writing, according to those reports⁶¹. We could not open the newspaper’s article and read two reports about it.
• Amazon’s 2022 article profiles two authors. One, a romance writer with 12 books, wrote for 90 minutes a day⁵⁷. The other had 67 science-fiction and fantasy books, went full-time in 2020 and says he makes “well into six figures annually”⁵⁷. Neither is described as selling a course⁵⁷.
• @TommiPedruzzi’s own timeline, quoted above, runs from $0 in month 1 to $114,000 in month 18¹. Nespola’s site shows students with large monthly royalties and 10 to 49 books²²²³. The company sells a programme and says the results “are not typical”⁴⁵.
• A tweet from a small account, labelled in our collection as having nothing on offer, describes a friend who passed $1,000 in each of two months after writing a series of four or five books and running adverts¹.

A realistic good result (our estimate). This is our reading of the figures above. None of it is measured for AI-written non-fiction, and every survey leaves out people who gave up.

• The common start. Under $100 a month for the first one to three books, and $0 to $300 a month for the first six months²²²⁶.
• A good first year. A few hundred dollars to about $2,000 in royalties, before costs. $2,000 a year was the median for committed authors in their first three years⁵⁸.
• A good result after two to three years, with 5 to 10 decent books. About $4,600 a year, before costs, is the median for committed authors with that many books⁵⁸.
• Well above that: the businesses that reached a sale. The lower eight of the eleven sold non-fiction catalogues netted about $15,000 to $58,000 a year after costs⁵⁰. Where the listing gives a count they had from 11 to more than 300 books, and we estimate they were two to five years old (our estimate)⁵⁰. They are the ones that did well enough to be sold, not typical results.
• Full-time, the middle result. About $1,000 a month is the middle result for full-time authors, before marketing costs as far as the surveys say²⁷.
• Six figures. $10,000 a month is reported by about one in twelve of a committed group, usually with many books and a four-figure monthly advert budget²⁶. The Alliance’s release on the 2023 survey says almost a fifth of committed authors “ran six-figure publishing businesses”, before costs⁵⁸. It typically comes with more than 20 books and several years²⁹⁵⁸.

What it takes to compete

The people who earn well differ from the rest in ways the evidence can show. The surveys are of people who mostly write their own fiction. The broker’s records are mostly of outsourced catalogues.

1. A catalogue. In the 2023 survey of committed authors, median income for 2022 rose with the number of books⁵⁸:

   Books published	Median income, 2022
   1	$618
   2 to 4	$1,254
   5 to 10	$4,600
   11 to 20	$21,000
   21 to 30	$50,000
   Over 30	$95,385

   These figures are before costs. The researchers add two cautions: more books means more streams of income, so the two feed each other, and some genres (they name short fiction and erotica) lend themselves to many titles⁵⁸. The 2025 survey shows the same shape. About 80% of authors with one to three books were under $100 a month; the median at five to nine books was about $50 a month; authors with 25 or more reported a median of about $3,000 a month, and over 40% of them earned $5,000 or more²⁶. In the 2019 survey, authors who had earned over $100,000 in a year had a median of 28 books²⁹. With a limit of two new titles per format a week, a large catalogue now takes years to build⁷.

2. A hit, not a pile. In the broker’s records, one or two books carry most of the income. One spirituality business gets over 96% of its revenue from a single book with almost 2,000 reviews⁵⁵. Others get about 90% from two books, 65% from three, and about a third from the top title⁴⁹⁵⁰. The clearest counter-example is a sold business with about 2,090 books under more than 100 pen names. It netted $34,466 a year, about $2,872 a month, or roughly $16 a book a year (our sums)⁵⁴. Its listing says: “Around 75% of the earning comes from five/six evergreen books”⁵⁴. So the skill being paid for is choosing and presenting titles, not volume alone (our reading).

3. Years. Of authors who had earned over $60,000 in a year, 69% had published their first book five or more years earlier²⁹. In a 2017 survey by the same company, 88% of those who had earned over $100,000 in a year had been publishing for more than three years⁵⁹. Full-time self-published authors publishing since at least 2018 reported a median income of $24,000 in 2022²⁷. There are faster cases among the broker’s listings, such as the two launched in December 2024 and May 2025, but they are the exceptions that reached a sale listing⁴⁸⁵⁵.

