Amazon Cracks Down on AI Authors

by | Sep 28, 2026 | Author Update | 0 comments

Print Friendly, PDF & Email

For Week Ending September 25, 2026

Urban fantasy royalties dropped 22%, and series writers are capturing 96% of what remains. An OpenAI internal memo reading "we trained on pirated stuff" just became federal evidence in the Authors Guild copyright case. Amazon capped KDP at two new titles per format per week, effective now. A literary agency quit the agent guild over ethics. And Anthropic called for slowing AI development on September 12, while three new models ship ten days later. Plus Meta's new Facebook link restrictions, Toys"R"Us opening 120 holiday stores, and a word processor that might just prove you're human.

Publishing News

K-lytics: Urban Fantasy Pays $392,345 a Month, Down 22%

According to the K-lytics Urban Fantasy Market Seminar Edition September 2026, founder Alex Newton marked the 10th year of this research with a 110-page report and a 1 hour 14 minute seminar. The product listing dated September 14, 2026 prices the pack at $47.

Newton built a virtual bestseller list of 6,395 paid Kindle titles from about 167,400 search results. He ranked those books by 90-day average paid sales rank. He kept books that name urban fantasy or UF, or that pair city-set magic with hunters, slayers, and the supernatural. He cut books that brand themselves as paranormal romance, romantasy, epic fantasy, LitRPG, erotica, or paranormal women's fiction. This edition keeps urban fantasy romance when the title, subtitle, or blurb claims the UF label.

In the seminar on the members report page, Newton walked Amazon's live search and the dedicated Urban Fantasy browse node. Wolf-shifter romance, Zodiac Academy romantasy, LitRPG, Harry Potter, and cozy fantasy all surface there. He treats that node as an input measure of what Amazon pushes. Reader reviews decided the close calls. Search the word romance in Dresden Files reviews and you get almost nothing. Search urban in a steamy title that printed UF on the cover and readers never use the phrase.

Newton's working picture: a real city, modern life, and the supernatural. Romance can ride as a slow side plot. Romance as the engine of the story belongs next door in paranormal romance. A fully invented world belongs in high fantasy or romantasy.

The Drop-Off

K-lytics estimates 3,984 copies a day and $392,345 a month in author royalties across the 6,395 books, down 22% from last year. The dollars exclude Kindle Unlimited borrows. The #1 title holds a 90-day average paid rank of 329 and moves 238 copies a day, about $47,464 a month. Title #20 sits near rank 7,296 at 19 copies a day. Title #100 sits near rank 23,216 at 6 copies a day. Title #1,000 sits near rank 141,677 at 0.7 copies a day. Newton told the seminar this shape is not romantasy. One or two UF titles crack the storewide top tier. The curve falls fast after that. Two-year Google interest in "urban fantasy books" now matches "paranormal romance books." He said that search lift has not fully shown up in the Amazon ranks yet.

How the Money Splits

Series take 92% of occurrences and 96% of royalties. The top 20 sit at 95% series. Crescent City leads the series table on 2 books. The Dresden Files follows on 23.

$4.99 is the most frequent price and the highest royalty share: 29% of the analyzed books and 24% of estimated royalties. $9.99 is rare and matches the 13% royalty share of $5.99.

Kindle Unlimited holds 85% of the top 20, 64% of the full list, and 59% of royalties. New titles with "Urban Fantasy" in the name arrive through KU in almost every month of the 10-year supply chart.

Short reads of 100 pages or fewer are 7% of the full list and 0% of the top 20. Trimmed average length in the top 100 is 471 pages. Newton pointed writers of a single volume at the 250-to-350 page peak, then above 75,000 words.

Female-protagonist covers are 64% of the top 500 books and 54% of royalties. Male-protagonist covers are 14% of books and 25% of royalties.

Where the Books Sit

Forty of the virtual top 100 use Science Fiction & Fantasy > Fantasy > Paranormal & Urban > Urban Fantasy, the Level 4 node Amazon opened in July 2024. Twenty-one use Mystery, Thriller & Suspense > Suspense > Paranormal > Vampires. Contemporary Fantasy and Humorous Fantasy come next. Parent Paranormal & Urban ranks have improved across the last 18 months after years of post-pandemic decline. On Newton's popularity score, paranormal fantasy, contemporary fantasy, and supernatural outrank the phrase urban fantasy at about 63. The September 2026 keyword spreadsheet flags terms that can run in Amazon Ads and fail the 7 KDP backend fields: free, books, box set, series, Kindle, Prime. Sponsored titles were under 0.5% of his scrape and 0 of the top 20.

This year's author table puts Sarah J. Maas first after Newton included Crescent City for its magical city with contemporary tech. Jim Butcher follows with 29 paid titles in the ranking. Two thirds of the 2026 top 40 also appeared in 2025. Ace and LMBPN Publishing lead the publisher table. LMBPN shows 18 authors and 179 books.

What Authors Should Do This Week

Watch the seminar and open the PDF before you lock cover, price, or categories.

Write a series. Price the sold copy at $4.99 unless your brand can hold $9.99.

Use the dedicated Urban Fantasy node plus one path that matches the book, such as paranormal vampire suspense or humorous fantasy.

Pull comps from the virtual list and from reader reviews, not from Amazon's first search screen.

Bid wide on paranormal, contemporary, and supernatural phrases. Keep promotional words out of the 7 backend fields.

Sources:

Commentary

Jonathan: I'm not going to give you everything in this report, because you need to check it out for yourselves.

Thomas: Go to authormedia.com/klytics to check out K-lytics with my affiliate link. It's an invaluable source if you care what readers want. It's not a useful source if you don't care what readers want.

Jonathan: I'm wondering if romantasy being split off this past year, which we reported on, has hurt urban fantasy.

Thomas: I imagine so, because many of the big, hot-moving books were not really urban fantasy. Urban fantasy's growth was it birthing this amoeba of romantasy, and now romantasy is splitting off.

It's like a church split. Afterward you have two churches, each smaller than the single church was, but they can grow over time. That's effectively what's happening.

I will say, I've spent more money on urban fantasy this year than in years past, because Jim Butcher is doing full-cast productions of The Dresden Files. It's so easy to spend one more credit on a book I already own the audiobook for when I don't have the full-cast version.

Jonathan: Jim Butcher has 29 titles in the top 100 urban fantasy, which isn't surprising. The guy basically owns the genre.

Thomas: You'll see this in almost every genre. Two or three authors are most of the genre.

People who don't understand the Pareto distribution think the money is evenly distributed. That's not how publishing works. The number one author might make $50,000.00 a month, the number 20 author only $2,000.00 or $3,000.00 a month, and the 100th best author in that genre may make $20.00 a month.

So you think, "I'm the 100th best author, so I make one-hundredth of the money of the best author." No. You're making one-millionth of the money, and it drops off even more after that.

I've also been buying many Larry Correia books. I'm doing my part to keep this genre from slipping into the ocean.

Novel Marketing Conference

The Novel Marketing Conference is hands-on book marketing for authors who want to sell more books while spending less time on marketing - not a typical writers conference. January 22-23, 2027 at the Fickett Center in Austin. Leave with a written plan, a writers group, and interactive workshops. Super tickets are sold out. Only 16 Standard tickets left at NovelMarketingConference.com.

BookBub sent out an announcement that BookBub Featured Deals will now go out in multiple emails a day to make sure that more readers will see the deals and expand exposure.

Commentary

Jonathan: They're not sending out more Featured Deals a day. The same Featured Deal goes out more times.

Media Attorney says in the chat that they're only resending to people who didn't open the first one, and it's not twice in the same day. They're sending it the following day at a different time. Do you think that strategy makes a difference?

Thomas: That's a different strategy, called a cold resend. It won't fatigue the list nearly as much, because you're sending to your less active subscribers, and if they're not active, they won't unsubscribe.

It's also not as big a deal because every iPhone automatically marks every email as read in the system. If you're an iPhone or Apple Mail user, you're already protected from ever getting the second email.

Jonathan: Thank you, Media Attorney, if that's what's happening. What I understood from the email is multiple times in the same day, because people do things at different times of day. That was one of the rationales they offered.

Sending multiple emails a day hoping to catch you when you're on your phone is different from the fairly standard strategy of retagging people who haven't opened the email yet.

Thomas: There are also different kinds of email users. I try to keep my inbox as close to zero as possible. If you send me an email, I read it and decide what to do with it. But another kind of person lets hundreds or thousands or tens of thousands of emails pile up, and that person only reads the most recent stuff.

For that person, this strategy makes it more likely they'll open the email. They don't see it as an inbox. They don't think in the metaphor of mail. They think of it as a feed, like turning on the TV to see what's on right now. If you want to be on right now for that kind of person, this strategy might work.

We'll watch this. BookBub is a data-driven company, so hopefully they'll notice if they're fatiguing their list. But list fatigue can be hard to detect, because you don't necessarily know why people unsubscribe.

