In an era when anyone can generate a book, being able to create one worth reading becomes more valuable than ever.
Headlines vs. Hard Data in the Age of Algorithmic Authorship
The publishing industry loves a good crisis narrative. From the “death of the novel” to the “bookstore apocalypse,” our sector has weathered countless predictions of doom, most of which proved far less catastrophic than initially feared. Today’s bogeyman is the supposed “flood of AI-generated books” overwhelming retail platforms, threatening to drown authentic human creativity in a sea of algorithmic mediocrity.
A post here on LinkedIn by Thad McIlroy cited industry folk as being 81% concerned about the “flood of AI-generated books on retail platforms.” Nadim Sadek asked, “Are there any data to calibrate us in understanding how big a phenomenon this actually is?” McIlroy and David Barker agreed the AI tsunami was more noise than substance.
The 2007-12 Tsunami of Crap Scare
For me, a reminder of those heady days of 2007-2012 when self-publishing became a reality, and the “tsunami of crap” became the the bedtime story to scare the shit out of tuxedoed publishing folk. It emerged in the USA when the Kindle store launched there, and swept cross the Atlantic when the Kindle UK store opened in 2010.
Underlying all this, a simple reality: yes there is a flood, and most of the new content being thrown out is not worth the paper it will never be printed on, but there will be gems among them. As a Brit entering publishing in 2010 I recall (now with amusement, then with deep sighs of frustration) how The Bookseller‘s new E-in-C, Philip Jones, too British to bandy about American terms like “tsunami of crap”, dismissed self-published books as “Penny Dreadfuls”.
Yet somehow the publishing industry survived, self-publishing is now mainstream, many self-publishers out-earn traditionally-published authors and top-selling self-publishers are lauded, not laughed at.
To the point, dismissing AI with the trite pret-a-porter BS about how “a machine will never write as good as a human” is on par with the trite pret-a-porter BS about how “no author can write a book without a professional editor and publisher.”
The reality is, AI can already write better than many humans, otherwise humans would not be using AI to tidy up their drafts, or even write their drafts for them. If AI prose was as bad as the Luddite Fringe would have us believe, there would be no need to rant and rant again about how it is a threat to civilisation as we know it.
Rather, the Luddite blustering and posturing and tilting at windmills shows just how strong AI is, that is can instill such blind panic among those in that den of insecurity and self-doubt that is the publishing industry.
Unpacking “unpack”
As the rainy season subsides here and some semblance of regular electricity and internet returns, I’ll unpack more of this.
And yes, “unpack” is one of those words AI tends to throw up a lot, and no doubt I’ll now get inundated with emails from Luddites R Us telling me my use of “unpack” means my secret use of AI has been exposed. So, memo to the Fringe – “unpack” has been in use in the English language since the 1400s and has since evolved to cover many less literal meanings than just to remove something from a pack.
One of the most mind-numbingly stupid ideas merging from the Luddite Fringe this decade is the notion that AI is inventing words that have never been used before in human history and therefore are living proof that a given sentence was artificially created. There are so many mind-numbingly stupid ideas from the Luddite Fringe that it’s difficult to rank just where this one should be positioned, but right near the top, for sure, because it fails (they fail) to understand the most basic thing about AI writing: that it is trained on what humans have already written.
Ditto grammar. The em-dash has ben about since way before nineteen-bow-and-arrow, so it should be no surprise that AI has identified it as a common grammar point, and uses it accordingly. It’s just a matter of time before the Grand Luddite Pooh-Bar looks at a piece of AI writing and sees the word “the” being deployed with alarming regularity and sends out a message to the cellar-dwellers declaring any sentence with the word “the” has been written by AI and the author is a fraud and should be hung, drawn and quartered.
But lets get back to the flood of AI books on digital platforms that has vexed even those who are open to the idea that AI does not eat babies for breakfast, and might actually be a positive force in publishing.
So before we sound the alarm bells or retreat to our editorial fortresses, it’s worth asking a fundamental question: Is this actually the industry-threatening deluge the headlines suggest, or are we witnessing another case of technological moral panic outpacing empirical reality?
The Scale Question: Numbers in Context
The most striking aspect of the AI book debate is how little concrete data exists to support the apocalyptic rhetoric. When we examine the available evidence, a more nuanced picture emerges.