4. Money for adverts, and room for a losing start. In the 2025 survey, monthly marketing spend rose with income: about $81 for those earning under $100 a month, $275 for those at $500 to $1,000, $478 at $1,000 to $5,000, $1,362 at $5,000 to $10,000, and $4,500 above $10,000²⁶. The share of revenue kept varies widely. Three recent advert-led non-fiction listings kept 30%, 35% and 52%, and a fourth kept 76%. The eleven sold catalogues kept from 36% to 97%, and nine of them kept more than half (our count)⁴⁹⁵⁰⁵⁵. An earlier version of this page called the share thin; the listings do not support that. Nespola’s own site says net profit can be near zero in months 1 to 4²².

5. Readers they can reach directly. Surveyed authors with an email list reported a median of about $300 a month, against about $15 without one²⁶. The average list ran from 902 subscribers for those under $100 a month to 18,327 for those over $10,000²⁶. The spirituality business has a list of more than 40,000⁵⁵. The survey shows that the two go together, not that a list causes the income.

6. Books that sell each other. In an Amazon advertising case study, a fiction publisher earns $3.49 on the first book in a series but values a reader who finishes all eight at $42⁶⁷. That is why a series can afford clicks that a single short book cannot: one $9.99 paperback has only its own $3.69 to pay for them⁵⁶. Romance is 21% of the authors surveyed and about 44% of those over $10,000 a month; general non-fiction is 6% and “Rare among high earners”²⁶. Kindle Unlimited pays by the page read, so long books that people finish earn more than short ones¹¹¹². Among those earning $5,000 or more a month, 43% had all their books in KDP Select²⁶.

7. Money for each book. High earners mostly paid $250 to $1,999 for editing and at least $250 for a cover²⁶. The tools cost about $350 once and $20 a month (our sum)³⁷⁷⁰. One reviewer says they paid Nespola “$9k for 3 ghostwritten books”, which is unverified⁴⁶.

8. Working hours. The 2019 survey’s top group wrote about 32 hours a week²⁹. Full-time self-published authors had a middle income of about $13,000 a year²⁷. Sellers of outsourced catalogues report far fewer hours, from 1 to 10 a week, for running a business that is already built^48505154.

9. A subject that is not already full. A detector vendor judged 77% of new paperbacks in one self-help category likely AI-written⁶⁴. In fiction, the number of books selling grew more than twice as fast as revenue over three years⁵⁶.

10. Books that pass Amazon’s rules. Declared, different from what is already on sale, accurately described, on one account⁸⁹¹³.

The route in the tweets is non-fiction paperbacks sold with Amazon adverts. An earlier draft said the only evidence for it was the company’s own account. There is now a broker’s record as well: sold catalogues, most of them ghostwritten, netting about $15,000 to $117,000 a year, and one team-built business for sale at $152,780 a year⁴⁹⁵⁰. The company’s account adds named students with 10 to 49 books, one student’s first two months of losses, and near-zero profit for the first four months²².

The pitch to beginners in the tweet texts we collected (an hour a day, the first tools at $0, $3,000 in the first month) leaves out the catalogue, the hit, the years and the advert money. The promoter’s own account of his method includes two of them: a catalogue of hundreds of books, and paid Amazon adverts¹²².

How to tell early which side you are on. These checks are our judgement, built from the figures above.

• Copies a day against the table. $1,000 a month from a $9.99 paperback is about 9 sales a day; $3,000 is about 27. If your books together sell only one or two copies a day after three months, you are on the common side of the line (our judgement; we found no measured figure for what a typical new book sells).
• Advert cost per sale against the royalty. A 100-page paperback at $9.99 pays $3.69⁵⁶. If you spend more than that on clicks for each sale, every sale loses money. The one measured case we have ran adverts at 25% to 33% of the sales they brought⁵⁵.
• Is one title pulling ahead? In the checked businesses a lead title has hundreds or thousands of reviews and brings a third or more of the income⁴⁹⁵¹⁵⁵. A catalogue that keeps growing while no title breaks out looks like the business with about 2,090 books and $2,872 a month⁵⁴.
• Too early, or nobody wants it? A new book with no reviews and no adverts may sell nothing whatever its quality. If adverts bring clicks and still no sales, more waiting is unlikely to help.
• Do buyers come back? Later books selling to readers of earlier ones, and an email list that grows, are what the higher earners have²⁶⁶⁷.
• Money out before money in. Royalties arrive about two months after the month of sale¹⁴. Add up what you will spend on adverts and covers in the first four months and ask whether you can lose all of it.
• Would a shopper see the difference? If your book covers the same ground as ten others, in similar words, it matches what Amazon’s quality guide describes⁹.
• Whether the plan needs you to buy something first. Ask any seller what share of their students reach the income in the advert.