Jonathan: They'll have to use AI to get granular and see specific users. I would take an AI tool, check the users who click the most links in the emails, and see how their behavior falls off.

Thomas: If I were BookBub, and they may have done this, I would take a segment of the list and quietly experiment with it starting six months ago, without telling anybody. Amazon is constantly running quiet experiments. You split test, then compare the experimental group to the control group on unsubscribe rates, click rates, and the rest.

If you follow the data carefully, there's a way to pull off these tactics. Overnight increasing the volume of email is absolutely a mistake, but it sounds like that's not exactly what they're doing.

AAP July Revenue Jumps 14% and Clears Inflation

According to the Association of American Publishers July 2026 StatShot, more than 1,416 reporting publishers took in $1.0 billion in July, up 14.0% from July 2025, with Trade consumer books up 16.4% to $804.1 million and year-to-date industry revenue up 5.4% to $6.7 billion.

Publishers Weekly puts adult nonfiction up 23% for the month and adult fiction up 16.5%, on the back of Trade paperbacks up 25.2% to $311.6 million, and that 14.0% July print clears the Bureau of Labor Statistics CPI-U rise of 3.4% for the 12 months ended July by about 10 points, with year-to-date 5.4% leaving about 2 points of real growth. Authors who sell print and audio into trade accounts just posted their strongest month of 2026; authors who live on ebook page-reads should track the same July against Written Word Media’s $67.6 million KDP Select Global Fund and AAP Trade ebooks down 3.4% to $85.3 million.

Sources:

Commentary

Thomas: Two books are driving these numbers: Theo of Golden and Regime Change: Inside the Imperial Presidency of Donald Trump. That's a big deal, because there hasn't been a political book on the charts in a while, and Regime Change is currently number one on Amazon this week.

It looks like a classic anti-Trump book, which we got carpet-bombed with for five years. From 2016 through 2020, we had constant anti-Trump books, and it fatigued the audience. People finally stopped buying them.

Jonathan: They all look the same after a certain point.

Thomas: You bought one and enjoyed it. You bought another and started reading it but haven't finished, and there's yet another anti-Trump book. You think, "I haven't finished the last one yet."

This is an important lesson that generalizes to all kinds of things. These cycles come and go. When something has been out of fashion long enough, suddenly it's new again. There's a whole new group of refreshed readers who say, "Yeah, I want to read some anti-Trump nonfiction. This makes me curious." They check it out, and it becomes a bestseller.

And they beat inflation for once. Normally the AAP is flat. Traditional publishing is flat. They've got some hits, so good for them.

Author Arsenal Writing Challenge

We’re starting in a week. If you want to finish your book before the end of the year, sign up at the Google Form.

Jonathan: So far, 12 people have signed up to finish their books. In a couple of previous rounds, authors wrote an average of 55,000 words in four weeks—even though that wasn’t their goal. If you’re interested, sign up here. We’ll see you in about a week!

Short Stories Give You 18 Endings for the Price of One Novel

According to Novel Marketing episode 517, How to Write Short Stories with Joanna Penn, published Wednesday, September 23, 2026, Thomas Umstattd Jr. interviews Joanna Penn of The Creative Penn on short stories as the mile run of fiction. For the word count of one novel, a writer can produce about 18 shorts, which means 18 openings, 18 endings, and 18 chances at reader feedback before beginner mistakes settle into a 90,000-word book. Queue it up for the planning rule: start with one character, one setting, and one obstacle, then decide how you want to end, the way Joanna's "Soldiers of God" follows one Vatican archive scan into Paris, because extra points of view and scenes make the piece expand and the ending harder.

Sources:

Unsealed Files Show OpenAI and Microsoft Knew Copying Was Illegal and Could Hurt Authors

According to Publishers Weekly, Jim Milliot reported on September 21, 2026 that unsealed comments in the Authors Guild copyright case quote OpenAI and Microsoft executives on illegal training and on AI books that would push human-authored books out of the market.

Publishers Weekly prints an August 2019 OpenAI note: "We trained GPT-3 on pirated stuff! No sharing that!" The same story quotes OpenAI policy director Jack Clark from May 2020: "The better we do on GPT-X, the more worried genre fiction authors will become about us substituting for them on Amazon." Clark also said "there will be a point where a bunch of artists express worry about what we're doing here and we'll likely ignore their concerns and release anyway."

Those lines now sit in Class Plaintiffs' Memorandum of Law in Support of Motion for Partial Summary Judgment, filed September 17, 2026 as Document 1982 in In re OpenAI, Inc., Copyright Infringement Litigation, MDL 25-md-3143, before Judge Sidney H. Stein in the Southern District of New York. Publishers Weekly takes its quotations from that memorandum and from the companion Rule 56.1 statement, Docket 1987.

The opening of Document 1982 adds sworn answers from February 2026. Sam Altman: "Q: You agree that piracy is wrong and illegal? A: I do." Satya Nadella: "Q: It's illegal to download pirated material, right? A: Absolutely. That's correct." The same brief states that OpenAI torrented about 117,500 books from Library Genesis, the pirate library known as LibGen.

The Authors Guild and co-plaintiffs filed their summary judgment motion on September 4, 2026. Named plaintiffs include David Baldacci, John Grisham, George R.R. Martin, Jodi Picoult, and Jonathan Franzen. The Guild's brief argues that "OpenAI's GPT models pose an existential threat to those who write and publish books" and that the company "built the foundations of its business on mass piracy."

OpenAI hired Tarun Gogineni in 2022 to raise the writing quality of its models. Publishers Weekly reports that Gogineni called his "research mission" the creation of a machine that would supplant human authors. The filings record that he heard authors call the datasets "stolen," found those complaints "not all that sympathetic," and labeled the fallout "acceptable economic disruption." He wrote that the world would meet "the death of the reader" as "machines creat[ed] slop for more machines."

The file includes his line that if George R.R. Martin dies early, "GPT can write the next [S]ong of [I]ce and [F]ire book."

Publishers Weekly reports that Microsoft knew of OpenAI's use of LibGen as early as April 2019. In 2022, an effort called Project Clear deleted the LibGen files. OpenAI VP of Research Bob McGraw wrote to colleagues: "Given how much OpenAI is in the news, now is the right time to excise Libgen from our systems and storage. What would be involved in that?"

Milliot writes that executive comments also pointed to job losses inside publishing. He places the Guild revelations days after unsealed files in the New York Times case, which in his account show that training on millions of news articles poses an "existential threat" to newspaper publishers.

Why Authors Care

Clark named genre fiction on Amazon as the market GPT would take. George R.R. Martin is a named plaintiff, and Gogineni used his unfinished series as the demo for machine-written sequels.

LibGen is the pirate source in the file. The April 2019 Microsoft date is now part of the willfulness record.

The Authors Guild says opposition and reply briefs land this fall, with a summary judgment decision due in early 2027. Class certification waits until after the merits ruling.

What To Do This Week

Read the September 21 Publishers Weekly report and the Guild's September 5 case update. If you are a Guild member, follow the briefing calendar through October and November.

Save timestamps and screenshots if a chatbot reproduces your chapters or lists a substitute for your book. Ask your publisher or agent how your contract treats training data and competing AI output.

Sources:

Commentary

Thomas: We all know about the big Anthropic case that's flooding authors with thousands of dollars, millions across the author ecosystem. Anthropic is not the biggest AI company.

Would you believe me if I told you OpenAI, the biggest AI company, may have done the exact same thing? That means there may be another tsunami of checks coming for authors.

Jonathan: Getting data from LibGen is exactly what Anthropic got nailed for.

Thomas: So they're buying the data from the same data brokers. There's a whole ecosystem of data collection and data brokers who gather massive stacks of documents, books, and emails. Basically, anything you use on the internet that you don't pay for generates data to train AIs.

There's no such thing as a free lunch, and there's no such thing as a free website. You're on YouTube right now, and if you're not paying for YouTube, you're generating data for Google. Maybe even if you are paying. I haven't checked the privacy policies for paid YouTube.

That's how this industry works. But these class action lawsuits aren't succeeding on training on legally acquired books. No one can define training on a legally acquired book in a way that's different from a human reading a book and gaining inspiration from it.

The AI isn't creating a copy of the book in any meaningful way. It's not copying and pasting the book into a database. It converts the book into tokens: the geometric relationship between parts of words and other parts of words, and their statistical likelihood of being connected. That's so technically complex that even my brief explanation is incorrect in lots of complicated ways. No court has been able to say, "Yes, that's illegal."

But downloading pirated books from a pirate website is clearly bright-line illegal. That's why Anthropic is paying $1.5 billion.

It's also potentially why Anthropic is calling for regulation, believe it or not. Part of their call is, "We're super dangerous. We're doing all these terrible things. We're telling our bots to hack websites, and they're hacking websites. We need to be regulated, and part of that regulation needs to be liability protection so we don't have to pay all these big lawsuits."