Amazon’s Kindle Direct Publishing platform processes approximately 1.4 million self-published titles annually, generating around $28 billion in worldwide book sales revenue. Against this massive backdrop, the platform’s response to AI-generated content has been notably measured: a limit of three self-published books per author per day, implemented in September 2024.
Savour that date a while. September 2024. That’s a whole year ago. Which is several lifetimes in AI terms, where yesterday’s hot news is out of date before the cyber ink has had a chance to dry.
This is telling. If AI books truly represented an existential threat to Amazon’s $28 billion book business, we might expect more draconian measures. Outright bans, mandatory AI disclosure requirements, or working algorithmic detection systems. Instead, Amazon opted for what amounts to a gentle speed bump, suggesting they view this as a quality control issue rather than a fundamental threat to their business model.
Draft2Digital, another major self-publishing platform, reported that 2024 saw publishing volumes trending roughly 50% higher than usual. Lo! The AI Lord has risen and is flooding D2D with titles full of em-dashes and invented words like unpack and crucial and the! Run for the hills!
But hold on. This increase encompasses all heightened publishing activity, not exclusively AI-generated content. The platform hasn’t provided specific breakdowns of AI versus human authorship, which itself suggests the AI component may not be as dominant as feared.
Historical Precedent: The Piracy Parallel
The AI book phenomenon bears striking similarities to previous technological disruptions that initially terrified the publishing industry. The music industry’s battle with file-sharing services offers particularly relevant lessons.
In the early 2000s, record labels claimed that piracy would destroy music creation and distribution. The rhetoric was apocalyptic: artists would starve, innovation would cease, and the industry would collapse. Yet today’s music industry, while transformed, remains vibrant and profitable. Streaming services ultimately provided a business model that satisfied both creators and consumers, while most illegal downloading was driven more by access and convenience issues than pure price sensitivity.
Similarly, the publishing industry has survived previous “existential threats”: the rise of paperbacks (which would supposedly cheapen literature), television (which would kill reading), video games (ditto), and e-books (which would eliminate physical books). Each time, the industry adapted, evolved, and often discovered new revenue streams and audiences.
The AI book concern follows a familiar pattern: a new technology enables rapid content creation, quality concerns arise, and industry veterans predict catastrophe. But like previous disruptions, this one is more likely to reshape the landscape than destroy it.
Platform Economics and Quality Filtering
Okay, so let’s look at some cold realities and why a million AI-created books with no substance is not a threat to anybody except the Luddite Fringe, whose blood pressure will go through the roof.
Here’s the thing: The economics of platform publishing provide natural barriers to truly damaging AI book proliferation. Amazon’s algorithm, like most retail platforms, prioritises books that generate engagement, sales, and positive reviews. AI-generated books, particularly low-quality ones, typically struggle on these metrics.
Customer reviews, while imperfect, provide a crowdsourced quality filter. Readers quickly identify and punish obviously AI-generated content with poor reviews, reducing the book’s visibility in Amazon’s recommendation systems. This creates a natural selection pressure favoring quality content, regardless of its origin.
Yes, you can buy your way around these hurdles by using paid ads and Bookbub and such, but no-one churning out AI crap is going to pay money out. And note the use of the word “no-one”. Let’s never lose sight of the fact that publishing scams are human-created, not AI-created.
And one more reality check – one we might prefer not to hear, but real it is: The publishing industry has always contained vast quantities of low-quality content. Visit any used bookstore’s clearance section, and you’ll find thousands of forgettable human-authored books that failed to find audiences. The difference with AI-generated content isn’t necessarily quality. It’s volume and speed of production.
And let’s be honest, this is what scares the shit out of us.
I use AI a lot. Not to write TNPS so much (but full disclosure, spellings, subtitles and formatting of this post all will have had an AI assist before you see it), but in my other life as a Third Word teacher AI, never actually in the classroom, but at home with preparation and development, is something I struggle to imagine life without.
Is it scary? Yes. But in a rollercoaster sort of way. Enjoyable fear, not flight-or-fight fear. In education I can harness that fear, tame it, and make it work for me. In publishing, while I haven’t yet gone down that road myself, I have doubt the same applies.