What we did not verify

• Searches. The first round ran no web searches. The second round ran 29, which covered the three that were owed: Amazon’s most recent statements on author earnings; reported results of AI-written KDP books and account closures; and court or regulator actions against other KDP course sellers. Finding nothing in a search does not prove nothing exists. We ran no search by name for regulator or court actions against the accounts quoted or named on this page.
• The broker’s records. The proof behind every Empire Flippers listing needs a buyer’s account and was not seen. “Verified” is the broker’s own vetting. The period the profit figures cover, and the dates of the sold listings, are not on the pages. We do not know whether the two businesses for sale will sell at their asking prices, whether the AI-assisted one declares its books to Amazon as AI-generated, or whether it can go on under the new publishing limit. Two listings show 98% of revenue kept as profit, which may reflect a short period or costs left out. How old the sold businesses were at sale is our estimate from the listings’ details, not a stated date.
• How AI-written non-fiction sells. The one study covers genre fiction, leaves out books with almost no sales, counts what readers spent and has not been peer reviewed. The self-help test measures reviews, not sales, and comes from a company that sells a detector. We found no figure for how many AI-written books sell nothing, and no count or rate of account closures. Two reports of stricter checks and closures in 2026 were seen only as search-result summaries and are not used.
• The share of all new publishers who reach any income level. Every distribution we found is a survey of committed authors or a sample that leaves out books with no sales.
• How many copies a typical new book sells, and what a click costs. This is still the largest gap in the arithmetic. We have the copies needed, not the copies a newcomer gets. Click prices of $0.30 to $0.80 appear on vendor blogs; we could not trace them to data and do not use them.
• The income claims in the tweets. “$50,000/month”, “$53,000 in a single month”, “$3M” from 1,500 books and the rest were neither confirmed nor disproved. We did not view the tweets’ attached images, videos, threads or guides. Several tweet texts are cut off in our copy. Tweet dates are worked out from ID numbers. No independently checked result was found for the two main accounts or for any Nespola student.
• Three tweets we left out. Three high-view tweets from the two main accounts (“$8,000 in your first 60 days”, “$3k in first month”, a “first $10k month by the end of December”) do not name the method in their text, so we did not use them as claims about this scheme.
• Nespola’s figures. Its student earnings, its $23-a-day target, its $300 to $500 a book and its profit margins are its own statements. The programme’s price was not found on its programme or pricing pages. Its Trustpilot page was first read through a tool that listed 19 reviews; a later reading on 7 October 2026 showed 22. The star shares were not confirmed. We could not tell whether the reviewer’s or the company’s account of the closed KDP account is right.
• The other accounts. For the accounts other than the two Nespola co-founders, what is on offer comes only from the tweet text and from labels in our collection. We did not open their profiles or pages, and do not know whether they sell anything or at what price. How the account sale described in the @NickDiFabio1 tweet was made is not stated in the tweet and was not checked.
• “Coral Hart”. The New York Times article refused our tools. Her income and sales are her own statements as relayed by two other sites.
• Exact quotes. Trustpilot refuses direct reading, so its reviews were read through a reader service (r.jina.ai), not in a browser. The Publishing.com review dated 20 August 2025 and the company’s reply to it were read again that way on 7 October 2026 and match what is quoted. The Nespola page was read again the same way on 7 October 2026: the review we cite, the company’s reply and the 3.7 rating match what is given here; that reading listed 19 reviews, so the count of 22 and the star shares rest on the earlier reading. The other Publishing.com reviews could not be opened again that day, so their wording rests on the earlier readings. Publisher Rocket and some Empire Flippers and Amazon pages were read the same way. The Written Word Media survey details added in the second round, the Originality.ai study, the Publishers Weekly report, the Alliance podcast and Amazon’s two author profiles were read through a tool that summarises pages, so their exact wording was not confirmed against the pages themselves.
• The FTC case. We read the case page, the press release and the final complaint, but not the final order itself, the public comments, or any statement by the respondents.
• The Kindle Unlimited rate. It comes from a third-party tracker. Amazon’s $69.4 million is labelled “Total KDP Select Author Earnings”; the tracker reports a “Global Fund” of $66 million for the same month. Amazon does not define its figure on that page, and we cannot tell whether the two measure the same thing. How Amazon counts pages for a short book was not checked.
• Survey details. The Alliance of Independent Authors’ 2025 survey report itself was still not found; we have only its summary and the announcement of the survey. The 2016 estimate of high earners was read through a news report.
• Earlier publishing limits. The start date of the two-titles-a-week limit comes from a podcast, not from Amazon, whose help page gives no date. Whether the 10-title limit before it was per week or per day was not confirmed, and one site’s “three per day” was not traced to Amazon.
• Amazon’s earnings counts. Nothing from Amazon later than May 2022 was found.
• Publishing.com. How many students it had and what they earned. Better Business Bureau content³⁵ did not load in full and is not used in the text.
• News reports. Vox, The Guardian, Wired, Reuters and Reddit refused the agents in the first round. The one Reuters report used was read in a syndicated copy.
• Other rules and costs. Whether the 70% ebook rate, which Amazon says “is for in-copyright works only”², is at risk for a fully AI-written book; Amazon’s account-level penalties for review abuse; the text of the FTC’s business-opportunity rule itself; ghostwriters’ prices on their own pages; audiobook rules; copyright law outside the United States; tax beyond US withholding; whether Amazon’s terms differ by country (we read the European page).
• Whether the people named have seen this page. The people and companies named here were not contacted.