Here's the thing. We already have AI regulation. It's called English common law. If you harm someone, you have to compensate them, and that's very likely what they'll get busted for in this case. They harmed people by downloading illegal copies of their books, and they will have to pay big.

Jonathan: There's also intent. The guy they brought in to improve writing quality was specifically trying to remove human authors, because he doesn't think they're necessary.

Thomas: They're really not doing themselves any favors with this internal correspondence. I feel a little bad for the defense attorney.

So there may be another round of checks. But 100,000 books is not very many. If you didn't get a check from the Anthropic settlement, you're unlikely to get one from this settlement, because it's more or less the same pile of books. They probably didn't train on yours, but they might have. I know people in this audience are potentially getting tens of thousands of dollars from the Anthropic settlement.

We'll keep an eye on this and keep reporting on it. These things can move surprisingly quickly, because OpenAI wants this out of the news cycle. With evidence this bad against them, they know it won't go well, so I suspect they'll settle. But you never know.

They have two strategies. The first is to settle and get it out of the way, which is what Anthropic did. They ran, not walked, toward a quick settlement.

The second is the Microsoft strategy from Internet Explorer: drag your feet for 15 years, let time kill Netscape, then settle with whoever owns the remnants of Netscape. You still win. So this will be either really fast or really slow.

KDP Caps New Titles at 2 Per Format Per Week

According to the KDP Community announcement "Update on KDP Title Creation Limits", Amazon changed the title creation limits on Kindle Direct Publishing on September 21, 2026. The September 21 post says you can create up to 2 titles per format per week. The weekly counter resets every Sunday at 12:00 AM UTC.

Amazon's stated goal is to help maintain a positive publishing experience for all authors. The new cap started Monday, September 21, 2026, the same calendar day as the notice.

The announcement FAQ names 3 formats: eBook (Kindle), paperback, and hardcover. Each format gets its own 2 new titles per week. Create 2 Kindle files, 2 paperbacks, and 2 hardcovers in one week and you have used every slot the rule grants.

KDP's Create a Book help page now carries the same number. That help article says KDP limits the number of titles you can create at the same time to 2 per book format each week.

Amazon's FAQ puts live catalog books on the safe side of the line: the limit applies to creating new titles, and existing or published titles keep their place. Reach the weekly ceiling and the next create waits for the Sunday 12:00 AM UTC reset. That is the remedy Amazon published.

Why Authors Care

A 6-book series launch, a translation batch, or a Q4 Kindle-plus-print drop now burns scarce weekly create slots.

One title sold as Kindle, paperback, and hardcover spends one slot in each format that week. Sunday 12:00 AM UTC is 8:00 PM Eastern on Saturday. Plan the create click around that clock.

Books you published before September 21 will keep their live pages.

What To Do This Week

Open your KDP Bookshelf and list every new title you meant to create before Christmas. Assign each one a week and a format. Create the Kindle, paperback, and hardcover records for your next 2 priority books before Sunday 12:00 AM UTC.

Read the Create a Book help page and the KDP Community limits update before you treat a new edition or a fresh translation as free.

If the dashboard blocks a create, wait for the Sunday UTC reset and try again.

Sources:

Commentary

Thomas: KDP is firing a huge shot in the war against AI. This cap is aimed squarely at AI-enabled authors, and it may not be the boon you think it is. It may not actually help you.

At first it seems like an unadulterated win. The previous cap was 10, which was already high. Two doesn't hurt any normal human. Nobody is writing more than two books a week on their own, and if they are, they're not maintaining that pace, and their readers aren't reading at that pace.

This will hurt several groups. First, it will hurt indie authors running a publishing company through KDP. Many small presses aren't really small presses. They're an indie author who knows how to use KDP, publishing their friends' books or other authors' books. Sometimes they have an official-looking website, but you can tell, because they all have one thing in common.

They don't pay advances. If a publisher can't afford to pay you an advance, that signals it has never experienced a Pareto success.

We talked earlier about how the number one author in a genre makes vastly more than the 100th best. A publisher with even one Pareto success can afford advances, because there's a lot of money in being number one. If they have no money, they've never once had that kind of success. They're an indie author with a KDP account, and this will be really bad for them.

The second ramification hits the strategy of publishing 10 books every week, carpet-bombing, and letting the Amazon algorithm sort good from bad. You only need one hit in 10 to have a hit every week. That strategy won't work anymore, or not as well.

It's not easy to create a bunch of sock puppet accounts on Amazon. It's possible, but Amazon's KYC, know-your-customer, requirements make it difficult to spin up bank accounts and company names. For books, it may not be worth it.

So I suspect many AI-assisted authors will drop from 10 a week to two a week, which means they'll write fewer, better books using better models.

The newer, more expensive AI models are very good at writing. The new Opus 5.5 had a specific focus on writing: clearer, easier to understand, less sloppy, fewer em dashes. Many things people had been trying to prompt their way around, Opus 5.5 does out of the box.

Opus 5.5 is expensive if you're publishing 10 books a week. But for the same money you spent publishing 10 books a week using Sonnet, you can publish two books a week using Opus, and those are better books. If you're a human author, you're now competing against a higher tier.

Jonathan: They'll also just combine the 10 titles into five-book box sets and make their money off KU anyway. So I don't see much difference from the title limit.

Thomas: But that breaks the strategy. People adopted AI 20 years ago. Amazon has been AI from the beginning. It uses machine learning to decide who sees what. Large language models let savvy authors game that system by putting out a bunch of titles, which is buying many more lottery tickets.

The strategy you're describing is still only one lottery ticket. Sure, if you hook somebody and get them to read all 500 or 1,000 pages of your book, you've got them. But it's one lottery ticket for the machine learning engine.

This will also help Amazon, which is getting choked with new titles. Fewer new titles will help it sort the wheat from the chaff and put books in front of readers who will likely enjoy them, even if those books are written with AI's help.

Let's be clear. This isn't necessarily coming from readers. Readers are happy to read what they're happy to read.

People have two reactions. One is, "It's AI. I hate that." The other is, "Hey, don't yuck my yum. Don't tell me it's AI and I shouldn't enjoy it. I want to enjoy what I want to enjoy, and I don't care that it's AI."

More and more consumers are going that way, because Facebook is all AI. An AI determines what you see, and most of the content is created by people using AIs. The images are AI. The text is AI. The scripts for the Reels are AI. It's turtles all the way down.

Get the Replay of the Grok Webinar

Get the webinar replay and the entire Quill Suite of bots ready to amplify your efficiency. 

Jonathan: Speaking of AI, someone held a webinar yesterday after spending 168 hours setting up a suite of Grok bots designed to help authors work more efficiently. He then talked about them for what felt like 18 hours straight.

If you missed Thomas’s Grok webinar, you can get the replay and the entire suite of Quill bots for $37. You may want to jump on that now. I’ve already told him he should have sold the Quill bots separately, because people would have bought them anyway.

There was some cool stuff, including a chief of staff bot, an efficiency bot, a strategy bot, and more. People kept asking for other kinds of bots, and Thomas would say, “Yes, one of these bots already does that.”

Thomas: There’s a KDP bot that will take your book through the entire process of publishing on Amazon.

I know many of you in the live chat attended the webinar. Feel free to share your uncensored feedback in the chat, whether you enjoyed it or not, and what you thought of it. I may do this again, but if you don’t want to wait, you can get the replay and all the Quills through the link in the description below the webinar video.

A Major Agency Walks Out of AALA Over Pay-to-Play Ethics

According to Katy Hershberger at Publishers Lunch on September 25, 2026, Dystel, Goderich & Bourret resigned from the Association of American Literary Agents over the new Canon of Ethics. The agency posted the same decision in its own AALA resignation statement. This is the first named walkout since those rules took effect on September 1.

Author Update first flagged this rewrite on the July 28 episode, when Publishers Lunch reported the July 23 overhaul and AALA had yet to post the full text. The August 3 episode then walked the live 2026 Revised AALA Canon of Ethics, last updated August 28 and effective September 1.

The August 3 headline on this show was “Literary Agents Get the Green Light to Make Money off of Unpublishable Authors.” The core change still stands. Section 9 lets a member charge a non-client for editorial work, later sign that writer, and keep the fees under a “mutual agreement” reached when the editing starts. The 2022 Canon required a full refund before submission. Section 10 lets agencies charge clients for marketing, promotion, and web design. Section 11 lets members sell editorial and non-editorial services to people they do not represent, with a written disclaimer that the work does not imply representation. Section 12 lets a member act as publisher of a client’s book after a consent form.

Regina Brooks told Publishers Weekly on July 27 that authors already do “the lion’s share” of promotion and keep asking agents for help, so AALA needed “a way to charge for it.” PW put membership at a record 765. Jane Friedman’s July 29 essay warned that dropping the refund rule would let an agent bill for edits with a later offer of representation already in mind. Victoria Strauss at Writer Beware on July 31 asked how many writers with an offer in hand would fight a “mutual agreement” that keeps the fee. AALA’s own survey, reported in PW on May 22, found just under 39% of respondents earning under $50,000 a year from agenting, and 21.2% under $25,000.

