This week, after nearly three years of drafting, structural editing, and production, my hard science fiction novel, RIFT: Gestalt, launched worldwide.

Over the past several months, I took the book through the full modern publishing stack: developmental editing, typesetting, metadata optimization, print-on-demand (POD), Amazon KDP setup, ARC campaigns, algorithmic advertising, and audience discovery in an ocean of millions of competing titles.

For years, I have analyzed technology shifts and platform economics from an investor’s perspective. But this time was different. I was inside the machine as an operator, feeling every micro-friction of the supply chain firsthand.

Shortly after launch, I came across a rapid evidence review analyzing 89 articles on AI and book publishing, published across multiple languages between November 2025 and August 2026. Much of the discussion clustered around rights, licensing, governance, reader trust, workflow adoption, and product announcements. Only ten of the 89 articles offered sustained technical scrutiny. Far less attention was paid to capability elicitation, agent reliability, inference economics, model drift, or what these technologies were doing to the underlying economics of publishing.

The industry is litigating the past while the ground shifts beneath its feet.

AI has already permeated ideation, structural editing, localization, metadata generation, marketing execution, and visual production. Debating whether the adoption rate is 40% or 70% misses the broader reality. When smartphones crossed their adoption threshold, the debate ceased to be about whether humanity “should” adopt mobile computing. Instead, photography, media, retail, payments, transportation, and human communication were rebuilt around it. Publishing is approaching the same kind of structural reset.

The defining question is no longer whether an LLM can write a better novel than a human. The real question is economic:

When AI radically lowers the marginal cost of producing and launching a book, where does value migrate across the publishing ecosystem?

The Supply Shock: Spam Economics on Open Distribution Rails

To understand where value may move, we first need to discard an outdated assumption: that self-publishing is an amateur fringe.

According to Bowker data reported by Publishers Weekly, more than four million ISBN-bearing books were published in the United States in 2025, up 32.5% from the previous year. More than 3.5 million were self-published titles, an increase of 38.7%, compared with 642,242 traditionally published titles. Independent publishing is no longer a peripheral corner of the book economy.

And AI is changing the economics of that supply. A 2026 empirical study of 14,419 self-published genre-fiction books sold on Amazon between 2023 and 2026, matched with daily sales records through June 2026, found a striking divergence: the number of books recording sales in a quarter increased 19.2-fold, while quarterly revenue increased only 8.9-fold.

The researchers also found that books containing substantial detected AI text were capturing a growing share of sales and scarce top-ranking positions despite weaker average commercial performance. Their conclusion is important: generative AI can reshape a creative market through scale rather than quality.

The signal is not that AI books are “winning.” It is that supply is scaling much faster than reader demand. That creates the conditions for what I would call Spam Economics.

A serious author may spend months—or years—developing a book. An automated production system can potentially generate, package, test, and replace titles at a radically different cadence. The economics do not require a bestseller.

Imagine a production system capable of releasing hundreds of lightly reviewed titles. If each title costs very little to generate and only a small percentage captures modest long-tail revenue, the portfolio can still produce a positive expected return. That is the same economic logic that made web spam persistent:

Near-zero marginal production cost changes the payoff structure of failure.

Amazon has already responded to the problem. Its current KDP Content Guidelines require publishers to disclose AI-generated text, images, and translations, while AI-assisted content does not require disclosure. Amazon also states that it may reject or remove content that creates a poor customer experience and that it uses machine learning, automation, and human reviewers to enforce its guidelines.

But governance faces a familiar problem. Enforcement can raise the cost of spam; it does not necessarily eliminate the incentive to produce it as long as expected returns remain positive. The market consequence is a new cost imposed on everyone else.

The Discovery Tax

More titles compete for the same search results, recommendation slots, category rankings, advertising inventory, reviewers, and reader attention.

For serious authors, this changes the economics of publishing in a very practical way. Producing a book may be getting cheaper, but getting that book noticed is not. In a market flooded with new supply, more of the cost simply moves downstream from production into discovery.  

The Great Unbundling of the Publishing Stack

At the other end of the quality spectrum, the same technology is producing an almost opposite effect: it is empowering professional independent authors.

Historically, a traditional publisher offered authors a bundled infrastructure:

Capital + Editing + Design + Production + Metadata + Distribution + Marketing + Rights Management

That bundle was difficult—and often expensive—for an individual author to replicate. AI and digital publishing platforms are now breaking that stack into accessible modules. A professional independent creator can increasingly assemble a sophisticated publishing operation without building a publishing company around it.

The emerging stack looks very different:

• AI leverage for narrative auditing, research, metadata, marketing variants, translation drafts, workflow coordination, and data analysis;

• Specialized human expertise for developmental judgment, nuanced editing, high-concept design, legal review, and other areas where quality depends on context and taste;

• Direct publishing infrastructure through platforms such as KDP, IngramSpark, and direct-to-consumer storefronts;

• Owned audience infrastructure through newsletters, websites, reader communities, and social channels.

The result is not the disappearance of publishers. It is the disappearance of the assumption that all of these functions must come bundled together. The author-publisher negotiation changes with it.

For decades, an aspiring author’s question was:

How do I convince a publisher to choose me?

A professional independent author can increasingly ask a different question:

If I can produce at professional quality, retain substantially more of the economics, maintain direct access to my readers, and iterate faster—what exact problem does a publisher solve for me?

That is not an ideological question. It is a value-chain question. And publishers with strong answers will remain extremely valuable.

The Barbell Economy

The result may not be a simple migration from traditional publishing to self-publishing. A more plausible outcome is a barbell structure in which two very different models gain leverage while standardized execution in the middle loses pricing power.