Sources

“Read at source” means we opened the page itself on 7 October 2026. “Secondary” means we read an article or summary about the source, not the source itself. “Estimate” means our own sum.

1. Does It Pay collection of 52 tweets on this scheme, collected 2026-10-06; each tweet quoted is linked where it appears. Read from our collected copy on 2026-10-07; four were also read on X’s embed service. Counts by account and by label are estimates.
2. Amazon KDP Help, “Digital Book Royalties”. https://kdp.amazon.com/en_US/help/topic/G200644210 . Undated page. Read at source on 2026-10-07. Royalties per copy are our estimate.
3. Amazon KDP Help, ebook list price requirements. https://kdp.amazon.com/en_US/help/topic/G200634560 . Undated page; states a change effective 2026-07-07. Read at source on 2026-10-07.
4. Amazon KDP Help, ebook delivery costs. https://kdp.amazon.com/en_US/help/topic/G200634500 . Undated page. Read at source on 2026-10-07.
5. Amazon KDP Help, “Paperback Royalty”. https://kdp.amazon.com/en_US/help/topic/G201834330 . Undated page. Read at source on 2026-10-07.
6. Amazon KDP Help, paperback printing cost. https://kdp.amazon.com/en_US/help/topic/G201834340 . Undated page. Read at source on 2026-10-07. Royalties per copy are our estimate.
7. Amazon KDP Help, title creation and submission limits. https://kdp.amazon.com/en_US/help/topic/G200635650 . Undated page. Read at source on 2026-10-07. The years needed are our estimate.
8. Amazon KDP Help, “Content Guidelines” (includes AI content). https://kdp.amazon.com/en_US/help/topic/G200672390 . Undated page. Read at source on 2026-10-07.
9. Amazon KDP Help, guide to Kindle content quality. https://kdp.amazon.com/en_US/help/topic/G200952510 . Undated page. Read at source on 2026-10-07.
10. Amazon KDP, KDP Select page (earnings figure and exclusivity). https://kdp.amazon.com/en_US/select and https://kdp.amazon.com/en_US/help/topic/G200798990 . Figure for August 2026. Read at source on 2026-10-07.
11. Amazon KDP Help, Kindle Unlimited royalties and pages read. https://kdp.amazon.com/en_US/help/topic/G201541130 . Undated page. Read at source on 2026-10-07.
12. Written Word Media, tracker of Kindle Unlimited payouts per page. https://www.writtenwordmedia.com/kdp-global-fund-payouts/ . Updated 2026-09-18. Secondary (a third party’s calculation, not Amazon’s). Payouts per book are our estimate.
13. Amazon, Kindle Direct Publishing Terms and Conditions. https://kdp.amazon.com/terms-and-conditions (redirected to kdp-eu.amazon.com/agreement). Last updated 2024-09-27. Read at source on 2026-10-07.
14. Amazon KDP Help, when royalties are paid. https://kdp.amazon.com/en_US/help/topic/GK2MKZUL6U3SFBPZ . Undated page. Read at source on 2026-10-07.
15. Amazon KDP Help, metadata guidelines for books. https://kdp.amazon.com/en_US/help/topic/G201097560 . Undated page. Read at source on 2026-10-07.
16. Amazon KDP Help, customer reviews. https://kdp.amazon.com/en_US/help/topic/G202101910 . Undated page. Read at source on 2026-10-07.
17. Amazon KDP Help, low-content books. https://kdp.amazon.com/en_US/help/topic/GGE5T76TWKA85DJM . Undated page. Read at source on 2026-10-07.
18. Amazon KDP Help, advertising for KDP books. https://kdp.amazon.com/en_US/help/topic/G201499010 . Undated page. Read at source on 2026-10-07.
19. Amazon KDP Help, tax withholding. https://kdp.amazon.com/en_US/help/topic/G201274690 . Undated page. Read at source on 2026-10-07.
20. US Copyright Office, “Copyright and Artificial Intelligence, Part 2: Copyrightability”. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf . 2025-01-29. Read at source on 2026-10-07.
21. Authors Guild, on Amazon’s disclosure policy for AI-generated content. https://authorsguild.org/news/amazons-new-disclosure-policy-for-ai-generated-book-content-is-a-welcome-first-step/ . 2023-09-07. Read at source on 2026-10-07; secondary for Amazon’s policy.
22. Nespola blog, on how much can be made on Amazon KDP. https://nespola.io/blog/how-much-can-you-make-amazon-kdp . Updated May 2026. Read at source on 2026-10-07. Books needed at the per-book target are our estimate.