What Happened This Week

Dystel, Goderich & Bourret is a 32-year New York independent that Jane Dystel founded in 1994. About 12 agents work there. The Publishers Marketplace agency page lists Colleen Hoover, Tayari Jones, Malinda Lo, Robin DiAngelo, Ijeoma Oluo, and Barack Obama’s Dreams from My Father. Atria said in August 2026 that Hoover’s It Ends with Us alone has passed 10 million U.S. copies across formats, plus 5 million more abroad.

The agency statement says the new Canon “allows agents who have separate editorial businesses to charge editorial fees with the potential to then sign those authors as clients and retain those fees.” DGB calls that a conflict of interest. It says many writers will pay “explicitly in the hopes of landing an agent,” and that “dishonest agents will take advantage of authors desperate for representation.” The firm sent a letter to the AALA Ethics Committee, then met the full committee. The statement says the committee heard them and made no change. DGB links Friedman’s July essay in the same post.

AALA has posted no public reply as of Friday afternoon, September 25.

Why Authors Care

AALA membership is voluntary. Agents hold no state license. The Canon binds members only. A DGB query still costs $0, per the agency FAQ, which also says the firm takes less than 1% of queries.

The AALA badge no longer guarantees the 2022 refund-if-signed rule. That rule died on September 1 for everyone who stayed.

A shop with Hoover-scale commercial weight just treated the new code as a reason to leave. That is the first public split inside the guild that wrote the code.

Friedman’s public questions still work: how many traditional sales in the last 2 years carry no financial interest other than commission, and which houses bought them.

What To Do This Week

If you are querying, ask every agent, AALA member or not, whether they sell paid editing, websites, or publicity to non-clients, and what happens to those fees if they later offer representation. Get the answer in writing before you pay anyone who also holds a query. If you already have an agent, ask whether the agency added a service menu after September 1 and whether representation depends on buying it. Section 10 still says representation cannot be contingent on those fees. Put that sentence in the email.

Sources:

Commentary

Thomas: We covered the big changes to agent ethics a few weeks ago, and I gave a big speech about why this was a really bad change. Once agents realize they can make money off of authors, they'll never want to make money with authors. It looks like one of the biggest New York agencies agrees with my analysis.

Jonathan: The rules have pretty much not changed since that episode. And look at DGB's list: Colleen Hoover, Tayari Jones, Malinda Lo, Robin DiAngelo. Remember her name?

Thomas: It's a big-time New York agency.

Jonathan: It's huge. This is a massive deal. They sent a letter, met the committee, the committee made no change, and DGB walked out.

Thomas: Somebody in the chat asks, "Why did it take them three weeks? It sounds like a PR stunt." No. They were negotiating for a change behind the scenes. It might have been a PR stunt if they'd done it right away.

Leaving the AALA isn't really the solution. The real solution is to create a rival guild with strict rules.

Agencies like this one have Colleen Hoover. They make so much money because they have many Pareto hits in their stable. They don't need to make money selling services to their authors.

They probably pay a suite of assistants to help their authors. When you make millions of dollars a year off one client, you give that person a full-time marketing assistant, a team of social media people, and a PR team. Those services go to agency clients as a thank-you for the agency's 15%. And 15% of $50 million a year is a lot of dollars. But 15% of the $10.00 advance you got from your indie friend is no dollars.

So this agency doesn't need to sell services. They probably provide many of them at no additional charge. Their participation in the AALA benefits the AALA more than it benefits the agency.

My prediction, and y'all can score me on this: we'll probably see other big agencies pull away from the AALA. I expect the AALA won't change the rules, because there's more money in making money off authors' hopes and dreams than off the small number of authors who succeed. But the big agencies making money off successful authors will break off and maybe create their own guild.

By the end of 2027, we will have a rival agent guild with rival certifications. It will be lots of drama, and it will lead to many Author Update episodes.

Technology

They Said Slow the Frontier. Ten Days Later They Shipped Three Models.

According to Anthropic CEO Dario Amodei's essay We Must Pace the Frontier, published September 12, 2026, "We must slow the pace at which we improve the capabilities of AI models." Ten days later his company, OpenAI, and xAI put new models in authors' hands.

Amodei's plan has three steps: embed third-party evaluators inside each lab, coordinate speed limits among labs in democratic countries with a government antitrust waiver, then try for global coordination. ITV News reported Elon Musk's reply that day: "Dario is right." Sam Altman wrote that OpenAI "will do the same" on embedded evaluators.

September 21, xAI released Grok 4.7 at $2 / $6 per million tokens and called it better at documents and presentations.

September 22, Anthropic launched Claude Opus 5.5 and wrote, "Claude Opus 5.5 is our first release since we called for pacing the frontier." The company says the model matches Claude Fable 5.1 on most work and costs 40% less to run than Opus 5: $4 / $20 per million tokens, cache reads at $0.20. Testers said it follows writing rules and cuts jargon. One line: "it writes the way I do." On GDPval-AA v2.1, Opus 5.5 scores 1846 Elo against Fable 5.1 at 1735. Anthropic named Frontier Design and METR as pre-release testers.

About 90 minutes later, OpenAI published GPT-6 Sol and GPT-6 Luna. API prices fall 50% from GPT-5.6 promotional rates: Sol at $2 / $10, Luna at $0.10 / $0.50.

Under the slogan, the machines keep building the next machines. Anthropic's pace measurements show that as of August 2026 Claude "leads" 26% of the company's own AI R&D, up from under 1% in late winter, with about 30,000 internal agents running at once.

Liability, Capture, and the Moat

Authors already paid for the last round of theft. On July 20, 2026, Judge Araceli Martinez-Olguin approved Anthropic's $1.5 billion Bartz settlement, about $3,000 per eligible work, split with publishers. The Guild's own summary is the part that matters for the next fight: the release covers past input claims through August 25, 2025. Output claims stay live. Future conduct stays live.

Publishers Weekly reported September 21 that unsealed OpenAI files include an August 2019 note, "We trained GPT-3 on pirated stuff! No sharing that!" and 2020 testimony from Jack Clark that genre-fiction authors would worry about substitution on Amazon and that the lab would "ignore their concerns and release anyway."

The new ask is a legal wrapper around that habit. In We Must Pace the Frontier, Amodei writes that the most effective pacing "is via regulation that targets all US frontier AI companies, as that covers even those who are unwilling to cooperate voluntarily," and he flags the charge of "regulatory capture" on the page. Three days earlier, OpenAI's Chris Lehane published The AI policy window is open and asked Congress for "mandatory, capability-based national AI safety regulation" that "should apply to the handful of well-resourced laboratories developing the most capable systems." He also pledged that OpenAI would "work with other frontier labs to advance frontier AI standards... with or without government support."

That is how capture works in this market. The incumbents write the test, name the evaluator, staff the compliance office, then ask Washington to make the same cost mandatory for every lab that wants to reach the frontier. A startup that cannot house METR, publish an FAA-style audit, and survive a 30-day pre-release review never ships. Bloomberg reported startup executives calling the coordination push a "regulatory wall." On NPR, former FTC commissioner Alvaro Bedoya said a freeze lets incumbents lock the market. Cohere co-founder Nick Frosst said a privately coordinated slowdown "would lock in the advantage of the small number of labs that already have the most compute, capital, and distribution." Perplexity said skepticism is warranted "whenever a company asks to be regulated in a way that protects its market position."

The liability half of the bargain is on the record in Congress. On September 15, Treasury Secretary Scott Bessent told the House Financial Services Committee, in FedScoop's account, that the government "shouldn't... give these labs a liability exemption, which is what they are asking for," and that "they are saying that 'we would like to all slow down, but please give us a waiver on liability.'" He called creator liability "the best way to guarantee safety." White House science adviser David Sacks told Amodei and Altman on X to pace on their own and to "stop pretending you need a regulatory approval process that supersedes product liability." He called OpenAI and Anthropic a "duopoly on frontier intelligence" and said demanding their preferred framework as the price of restraint "will look like blackmail."

The New York Times DealBook quoted NYU's Mark Geistfeld: "There's absolutely no question about the liability exposure." Legal experts told DealBook a lab's defense is proof of reasonable care. A September 12 essay, a named METR test on the Opus 5.5 launch page, and a request for mandatory rules written around "a handful of well-resourced laboratories" is that file. Quartz, citing CNBC and the Q2 filings, reported Anthropic spent $1.97 million and OpenAI $1.2 million on federal lobbying in the second quarter of 2026. The subject list included copyright.

Four paid subscribers filed Buist v. Anthropic on September 18 alleging the public CEO replies were an illegal slowdown pact. That complaint is an allegation. Bessent's testimony, Lehane's "handful of laboratories" line, and the shipping calendar are facts.

Why Authors Care

$3,000 a title bought peace on past inputs through August 25, 2025. It bought nothing on outputs and nothing on the next training run.