AI holds no fear, bar one, for me. Just endless wonder. Endless opportunity. Endless possibility.
My one AI fear? That I won’t live long enough to see what happens next.
The Real Impact: Noise, Not Displacement
Rather than viewing AI books as a threat to human authors, it’s more accurate to see them as contributing to an already noisy marketplace. The fundamental challenge for any book – AI or human-authored – remains the same: finding and engaging readers in an oversaturated market.
The publishing industry has always been a hits-driven business where the vast majority of titles sell poorly while a small percentage generate most of the revenue. AI-generated books are more likely to join the ranks of unsuccessful titles than to displace successful human-authored works.
Consider the genres where AI books are most prevalent: quick-turnaround non-fiction like diet guides, basic how-to manuals, and formulaic romance. These categories already contained significant amounts of low-quality, hastily produced content. AI may have industrialized the process, but it hasn’t fundamentally changed the consumer experience or market dynamics.
Note to the many, many high-quality romance writers out there. Fear not, I have not changed my view on the fantastic work you guys do. Write till you drop! And embrace the fact that the scam-AI crews are targetting romance – it shows just how popular the genre is.
Quality Concerns: A Self-Regulating Problem
The most legitimate concern about AI-generated books isn’t their existence but their potential to mislead consumers. However, several factors suggest this problem may be largely self-regulating:
Consumer sophistication: Modern readers, particularly digital natives, have developed sophisticated skills for identifying low-quality content. Online shopping has trained consumers to scrutinize reviews, ratings, and sample pages before purchasing.
Platform incentives: Retail platforms have strong financial incentives to maintain customer satisfaction. Books that generate complaints, returns, or negative reviews hurt the platform’s reputation and reduce long-term revenue.
Author branding: Successful authors increasingly build direct relationships with their audiences through social media, newsletters, and websites. This direct connection provides authenticity verification that’s difficult for AI-generated content to replicate.
Industry Adaptation Strategies
Rather than panic, the publishing industry should view AI-generated content as an opportunity to clarify its value propositions. Several strategic responses emerge:
- Emphasis on curation: Traditional publishers can emphasize their role as quality gatekeepers, using AI books as a foil to highlight the value of professional editing, fact-checking, and author development.
- Author platform development: Publishers should invest more heavily in helping authors build authentic, direct relationships with readers. These relationships provide competitive advantages that AI cannot easily replicate.
- Transparency initiatives: The industry could develop voluntary standards for AI disclosure, similar to how the film industry credits CGI work. This approach provides consumer information without stifling innovation.
- Quality differentiation: Rather than competing on volume or speed, human authors and traditional publishers can emphasise qualities that AI struggles with: nuanced argumentation, personal experience, cultural insight, and emotional authenticity.
The Innovation Opportunity
Perhaps most importantly, the publishing industry should recognise AI as a tool rather than just a threat. Progressive publishers are already experimenting with AI for:
- Translation services: Enabling faster, more cost-effective translation of books into multiple languages
- Research assistance: Helping authors quickly gather background information and verify facts
- Marketing copy: Generating initial drafts of book descriptions, promotional materials, and social media content
- Editing support: Providing first-pass grammar and style suggestions to reduce editorial costs
These applications suggest that AI’s greatest impact on publishing may be as a productivity enhancer rather than a replacement for human creativity.
Measured Response Over Moral Panic
The available data suggests that concerns about AI-generated books have been significantly overblown. While the phenomenon is real and deserves attention, it’s neither the industry-ending flood nor the creative apocalypse that headlines suggest.
The publishing industry’s strength has always been its adaptability. We’ve survived and thrived through countless technological disruptions, from the printing press to the internet. Each change initially seemed threatening but ultimately expanded markets and created new opportunities.
The AI book phenomenon fits this pattern. Rather than representing an existential threat, look upon it as another evolutionary pressure that will drive the industry toward greater efficiency, better quality control, and clearer value propositions.
The most productive response is neither panic nor denial, but measured adaptation. By maintaining focus on quality, authenticity, and reader satisfaction – while remaining open to AI’s legitimate applications – the publishing industry can navigate this change as successfully as it has all the others.
After all, in an era when anyone can generate a book, being able to create one worth reading becomes more valuable than ever.
This post first appeared in the TNPS LinkedIn newsletter.