23. Nespola, home page and Velocity programme page. https://nespola.io/ and https://nespola.io/velocity . Undated pages. Read at source on 2026-10-07.
24. Nespola and PublishingOS terms and application page. https://nespola.io/terms-of-service (updated 2026-04-03), https://www.publishingos.io/terms (May 2026), https://www.publishingos.io/home (undated). Read at source on 2026-10-07.
25. Skool, “PublishingOS (Free)” community page. https://www.skool.com/publishingos-free/about . Undated page. Read at source on 2026-10-07.
26. Written Word Media, 2025 indie author survey results. https://www.writtenwordmedia.com/2025-indie-author-survey-results/ . 2025-12-02. Read at source on 2026-10-07.
27. Authors Guild, “Key Takeaways from 2023 Author Income Survey”. https://authorsguild.org/news/key-takeaways-from-2023-author-income-survey/ . 2023-09-27. Read at source on 2026-10-07. Monthly figures are our estimate.
28. Alliance of Independent Authors, facts page citing its 2025 income survey. https://www.allianceindependentauthors.org/facts/ . Undated page. Read at source on 2026-10-07; secondary for the survey itself, which was not opened. Who the 2023 survey asked: https://selfpublishingadvice.org/indie-author-income-survey/ , 2023-02-27, read at source on 2026-10-07.
29. Written Word Media, “Author Income: How to Make a Living from Your Writing”. https://www.writtenwordmedia.com/author-income-how-to-make-a-living-from-your-writing/ . 2019-10-22. Read at source on 2026-10-07.
30. Jane Friedman, “How Much Do Authors Earn? Here’s the Answer No One Likes.” https://janefriedman.com/how-much-do-authors-earn/ . 2021-05-19, updated 2025-03-27. Read at source on 2026-10-07.
31. Amazon press release, fourth-quarter 2019 results. https://press.aboutamazon.com/2020/1/amazon-com-announces-fourth-quarter-sales-up-21-to-87-4-billion . 2020-01-30. Read at source on 2026-10-07.
32. Amazon, 2017 letter to shareholders (exhibit 99.1, filed with the US Securities and Exchange Commission). https://www.sec.gov/Archives/edgar/data/1018724/000119312518121161/d456916dex991.htm . April 2018. Read at source on 2026-10-07.
33. Publishing.com, help centre and home page. https://help.publishing.com/ and https://www.publishing.com/ . Undated pages. Read at source on 2026-10-07.
34. Trustpilot, reviews of Publishing.com. https://www.trustpilot.com/review/publishing.com and https://www.trustpilot.com/reviews/68a6598e5b13144c26f709d1 . Reviews shown dated 2024-12-09 to 2026-06-19. Read on 2026-10-07 through the reader r.jina.ai (direct reading refused). Reviews are unverified.
35. Better Business Bureau, profile of Publishing.com. https://www.bbb.org/us/tx/austin/profile/publishing-consultant/publishingcom-0825-1000222655 . Undated page. Read at source on 2026-10-07; not relied on.
36. Skool, pricing and affiliate programme. https://www.skool.com/pricing and https://www.skool.com/affiliate-program . Undated pages. Read at source on 2026-10-07.
37. Publisher Rocket, price and affiliate programme. https://publisherrocket.com and https://publisherrocket.com/affiliate-program/ . Undated pages. Read on 2026-10-07 through the reader r.jina.ai (direct reading refused).
38. Federal Trade Commission press release, “FTC Announces Crackdown on Deceptive AI Claims and Schemes”. https://www.ftc.gov/news-events/news/press-releases/2024/09/ftc-announces-crackdown-deceptive-ai-claims-schemes . 2024-09-25. Read at source on 2026-10-07.
39. Federal Trade Commission, complaint in FTC v. Empire Holdings Group (E.D. Pa., 2:24-cv-04949). https://www.ftc.gov/system/files/ftc_gov/pdf/2423027ecommercebuilderscomplaint.pdf . Complaint 2024-09-18; final order 2025-05-09. Read at source on 2026-10-07.
40. Federal Trade Commission, case page for FBA Machine / Passive Scaling. https://www.ftc.gov/legal-library/browse/cases-proceedings/x240032-fba-machinepassive-scaling-ftc-v . Updated 2025-07-30. Read at source on 2026-10-07.