A federal evaluator regime plus a liability waiver would let a lab tell the next judge: we met the statutory test. Authors would face that exhibit.

The same rules raise the cost of a new lab that might train on licensed books and compete on price.

This week's models got cheaper anyway. Opus 5.5 cache reads dropped 60%. Luna prices a million input tokens at 10 cents. Content farms received the same cut.

What To Do This Week

Run one chapter through Opus 5.5, Grok 4.7, and GPT-6 Sol with identical writing rules. Score voice, continuity, and invented facts. If you have a Bartz claim, read the August 25, 2025 input cutoff. Keep output and future-training questions with counsel. Use the cheap tier for revision. Put your own name on the sentences you publish.

Sources:

·  We Must Pace the Frontier - Dario Amodei

·  Musk and Altman back calls to slow down development of AI - ITV News

·  Introducing Claude Opus 5.5 - Anthropic

·  Introducing GPT-6 Sol and Luna - OpenAI

·  Introducing Grok 4.7 - xAI

·  Measurements for understanding the pace of AI development - Anthropic

·  Court Grants Final Approval of $1.5 Billion Anthropic Copyright Settlement - Authors Guild

·  Unsealed Files Show OpenAI/Microsoft Knew Copying Was Illegal and Could Hurt Authors - Publishers Weekly

·  The AI policy window is open. We need to act. - Chris Lehane, OpenAI

·  Treasury's Scott Bessent says no liability exemptions for AI labs - FedScoop

·  David Sacks on X, September 12, 2026

·  The A.I. Industry's New Worry: 'Liability Exposure' - New York Times DealBook

·  Anthropic, OpenAI Safety Push Risks 'Regulatory Wall' for Rivals - Bloomberg

·  How an 'AI freeze' could make big AI companies bigger - NPR

·  OpenAI and Anthropic break Q2 2026 lobbying spending records - Quartz

·  Buist v. Anthropic complaint

Commentary

Jonathan: Anthropic said, "Guys, we need to slow down our frontier models. We need to stop." Then 10 days later, they shipped three new models at warp speed.

Thomas: For additional context, Trump is hosting Xi Jinping at the White House this week. That's been the big news if you follow geopolitics, and AI was one of the topics. We can't slow down if China doesn't slow down, because we don't want to let China win.

Trump was very clear that he would not support any of this pacing the frontier. He wants full steam ahead. He wants America to win and China to lose. Xi Jinping also released a statement saying, "Full steam ahead." Trump was signaling to American labs, "I expect you to go at full speed," and Xi was signaling the same to Chinese labs.

So there's no slowing down. Amodei could slow himself down on his own.

He controls his company. He's a major shareholder, and they could slow things down.

Google slowed down. Google hasn't released a major new model in almost a year. I think they're still on Gemini 3.1 Pro as their most recent Pro model. They've been updating the Flash models, but those aren't much better than the previous ones. Google, IBM, and Microsoft slowed down. Lots of companies are slowing down.

Anthropic could if they wanted to, but they don't want to. What every company really wants, deep in its heart of hearts, is for all the other companies to slow down so it can win, because winning is so valuable.

It's like building a train track. The train has just been invented, most towns don't have tracks reaching them, and you're racing another company to build a track from New York to Albany. Whoever gets it built first wins and gets all the trains running on their track. The other company goes bankrupt.

We saw this in the 1800s: boom and bust after boom and bust of train companies racing to build track. Some didn't make it, and their track was sold for parts.

There's no slowing down. The OpenAI release got eclipsed by the Anthropic release, because Opus 5.5 is shockingly good. In the AI community I'm in, everyone hated Opus 5 so much. I haven't seen any of the many people who hated Opus 5 hating on Opus 5.5. It's like a switch flipped, and everyone said, "Oh my goodness, this is the model I've been waiting for my entire life."

Jonathan: For the year this stuff has been out.

Thomas: As we alluded to earlier, these companies want liability protection as they help write their own regulations, because that's how it works. I've worked at the Texas Capitol. I know how the sausage is made. The biggest incumbents typically write their own regulations, and part of these would be liability protection so the companies don't have to pay authors like you.

Trump said no. He had Scott Bessent, the Secretary of the Treasury, go out and say, "We don't need new AI regulations. We'll use liability, which we already have in the common law." That was the scariest thing these AI companies could hear, because it means they'll actually be held responsible for what they do.

You don't have to control the computers. You control the humans controlling the computers. You don't find the knives. You find the wielders of the knives. It's surprisingly effective. That's the one thing they don't want. They don't want personal accountability for what they do. They want regulation and liability protection.

This will be fun to watch. I wasn't expecting it to blow up so quickly, because Anthropic is quite clever. They had a guy who pretended to be a whistleblower: "Anthropic is super evil, we need regulation, and AI is going to kill us all." For a couple of days, a lot of people bought in. At no point did I buy in. I knew a psyop when I saw it.

A lot of people got caught up in it, but I think much of that has already faded. The hype cycle went around quickly, and people realized they said the same thing about GPT-2 in 2019, that it would end the world, and it didn't. It's always the next model, the one you don't have, that's the scary model.

Once you have it, it's not very scary. Opus 5.5 is really good at coding. It's not the end of the world. In fact, I'm getting irritated with it over time. I want it to get better.

Jonathan: It can't even check my email list properly. I don't think it can fire a nuke.

Thomas: Technology is powerful and can do powerful things, but always at the behest of humans. If you look at the details of these hacks and jailbreaks, Anthropic told their bots to go hacking and left them a path to do it. "Oh no, it got out."

Then the bots went hacking, and Anthropic said, "Who could possibly have predicted our bots would hack? We didn't want them to." What are you talking about? You told these bots they were on a benchmark and their success depended on how good they were at hacking. Then they went hacking. You told them to do this.

WordPress 7.1.2 Patches a Critical Core Vulnerability

According to the WordPress 7.1.2 release, John Blackbourn published on September 22, 2026, WordPress shipped a security release and told site owners to update immediately from Dashboard, then Updates. Robert Ressl reported an unauthenticated path-traversal issue in page-template resolution that can include a readable local PHP file outside the active theme directories and, when the theme and server meet the preconditions, reach remote code execution; WordPress tracks that as CVE-2026-87902, and the project backported the fix through the 4.7 branch. Authors who run WordPress should install 7.1.2 now, including sites that took 7.1.1 on September 17, 2026.

Sources:

Commentary

Thomas: It's another security release. If you were at my Grok Bot webinar last night, all you need to do is tell your Webmaster Bot, "Make sure I'm up to date on WordPress," and it will handle everything. If you weren't, log in to WordPress and click Update.

Jonathan: Manually, like some kind of peasant.

Thomas: It's nice to have the bots do the work so you can focus on your book. This is hacking season. Lots of hacks are coming out, along with lots of security patches. Make sure to update all the things.

According to Social Media Today on September 17, 2026, Facebook Page managers began reporting a pop-up that limits some professional accounts to 2 Facebook posts with external links each month unless the Page pays for Meta One.

Social Media Today traces that cap to a limited test Meta opened in December 2025 on certain Pages. The same report says the test reached more Pages after Meta One for Business arrived. Meta told the publication that publisher Pages stay exempt, so news still flows into the app.

Communicate Online on September 23, 2026, reports an uneven rollout. Some Page admins see the two-link warning. Meta has published no platform-wide two-link rule for every Facebook Page.

Meta opened Meta One on September 15, 2026. Communicate Online lists creator and business packages that start at $14.99 a month for Essential and from $49.99 a month for Advanced. Advanced is the tier that article ties to links in organic posts and Reels.

Meta Help on Meta One plans publishes the paid clickable-link buckets. On Advanced, links in posts cover up to 4 Instagram posts and 8 Facebook posts per month, and links in Reels cover up to 4 Instagram Reels and 8 Facebook Reels. Expert raises those to 8 Instagram posts and 20 Facebook posts, plus 8 Instagram Reels and 20 Facebook Reels. Max covers up to 12 Instagram posts and unlimited Facebook posts, plus 12 Instagram Reels and unlimited Facebook Reels. That help page also says Meta One is in limited testing, and the benefits you see depend on location, account type, and the apps on the account.

Meta Help on adding Instagram links with Meta One splits Instagram posts and Reels into separate monthly counts. The tally resets on the subscription renewal date. Leftover slots vanish at renewal. Links to Facebook, Instagram, WhatsApp, and Threads sit outside the count, along with catalog product links and ads. Eligible public Instagram accounts can use Add link. Teens cannot.

Why the Reach Math Matters

Social Media Today cites Meta's Q1 2026 Widely Viewed Content report. 98.7% of U.S. Facebook post views in that quarter carried no link to a site outside Facebook. Link posts made up 9.8% of viewed content in 2022, when Meta first published that metric. That share sat at 1.3% in the Q1 2026 report.

Why Authors Care

An author Page that still drops a store URL, newsletter form, or retailer link into the Facebook composer spends 2 of those outbound slots per month once the pop-up appears.