At one end sit large publishing groups. Their durable advantages are not routine production tasks. They are assets that remain difficult for individuals to replicate: risk capital, global rights infrastructure, retail distribution power, institutional reputation, major media access, and the ability to turn intellectual property into film, television, audio, licensing, and other formats.

In that sense, major publishers may gradually look less like book factories and more like cultural venture funds and IP asset managers. They select a portfolio of creative assets, deploy capital behind a small number of high-conviction bets, and use distribution and institutional leverage to increase the probability of outsized returns.

At the other end sits what I call the Sovereign Creator. These are not hobbyist self-publishers. They are professional creators who own their intellectual property, retain a larger share of the economics, use AI to reduce operational overhead, buy specialized human expertise when it adds value, and build direct relationships with readers. Their advantage is not scale. It is control, speed, intimacy with the audience, and lower organizational overhead.

The same AI infrastructure therefore strengthens two completely different ends of the market:

It allows a large publisher to operate more efficiently—and a single professional creator to operate more institutionally.

What gets squeezed in between is not necessarily the mid-sized publisher. A specialized independent press with strong editorial taste, a loyal niche community, or differentiated distribution may become more competitive with AI because its fixed production costs fall.

The real pressure falls on something else: execution without differentiated leverage.

Routine production, generic marketing, basic formatting, commoditized translation, undifferentiated catalog placement, and other services whose premium historically depended on friction become harder to defend when much of that friction becomes software.

AI does not necessarily destroy publishing. It destroys the premium on standardized execution.

The Platform Layer: Who Captures the Discovery Tax?

There is another player in this restructuring that is easy to overlook: the platform itself.

If AI lowers production costs, supply rises. If supply rises faster than reader attention, organic discovery becomes harder. And when organic discovery becomes harder, creators spend more resources trying to recover visibility through advertising, promotions, optimization, influencers, newsletters, and other acquisition channels.

Who captures the value created by rising discovery friction?

The answer will differ by platform, and it would be premature to claim that marketplaces deliberately tolerate low-quality supply because they profit from the resulting competition for visibility. But the incentive structure deserves attention.

A marketplace can potentially earn from both sides of abundance: from the transactions generated by a larger catalog and from the increasingly competitive infrastructure required to navigate that catalog. This is not unique to books. We have already seen versions of the same mechanism in search, social media, mobile apps, e-commerce, and online advertising.

The more abundant the supply becomes, the more economically valuable the gateway to demand can become.

That means part of the value AI removes from production may therefore not flow to authors or publishers at all. It may migrate upward into the discovery layer.

From the Right to Publish to the Right to Be Chosen

Digital platforms democratized access to publishing. AI is now democratizing access to production capability. Together, they create an extraordinary abundance of content. But abundance does not create more hours in the day. A reader can theoretically access millions of books. They may still read only ten, twenty, or fifty in a year.

The scarce resource is no longer simply access to publication. It is selection. And selection depends increasingly on trust. A book does not cost a reader only $9.99 or $19.99. It may cost eight or ten hours of their life. As content supply expands, anything that reliably reduces the probability of wasting those hours acquires economic value.

That is why scarcity may increasingly migrate toward provenance, curation, trusted distribution, and direct reader relationships. A major publishing imprint can provide that signal. So can a respected independent author. So can a trusted reviewer, a specialized newsletter, a BookTok or BookSky community, a bookstore, or a curator whose recommendations consistently prove worthwhile.

Trust, in this environment, becomes a form of reputational collateral. Every recommendation places some of that collateral at risk, and creators who repeatedly deliver on the promise accumulate it themselves.

Institutional trust is no longer the only form of publishing trust. Reputation is becoming increasingly distributed.

A professional independent author with a loyal direct audience can increasingly function as a small trust institution of their own.

Where the Value Goes

AI reduces the cost of standardized execution, which unbundles the traditional publishing stack. That unbundling empowers professional independent creators—but also enables industrial-scale low-cost content production. Supply expands faster than reader attention, making discovery more difficult and more valuable. Platforms controlling access to demand gain leverage.

At the same time, publishers that possess capital, distribution, rights infrastructure, and institutional reputation retain assets that are difficult to commoditize. Creators who own their audience and accumulate reputational trust gain leverage of their own.

So value does not disappear from publishing. It migrates away from standardized execution and toward risk capital, differentiated intellectual property, distribution power, direct audience ownership, curation, and reputational trust.

That is why the most interesting question facing publishing is no longer whether AI can generate a book. It is who controls the scarce assets once generating and launching one becomes dramatically easier.

The next publishing giants may still include century-old institutions. They may also include AI-native boutique publishers, powerful curators, discovery platforms, and individual creators who operate with the capabilities of small media companies. The structure is still forming. But one principle already looks increasingly clear:

When execution becomes software, value migrates toward whatever cannot be reduced to execution.

AI is not simply changing how books are made.

It is changing what the publishing industry gets paid for.

Sources & Further Reading

Fred Zimmerman, “Copyright Is the Headline; Capability Is the Blind Spot: AI Technology in the Book-Publishing Trade Press, November 2025–August 2026.” arXiv, August 2026.
https://arxiv.org/abs/2608.00964

Tuhin Chakrabarty, Xinyue Liu, Jane C. Ginsburg & Paramveer Dhillon, “Generative AI floods and dilutes the market for books.” arXiv, July 2026.
https://arxiv.org/abs/2607.20349

Jim Milliot, “Book Output Topped Four Million in 2025.” Publishers Weekly, March 17, 2026. Data: Bowker.
https://www.publishersweekly.com/pw/by-topic/industry-news/publisher-news/article/99943-book-output-topped-4-million-in-2025.html

Amazon Kindle Direct Publishing, “Content Guidelines — Artificial Intelligence (AI) Content.”
https://kdp.amazon.com/en_US/help/topic/G200672390Article content coming soon.