41. Federal Trade Commission, guidance on the Business Opportunity Rule. https://www.ftc.gov/business-guidance/resources/selling-work-home-or-other-business-opportunity-revised-rule-may-apply-you . Undated guidance on a rule made final in 2011. Read at source on 2026-10-07.
42. Federal Trade Commission, case page “Publishing.com, In the Matter of” (matter 242 3055, docket C-4836). https://www.ftc.gov/legal-library/browse/cases-proceedings/242-3055-publishingcom-matter . Final order 2026-07-06. Read at source on 2026-10-07.
43. Federal Trade Commission, final complaint, In the Matter of Publishing.com, LLC. https://www.ftc.gov/system/files/ftc_gov/pdf/2423055c4836publishingcomfinalcomplaint.pdf . 2026-07-06. Read at source (full text) on 2026-10-07.
44. Federal Trade Commission press release, “FTC Approves Final Order Against Publishing.com, Settling Allegations It Misled Consumers”. https://www.ftc.gov/news-events/news/press-releases/2026/07/ftc-approves-final-order-against-publishingcom-settling-allegations-it-misled-consumers . 2026-07-02. Read at source on 2026-10-07.
45. Nespola, “People” and “Approach” pages. https://nespola.io/people and https://nespola.io/approach . Undated pages (copyright 2026). Read at source on 2026-10-07.
46. Trustpilot, reviews of Nespola. https://www.trustpilot.com/review/nespola.io . Reviews shown dated 2023-03-12 to 2025-11-24. Read on 2026-10-07 through the reader r.jina.ai. Reviews are unverified.
47. Amazon KDP, home page. https://kdp.amazon.com/en_US/ . Undated page. Read at source on 2026-10-07.
48. Empire Flippers, listing 94912 (AI-assisted fiction KDP business, for sale). https://empireflippers.com/listing/94912/ . Listed 2026-09-03 (page metadata). Read at source on 2026-10-07. Broker-vetted figures; the proof documents were not seen.
49. Empire Flippers, listing 92207 (non-fiction KDP business of about 40 titles, for sale). https://empireflippers.com/listing/92207/ . Undated page. Read at source on 2026-10-07. Broker-vetted figures; the proof documents were not seen.
50. Empire Flippers, eleven sold non-fiction KDP listings. https://empireflippers.com/listing/56997/ , and the same address with 61723, 52005, 44501, 76485, 59055, 58057, 57647, 59387, 62894 and 62512. Undated pages; businesses created 2017 to 2020. Read on 2026-10-07 through the reader r.jina.ai. The range, the count under $58,000, the counts by who wrote the books and by adverts, the shares kept and the age at sale are our estimate.
51. Empire Flippers, three low-content KDP listings. https://empireflippers.com/listing/53884/ (sold), https://empireflippers.com/listing/92315/ (for sale) and https://empireflippers.com/listing/83091/ (for sale). Undated pages. Read at source on 2026-10-07; the first two through the reader r.jina.ai.
52. Empire Flippers, eight sold romance and fiction KDP listings. https://empireflippers.com/listing/58548/ , and the same address with 47079, 58156, 51457, 73757, 47452, 67919 and 78303. Undated pages. Read on 2026-10-07 through the reader r.jina.ai.
53. Empire Flippers, “Verified by Empire Flippers” and its marketplace page for KDP businesses. https://empireflippers.com/verified-by-empire-flippers/ and https://empireflippers.com/marketplace/?monetization=kdp . Undated pages; the marketplace is a live list. Read at source on 2026-10-07. Profit implied by a $70,000 sale is our estimate.
54. Empire Flippers, listing 52985 (sold KDP business of about 2,090 books). https://empireflippers.com/listing/52985/ . Undated page. Read at source on 2026-10-07. Monthly and per-book figures are our estimate.
55. Empire Flippers, listings 94013 (spirituality), 91097 (yoga, survival, cookbooks), 90998 (children and teens) and 94174, all for sale. https://empireflippers.com/listing/94013/ , https://empireflippers.com/listing/91097/ , https://empireflippers.com/listing/90998/ , https://empireflippers.com/listing/94174/ . Undated pages. Read at source on 2026-10-07. Monthly figures and shares kept are our estimate.
56. Chakrabarty, Liu, Ginsburg and Dhillon, “Generative AI floods and dilutes the market for books” (working paper). https://arxiv.org/abs/2607.20349 . 2026-07-22, revised 2026-08-03. Read at source on 2026-10-07. Not peer reviewed.