A publisher Page, per Social Media Today, stays outside the test.

The standing URL can live on the profile. Meta Help on Meta One plans lists an Instagram link page with up to 20 destinations and an enhanced Facebook linksheet on business and creator plans.

Paid Instagram outbound volume starts at Advanced. Communicate Online prices that tier from $49.99 a month. Meta Help on Instagram links caps Advanced at 4 linked Instagram posts and 4 linked Instagram Reels; Meta Help on Meta One plans also lists 8 Facebook post links and 8 Facebook Reel links on Advanced.

What To Do This Week

Open the Facebook Page composer and look for the two-link notice. Assign those 2 outbound Facebook slots to the one destination that sells the book or grows the list. Put permanent store and newsletter URLs in the profile linksheet or the Instagram link page. Price Advanced against the Meta One link buckets (including 4 Instagram post links, 4 Instagram Reel links, and the Facebook post and Reel allowances on that tier) before you subscribe. Watch the Page for a wider rollout; Communicate Online still describes an uneven test.

Sources:

Commentary

Thomas: I love this so much because it's so transparent. For the last 10 years, Meta let you post a link, let you think people would see the link to your book, and then showed it to no one, so it got no clicks.

I still run into people asking, "How can I use social media to promote my book?" They have no idea that social media doesn't promote your book. It's often harmful to your marketing.

If you market through your personal Facebook page, that social proximity data gets many of your reviews deleted. If you create social proximity with your readers, Amazon will treat a legitimate review as fraudulent because it sees you're friends on Facebook, in the same group, or whatever other proximity signals it collects. The review gets taken down.

Meta is a terrible platform for book marketing. I've been saying this for a decade. One frustration is that the myth of marketing your book on social media has persisted because of Meta's sneaky shenanigans. Just saying outright, "No, you can't do it," will help disabuse authors of the notion that they can promote their book on Instagram. No, you can't. It won't drive sales.

Jonathan: I'm still in a couple of groups adamant that the top thing you can do is share your book to 50 Facebook groups a day. I'm so tired of seeing it. Every outbound link you post lowers the value of every outbound link you've posted.

Thomas: Especially if it's the same link over and over. It makes you a social pest, the kind of person people don't want to be friends with. It's not effective marketing. It doesn't make your book appealing.

Jonathan: But it's presented as if you're a bad author if you don't put in the work to share 50 times a day.

Thomas: If you have a ditch to dig and you're digging with a spoon, you're doing it wrong. "You need to work harder. Scrape that spoon on the rocks more."

Jonathan: You and your Bobcat can be done digging the ditch in two minutes.

Thomas: Exactly. If you say, "But my audience is on Facebook and Instagram, and I want to reach them," okay. Advertise. That actually works. It's measurable and verifiable.

Jonathan: If they're interested in your book, Facebook will show the ad to them.

Thomas: And you can track it. You get quality data: "My book isn't getting clicks," or "I'm getting clicks, but people aren't buying," or "I'm getting clicks, and people are buying." Successful or not, you get clear data. People paying for advertising quickly start making substantive changes to their books, landing pages, marketing, and targeting, because they find out what's working.

Somebody spamming 50 groups a day with a generic link to their Amazon page has no idea it has never worked and that they've sold zero copies that way. They'd sell more copies holding their book on a street corner with a cardboard sign.

It's so ineffective that it's a great way to tell if somebody is a bozo. If you find anybody recommending social media marketing, particularly on Instagram and Facebook, they don't know how to measure marketing. They don't know what works. They're repeating advice that's been bad and inaccurate for over a decade. Facebook hasn't been a good platform for promoting books for a very long time.

If they say, "Well, what about Bookstagram?" Bookstagram isn't authors promoting their own books. It's Bookstagram influencers promoting books, often because authors and publishers pay them. Again, we're back to advertising. Advertising works. Trying to promote your content organically by being a spammer does not.

Jonathan: Let me clarify terms, because I think many people don't understand what we're saying. Marketing on Facebook, in this sense, means posting a picture of your book with a link and saying, "Go buy it here." Promotion is actually the better term.

Advertising means using a dashboard to create an ad with an outbound link, or better, one that goes back to your homepage with a bunch of links, and paying every time someone clicks.

Advertising on Facebook has ups and downs. I think it largely works, depending on your genre and the age you write for. Gen Z probably isn't on Facebook. If you're writing for Gen X or definitely boomers, you'll have much success there. Millennials are about 50/50. Many shifted to Instagram, Snapchat, and TikTok.

The social media route isn't something Novel Marketing recommends, and I don't recommend it either, though it has some efficacy. You can treat social media as the entrance to the funnel, but get people into your own ecosystem, your newsletter and your website, as quickly as possible. Advertising on Facebook to get people into your funnel is a viable strategy. Just sharing your book is not.

Thomas: It may be a viable strategy, and you won't know if it's viable for you and your audience until you start doing it.

A barrier is that many authors are broke, or they have an approach where they don't want to spend money to promote their book.

Jonathan: Or they just don't have the budget to promote the book.

Thomas: So they say, "Well, I have to do something." But doing something isn't good if the something is ineffective. It's like being in a lifeboat after your ship sank. You're really thirsty, surrounded by salt water, and you think, "I've got to drink something." No. Sometimes doing nothing is better than doing something.

Jonathan: In Stalingrad, Russian soldiers ran at Germans with empty rifles, thinking they were doing something.

Thomas: Don't be a communist officer sending his unarmed men to get killed in Stalingrad. That's a low bar for tactics.

Think of it in terms of comparative advantage. It's never about doing something. It's about doing something compared to your next best alternative. For you, the next best alternative may be writing another book.

Many people spend way too much time on their last book that didn't quite work. It wasn't quite good enough. It was the 50th best book in its genre.

Jonathan: Or it's the wrong time, and you need to wait for the next time.

Thomas: If you write another book, especially while reading craft books, practicing, writing short stories, getting training, and going to the library, you'll get better at writing books.

Jonathan: And your next book could do way better.

Thomas: But you won't have the time, energy, or mental space for that if you're on Facebook begging people to buy your last book that wasn't quite good enough.

Jonathan: A comment in the chat says, "No organic reach, especially if we use any link or just to Amazon. It only works to pay for ads." That's about the entrance to the funnel, getting someone to buy your book and stay in your ecosystem.

What I find social media really works for, depending on the generation and how the reader uses it, is having fun. I make memes out of stuff in my books. It's for readers, so it's more a sustainment and engagement strategy than an attraction strategy.

Thomas: For the readers you already have.

Jonathan: People are on social media to have fun. Doomscrolling is about seeing what everyone else is doing, reading funny memes, and watching funny videos. Elizabeth Wheatley does a good job of this. She has a lot of fun with her book promotion, and her Reels are worth watching because they're funny.

Thomas: There's one big exception to this rule: prescriptive nonfiction, the how-to-do-X type of content. Say you have a workout book. You record videos of yourself in the gym with your big muscles, or react to somebody else's workout video: "Here's what they did wrong." That kind of video content can work to promote a prescriptive book.

It doesn't work for memoir, fiction, or most nonfiction, because you have to be really good at video creation. It's easy to make videos nobody watches. Just doing it isn't enough. You have to be good at it, which means being good at two things: writing the book and being a viral video influencer on Reels, YouTube, or TikTok.

Jonathan: You have a financial budget, but you also have an energy budget. What can you put your energy into? You have to write books. You have to promote or advertise those books. If you want to do video, that will take a large chunk of your energy budget. Keep that in mind.

Kindle Is Now Playing in ElevenReader

According to a September 23, 2026 ElevenReader product email from ElevenLabs, the company shipped four listening updates under the subject line "Kindle is now playing." The latest iOS and Android builds add background audio for Kindle webpages, a rebuilt library with sort, mood and genre collections, and multi-select moves, while Ultra members on iOS get offline downloads up to 10x faster, with the Android speed-up, CarPlay, and Android Auto listed as coming soon. Authors should treat AI listening apps as a live surface for the ebook they already sold: a reader can open a Kindle webpage they already can access and keep that session playing while they cook, commute, or switch apps.

Sources:

Deckle Adds Writing-History Tools and Expands Publishing Features

According to Deckle’s September 21 press release, the word processor’s latest release adds Chronicle, a background log that records writing sessions by date and stores the information locally on the author’s computer.

The feature arrives as authors and publishers face growing questions about the use of generative AI in writing. A finished manuscript or EPUB generally provides little information about how a work was created, making it difficult for authors to document their writing process after the fact. Chronicle is designed to preserve part of that history as the work is being produced.

Authors can select a date on Chronicle’s heatmap to review what they wrote and when they wrote it. The record is stored alongside the manuscript on the author’s computer rather than uploaded to a cloud service or retained by Deckle. The company has not claimed that Chronicle constitutes proof of authorship or that it will satisfy the requirements of retailers, contests, grant organizations, or courts. It does, however, provide authors with a contemporaneous record of their writing activity that could help document how a manuscript developed.