57. Amazon, “How these authors found success with Amazon Kindle Direct Publishing”. https://www.aboutamazon.com/news/books-and-authors/how-these-authors-found-success-with-amazon-kindle-direct-publishing . 2022-05-17. Read at source on 2026-10-07; the two author profiles through a summarising tool.
58. CREATe, University of Glasgow, working paper 2023/4 on the Alliance of Independent Authors’ income survey. https://eprints.gla.ac.uk/301455/1/301455.pdf . June 2023. Read at source on 2026-10-07. The Alliance’s release on the survey: https://www.allianceindependentauthors.org/media-releases/alliance-independent-authors-survey-reveals-self-publishing-authors-earn-more/ , 2023-04-17.
59. Written Word Media, “What Makes a $100K Author”. https://www.writtenwordmedia.com/100k-author/ . 2017-06-07. Read on 2026-10-07 through the reader r.jina.ai.
60. Publishing Perspectives, report on Author Earnings’ May 2016 estimates. https://publishingperspectives.com/?p=89116 . 2016-06-06. Secondary (the Author Earnings report itself was not opened).
61. Aftermath and The Decoder, reports on a New York Times article about AI-written romance novels. https://aftermath.site/ai-romance-novel-new-york-times (2026-02-10) and https://the-decoder.com/human-writers-face-an-impossible-race-against-chatbots-that-finish-a-book-before-lunch/ (2026-02-09). Secondary (the New York Times article refused our tools).
62. Reuters, on AI-written ebooks on Amazon, as syndicated by Kitco. https://www.kitco.com/news/off-the-wire/2023-02-21/chatgpt-launches-boom-ai-written-e-books-amazon . 2023-02-21. Secondary (a syndicated copy).
63. The Decoder, on AI-generated books on Amazon. https://the-decoder.com/ai-generated-books-are-flooding-amazon-and-tanking-sales-for-human-authors/ . 2026-08-15. Secondary for Amazon’s publishing limit.
64. Originality.ai, study of likely AI-written books in Amazon’s “Success” self-help category. https://originality.ai/blog/likely-ai-success-self-help-book-study . 2026-01-28. Read at source on 2026-10-07 through a summarising tool. The publisher sells an AI detector.
65. Publishers Weekly, “Book Output Topped 4 Million in 2025”. https://www.publishersweekly.com/pw/print/20260323/99943-book-output-topped-4-million-in-2025.html . 2026-03-17 online. Read at source on 2026-10-07 through a summarising tool; secondary for Bowker’s count.
66. Alliance of Independent Authors, news podcast on Amazon’s limit of two titles a week. https://selfpublishingadvice.org/podcast-amazon-cuts-kdp-uploads-to-two-titles-a-week/ . 2026-10-02. Secondary for Amazon’s limits and their dates.
67. Amazon Ads, case study of an author-publisher’s Sponsored Products strategy. https://advertising.amazon.com/library/case-studies/authors-sponsored-products-advertising-strategy . 2023-06-26. Read at source on 2026-10-07. Amazon’s marketing material; the publisher’s own figures.
68. Amazon Ads, authors’ guide to Sponsored Products. https://advertising.amazon.com/library/guides/authors-guide-to-sponsored-products . 2023-09-07. Read at source on 2026-10-07. Clicks needed per sale are our estimate.
69. Reedsy, on the cost of self-publishing a book. https://reedsy.com/blog/guide/how-to-self-publish-a-book/cost-to-self-publish-a-book/ . Published 2020-06-23, updated 2026-08-13. Read at source on 2026-10-07. The publisher sells these services.
70. Pricing pages of ChatGPT, Claude and Atticus. https://chatgpt.com/pricing , https://claude.com/pricing and https://www.atticus.io/ . Undated pages. Read at source on 2026-10-07; ChatGPT through the reader r.jina.ai (direct reading refused). The total is our estimate.
71. Nespola, Accelerator programme page. https://nespola.io/accelerator . Undated page. Read at source on 2026-10-07. Books needed are our estimate.
72. Alliance of Independent Authors, announcement of its 2025 author income survey. https://selfpublishingadvice.org/2025-author-income-survey/ . 2025-09-11. Read at source on 2026-10-07.