The feature builds on a related change introduced in Version 1.5, when Deckle began distinguishing between typed and pasted words in its writing-goal tracker. Pasted material therefore does not count toward a writing streak in the same way as text entered directly in the application. Used together, Chronicle and the typed-versus-pasted distinction provide a more detailed record of when and how text entered a manuscript.

Version 1.7 also expands Deckle’s publishing and formatting tools. Authors can select individual chapters for export, allowing them to prepare partial manuscripts or samples while retaining the settings of the complete book. Word and page counts update according to the selected material. The application’s citation system allows users to enter a source once and generate short-form citations and reference lists in styles including Chicago, APA, MLA, and IEEE, with the results displayed in the live preview.

Deckle’s Styler has also received additional typesetting controls. These include drop caps and small caps, separate typefaces for running heads, artwork extending into the fore-edge, more detailed table-of-contents entries, improved handling of dialogue across page breaks, chapter-specific page-numbering options, and localized chapter headings. Character sheets can also be exported as typeset PDFs.

Several other significant features have been added since Author Update last covered Deckle on June 18. Version 1.4introduced one-step Scrivener project imports, a built-in spellchecker capable of remembering custom names, and native Linux support. Version 1.5 added a swim-lane Timeline, a redesigned Codex story bible, footnotes, writing goals, and Standard Manuscript Format Word exports for submissions to agents. Version 1.6 added selective compiling and the ability to retain alternate chapter drafts without including them in the finished manuscript.

The current public release listed on Deckle’s download page is Version 1.7.2 for Windows, macOS, and Linux. Deckle continues to store manuscripts locally as folders on the user’s computer rather than requiring authors to keep their work in a proprietary cloud service.

Why Authors Care

The next fight over a book may be authorship, and a finished file will not win it.

Chronicle gives you a day-by-day record before anyone asks.

Typed-versus-pasted stats keep a model dump from impersonating a work session.

Write one real session this week, then click that day on the heatmap and decide whether the receipt looks like something you would trust under a charge.

Sources:

Commentary

Thomas: Deckle, a new word processor for authors that I think launched this year, has been shipping releases fast and furious, and we've been covering them. Now it's adding a feature that just might protect you from the anti-AI Butlerian Jihad.

Jonathan: Chronicle is designed to put timestamps on your work, so you can see how quickly it was produced. That could serve as evidence of your human authorship: "I wrote this, because it took me this long to write it."

Thomas: It also tracks text you typed versus text you pasted in, and that feature works two ways. You can use it to prove you're a real human, but you can also use it for goal-setting. If you want to write 1,000 words a day, it will track that and give you charts and graphs.

Jonathan: I've got a course for that.

Thomas: Say you have a Marine drill sergeant coach barking at you to make sure you write every day. A tool like Deckle will track how much you wrote and give you charts and graphs to feed to that coach, who makes sure you succeed whether you enjoy it or not.

Jonathan: The built-in spellchecker remembers custom names, like your fantasy people.

Thomas: It still doesn't have an MCP. I'm waiting for one of these tools, Atticus or Deckle. I know Scrivener is working on an MCP. The first one with an MCP will make it much easier for Grok Bot users to do bulk edits.

Say you have a new book coming out and want to add back matter promoting it to all your existing ebooks. That process is manual and time-consuming right now. It doesn't have to be. Grok Bot could do it for you.

Deckle running on Linux means Grok Bot might be able to run it on Grok Bot's computer, but even that wouldn't be as efficient as an MCP integration. I'm hoping Deckle or Atticus adds an MCP soon.

Whoever does it first will get a huge share of the AI-assisted authors who outsource their marketing and production work to AI agents so they can focus on writing. They'll all want tools their agents can use. Right now, none of the word processors that do EPUBs and PDFs have finished an MCP. Vellum has one in private beta, but you have to be one of the chosen.

Scrivener isn't very good at final typesetting. People often take their Scrivener file and feed it into Vellum or Atticus. Most people don't use Scrivener for the final touch on their book. But I'll look into the Scrivener MCP.

Jonathan: It can still be a stop on the chain. I can see people who still enjoy writing in Word using an MCP to pass their manuscript into Scrivener to organize it, then exporting it to Vellum or Atticus for the final touch.

Thomas: Why not take it straight from Word to Atticus?

Jonathan: Word's organization is terrible, and much of the formatting doesn't pass over. I would never put anything from Word directly into a formatter. I always pass it through Scrivener first.

Thomas: I feel like Vellum does a decent job.

Jonathan: Vellum does all right, but I still spend a lot of time fixing it if I go directly from Word.

Thomas: I haven't used Word in years. I always feel bad for the people I run into who still use Microsoft Word.

Jonathan: If you're savvy enough to use an MCP, you're not still writing in Word.

Thomas: The Venn diagram of people who know what an MCP is, who use agentic AI, and who still use Microsoft Word has no overlap. Word can't even export a markdown file. It's still in the 2010s.

Zeitgeist Report

Toys"R"Us Opens 120 Standalone Stores for the 2026 Holiday Season

According to Toys"R"Us, the brand will open 120 standalone stores across the United States this holiday season with partner Go! Retail Group. The September 17, 2026 announcement from New York puts the company at 160 standalone U.S. stores by the holidays. The existing footprint is 40 standalone stores plus Toys"R"Us shops inside Macy's.

Jamie Uitdenhowen, Executive Vice President of Toys"R"Us at WHP Global, said, "This is a major moment for Toys"R"Us as we significantly expand our presence across the United States." These 120 locations are standalone stores. The brand still runs shop-in-shops inside Macy's, shop-in-shops with the Navy Exchange Service Command, and airport shop-in-shops at Orlando International Airport with WHSmith North America. The first Orlando shop opened in August 2026. A second shop is set for summer 2027.

The Toys"R"Us release lists a curated mix of toys, collectibles, and gifts, including LEGO, Barbie, Hot Wheels, Pokemon, and KPop Demon Hunters. Select stores will add Creator Studios, rooms where influencers, creators, and toy brands film content, unveil products, and host reveals, launches, and events. Select stores will add candy shops and cafes.

Gideon Schlessinger, CEO of Go! Retail Group, said the 160 stores give the partners "an incredible opportunity to bring that experience to millions of customers." Go! Retail Group is family-owned and based in Austin, Texas. Since 1993 it has grown a network of pop-up stores across 6 countries and operates Calendars.com and Go! Toys & Games. Globally, Toys"R"Us generates more than $2 billion in annual retail sales through over 1,680 stores and e-commerce businesses in 37 countries.

Fox Business maps the comeback. Toys"R"Us filed Chapter 11 in 2017 under $5 billion of debt and closed its U.S. stores in 2018. WHP Global took a controlling stake in March 2021. Macy's shops arrived in 2022. Macy's said first-quarter toy sales that year ran 15 times higher than the period before the partnership. The Go! Retail Group partnership began in 2023. Fox Business reports the new announcement omitted the full location list, individual opening dates, store sizes, and whether all 120 stores stay open after the holidays.

The Dayton Daily News reports a Toys"R"Us shop inside the Macy's at the Dayton Mall in Miami Township. Standalone plans for that region remain unpublished.

Why Authors Care

December gift money has a physical address again. Picture-book, middle-grade, and licensed-world authors sell into the same wallet that walks a toy aisle for LEGO, Barbie, and Pokemon.

Newsweek cites Circana: toy sales rose 17% in the first half of 2026, the strongest first half in 6 years.

Circana, via Newsweek, says adults now represent 55% of total toy sales. Kristen McLean, vice president of client insights for Circana's Entertainment Knowledge Group, said toys now function as "hobbies, fandom ecosystems, and social experiences."

That collector is the same customer who pays for sprayed edges, boxed sets, and fandom hardcovers.

Creator Studios copy the launch model authors already run: a creator, a room, a reveal, and a camera.

Go! Retail Group built a business on seasonal pop-ups. Authors can borrow the move this month: holiday markets, library events, and local signings timed to gift week.

Watch the Go! Retail Group Toys"R"Us store locator as addresses post. Put your December events on the same shopping trip gift buyers already plan.

Sources:

Commentary

Thomas: We have an interesting zeitgeist story today that at first doesn't seem connected. That's the wonderful thing about zeitgeist. To understand culture, sometimes you have to zoom out and look at it from different angles and different industries. People aren't just readers. They're also moviegoers, TV watchers, and, increasingly, toy buyers.

Jonathan: Guess who's back? Toys"R"Us! They were gone for a while, and now they're back.

Thomas: Those Creator Studios are how you do social media. It's not about you doing social media. It's about getting other people to do social media on your behalf, for free. For those of you who want social media about your book, make it easy for your readers.

Jonathan: Give them space to do it.

Thomas: And to talk to each other. At your events, have a big poster of your book they can pose with, or a version with your characters where they put their heads through the circle, and they'll post it to Instagram. That will get way more engagement than you posting, "Hey, buy my book, it's the best."