Corrections

If you are quoted or named on this page and think something is wrong, want your reply shown beside it, or want to be removed, write to [email protected]. We aim to reply within 14 days, and always within one month. Factual errors are corrected with a dated note here. A sentence that is seriously disputed comes down while we check it.

Change log:

• 2026-10-07, before publication: a second round of research with web searches replaced “When it can work” with “The upside” and “What it takes to compete”. It corrected four statements from the first draft: that no data existed on how AI-written books sell (one study of fiction now does); that no AI-book income had been checked against payments (one broker-vetted listing now has); that Amazon’s newest count of high earners was from 2019 (it is from May 2022); and that there was no income evidence for low-content books (three broker records).
• 2026-10-07, before publication: a review against the sources corrected the new sections. Not all eleven sold catalogues were ghostwritten or sold with adverts (seven were outsourced; two ran no paid adverts). Their age at sale is our estimate, not a stated figure. The share of revenue kept is wide (36% to 97%), not thin. “Almost 18%” over $100,000 became the Alliance’s “almost a fifth”. The paper’s caution on catalogue size was restated in its own terms. The Kindle Unlimited rate’s low point is about $0.0040, not $0.0041. Two listings, not one, show 98% kept.
• 2026-10-07: wording about named accounts and companies reviewed; a company’s full reply to a review was added, two accounts with nothing on offer are no longer quoted or linked, statements about what tweets omit now say what we did not view, and two accounts whose offers we did not check are no longer listed as earning from the method; the Nespola review page was read again and its review date and count are now given as read.