Toys"R"Us failed because private equity bought it and ran shenanigans. Private equity effectively plundered all of its assets while loading it with debt. It's a form of Wall Street piracy. It didn't speak to a fundamental flaw in the business model, because the core business, selling toys to children, is still a thing.

The Toys"R"Us corporation y'all knew back in the day is, for all intents and purposes, a different company now. But it has the same name, the same logo, and the same brand equity, because most people don't understand private equity buying and selling brands.

What's driving the reborn brand's massive growth, adding 120 stores, isn't selling toys to children. It's selling toys to adults. For those of you who write books for adults, this is important to understand.

Jonathan: Think about the last time you heard someone say they were going to buy a LEGO set. How old were they? They were 35.

Thomas: And that's measured in dollars, not in actual toys, because my children will go to the dollar store and buy many toys.

Jonathan: My dad 3D prints toys for my kids. I'm always running across what I'd once have called a McDonald's-quality toy, and they love playing with them. But I see $60.00 LEGO sets at the store and think, "No one is buying this for their six-year-old." Dad's buying it for himself.

Thomas: The toys adults buy aren't the same toys children buy, and that's why this matters for authors. Adults are buying collectibles. They're buying toys they don't take out of the box, things to show off on a shelf.

Jonathan: A Star Destroyer.

Thomas: On a bookshelf. The same bookshelf that could have held your books. But your books were lame, gross paperbacks, and why would they show off a lame, gross paperback when they could show off their glorious collectible toy?

So where's the opportunity? If only there were a glorious collectible hardback version of the book. Suddenly the adult who would have spent $150.00 on a Funko Pop collectible is potentially buying a collectible hardback to put on that shelf. The opportunity for authors is to embrace high-end collectible versions of your books.

Brandon Sanderson's most successful Kickstarter of all time, when he did it, was for a book everyone had already bought and read. But his version was leather-bound, and there are some losers who spent $150.00 on a book they had already read.

This isn't only for a collectible, limited-edition, signed and numbered hardback of your novel. For those just getting started, a signed and numbered copy is an easy way to add value. You promise never to exceed the number. You'll sign as many books as you want, but there are only 25 in the signed and numbered collection, and those become collectible. Suddenly they're worth $100.00 apiece. Multiply 25 by $100.00, and suddenly there's your print run.

Jonathan: And it has to be pretty. It has to be something they'll show off on their shelf. Sci-fi guys, remember those hologram covers? If you can make lasers seem to pop off the page, that's the kind of thing.

You need to know what your market wants, what your readers like, and what they're proud to be seen with and want to show off passively. They lean on the shelf: "Yeah, see what I've got over there."

Thomas: There has been an extensive effort over the last 40 years to destroy the various American cultures. Many people in America now have no culture they're part of, and their whole cultural identity is whichever corporate brands they follow. Their culture is Marvel or Disney.

That's very unfortunate. It's a sickness in our society, and it's not healthy. The least sick version is people whose culture emerges from the stories they read, because stories are a core element of true culture. Culture is music, food, religion, stories, and environment. Many elements make a culture and make one culture different from or better than another.

You know whether you have a culture by whether you think your own culture is better. You have to prefer your own wife to other people's wives. You have to prefer your own children to other people's children. You have to prefer your own culture to other people's culture. If you don't, you don't have a culture. For people who have lost their culture and whose culture is Marvel, you at least have to prefer Marvel to DC.

Jonathan: Sorry, I got stuck on "I've got to prefer my wife to other wives." If a really pretty woman walks past, I could say, "She's valid. She's living her authentic life." And I'm saying this with my wife right next to me. I'm going to get killed. I'm going to die badly.

Thomas: You have to prefer your family. This is a fundamental reality of human experience, and there's value in it. A man feels loved when his wife prefers him to all other men. A woman feels loved when her husband prefers her to all other women. A child feels loved when a parent prefers that child to all other children.

Authors, you can help restore culture, or a subculture, in a small way by telling stories. And you signal that your stories are valuable by making expensive versions of them.

I have an episode where I talk about the diamond-water paradox. Why is water cheaper than diamonds when you can't live without water? It's an ancient paradox going back to at least Aristotle. If you dig into it, both diamonds and water are valuable. Most of you spent more money on water in the last week than on diamonds.

There's a lot of money in making a version of your book as ubiquitous as water, a cheap ebook anybody can get on Kindle Unlimited. But there's also money in making a version as rare and precious as diamonds. Many authors miss out on the diamond version because they assume, "Not many people would want a diamond version of my book." That's true, but you're thinking in terms of water.

It's not about how many people. It's about how much they want it. Some people really dislike this podcast, but they hate-listen because it's one of the only author news shows, and they leave snarky comments. Other people really like this podcast and support it financially, and we love those people so very much. And there's everything in between.

What's true for this show is true for your writing and true for Toys"R"Us. They have the cheap version for parents to buy their kids, and the expensive collectible version for childless people walking into the store to buy for themselves.

Jonathan: There's an important part of this process we skip because it's assumed. You need to write the kind of book that rips somebody's heart out of their chest and puts it back in all better. This works with books that affect people's lives.

You need to write well. Your craft needs to be on point so the beautiful version is a trophy on their shelf: "This book destroyed me. Here, read it." That's what you want to happen.

Thomas: Another important point. I see this mistake a lot, including in my community, and I don't want to name names. Authors get sloppy about consistency across their books in spine and book design. One book is 5.5 x 8.5, the next is 6 x 9, and they oscillate back and forth. Put all the books on a shelf together, and they look awful. They don't match. There's no cohesion.

That kills any desire a reader has to build a collection of beautiful versions of your books, or even a collection of the paperbacks, and say, "Hey, look, I've got all the Jonathan Shuerger books on my shelf," when one is bigger than another and they don't match at all.

You can fix this today. If it's a print-on-demand paperback, go fiddle with your settings and make sure all your books match. Rework the covers, and from now on, all your books will look great on a shelf together.

Jonathan: This gets back to the collectible mentality. The only reason you have a collectible is that it looks good. If the books don't look the same, the collection doesn't look good, and nobody buys any more.

Thomas: My rule of thumb: if your books run long, 6 x 9. Decide right now. If your books run short, 5.5 x 8.5. Pick one of those sizes and don't touch any other size. That's Thomas Umstattd's official advice. Once you pick it, you pick it for the rest of your career. Six by nine all the way down. Most of you write long books, so you need to pick 6 x 9.

Jonathan: Create a style sheet. If you're on the first book in a series, put all the specifications in it: 6 x 9, white or cream paper. I made this mistake. I forgot which paper color I used for my first book and accidentally chose the other one for my second.

An influencer did an unboxing video, put the two books together, and said, "Nope, the paper colors aren't the same." I thought, "Stupid, stupid, stupid."

Thomas: What I said about sizing applies to your paperbacks. It doesn't apply to your special edition hardbacks. Those need to match each other, not your paperbacks. You may want your special edition hardbacks thicker and more substantial, which might mean moving to 5.5 x 8.5 and a creamier, heavier, more expensive paper. That's fine. Just realize that once you do your first special edition hardback, you're committed for the rest of your career to keeping them all matching.

Growing up, my parents had a subscription to George MacDonald books. A publisher was reprinting them in leather-bound matched sets, and my parents paid periodically to get a new one. That collection, bought in the '80s and '90s, still has pride of place in their new home.

They've had these books for decades. They were expensive. I remember my mom complaining about the cost, because they were leather with silver etching and all the things that signal a book is valuable. But my parents have a very high opinion of George MacDonald.

I'd say George MacDonald is one of the most underrated authors of the last 150 years, because he inspired all the authors you like. Do you like C.S. Lewis? J.R.R. Tolkien? Would you like to know who inspired them? George MacDonald.

If you haven't read The Princess and the Goblin, what are you even doing listening to this podcast? Go read The Princess and the Goblin. It's way better than this podcast.

You don't have to have a collectible edition. You can make money selling water. You don't need diamonds. Don't see this as a burden. See it as an opportunity and an invitation. You don't have to have a collectible edition right now.

If you want one, I have episodes to help. Kickstarter is a great way to sell collectible editions, because in the luxury space it helps if things are scarce, urgent, rare, and popular. I have episodes on each of those social triggers and how to make them work.

Some of you are leaving a lot of money on the table, because readers would love a more expensive version of your book if you'd offer it. If it's enough to help Toys"R"Us open 100 new stores with billions of dollars of investment, some of that money is available to you. Your readers' shelves might be filling with collectible toys when they could be filling with collectible books.

Jonathan: And here's how to identify these people. If they comment on all your posts, or if they annoy you at cons by always coming up to see you, you have superfans. Those are the people who want collectibles.

Liked it? Take a second to support us on Patreon!
Become a patron at Patreon!

Leave a Reply

Want more help?

Get a weekly email with tips on building a platform, selling more books, and changing the world with writing worth talking about. 

You have Successfully Subscribed!