Should Meta Become a Publisher?

The question

The idea is simple. Meta’s AI tools can already help someone draft a manuscript, generate illustrations, and design a cover. Meta also owns the largest audience network on earth — Facebook, Instagram, and Threads.

A publishing division would connect those two halves: create the book inside Meta AI, publish it through a Meta imprint (an imprint is a brand name a publisher uses for a line of books), print copies one at a time as orders come in — a method called print-on-demand, which avoids warehouses and unsold inventory — and sell it to the creator’s existing followers.

We asked five AI systems — Claude, Perplexity, ChatGPT, Gemini, and Meta AI itself — to analyze whether Meta should do this. Each was given the same neutral prompt, framed as platform-strategy research, with no indication that the person asking had any stake in the answer. Each was asked for the case for, the case against, the historical precedents, the likely shape of the business, and a verdict.

The market is real

Nobody disputed the opportunity itself. Self-publishing has become a major channel:

And every AI identified the same genuine gap: for independent authors, writing the book is no longer the hard part. Being discovered is. Meta owns discovery — billions of people, precise ad targeting, and direct relationships between creators and their audiences. That is the one asset no existing publisher can match.

The verdict: five to zero

Five systems, five different makers, one conclusion. Even Meta’s own AI — the one with every reason to cheer for its parent company — advised Meta not to do it. Three walls explain why.

Wall OneThe math is too small

Meta reported about $201 billion in revenue for 2025, almost all of it advertising. Against that scale, publishing barely registers. Run the generous case: suppose a Meta publishing service captured a tenth of the entire self-publishing market and kept a healthy platform share of it.

A business can be real, growing, and profitable — and still be too small for the company being asked to run it. Every hour of executive attention, engineering, and legal review spent on books is an hour not spent on the advertising and AI infrastructure businesses that are hundreds of times larger.

Wall TwoThe legal timing is terrible

Meta is currently being sued over books — by the book industry. In May 2026, five major publishers (Elsevier, Cengage, Hachette, Macmillan, and McGraw Hill) joined author Scott Turow in a class-action lawsuit alleging that Meta pirated millions of works to train its AI. French publishers sued in Paris in March 2025.

Meta did win an important ruling in June 2025 in a case called Kadrey v. Meta, but the win was narrow: the judge found for Meta largely because those particular authors had failed to build the argument that AI training harms the market for their books — and signaled that argument could succeed if made properly.

That is where publishing becomes dangerous for Meta in a way it wouldn’t be for anyone else. If Meta starts competing directly with publishers — selling books to the same readers — it hands its opponents the very market-harm evidence the Kadrey judge found missing.

Wall ThreeThe flood, and the follow-through

Open the door to AI-assisted publishing at Meta’s scale and the predictable result is a flood of low-effort, repetitive, or misleading books — the “AI slop” problem. Amazon, which runs the largest self-publishing platform in the world, has already been forced into defensive measures: mandatory disclosure of AI-generated content since September 2023, and a cap of three new titles per author per day. Meta would inherit the same moderation, rights-dispute, and refund workload from day one, attached to its own name.

Meta also has a track record here, and it is not encouraging. Its newsletter platform, Bulletin, was shut down. Facebook’s Instant Articles — a publishing partnership with major news organizations — failed and was discontinued. Its podcast push lasted barely a year. The pattern the AIs kept citing: Meta builds creator products enthusiastically and abandons them when they don’t strengthen the core business.

The experiment that already failed

The strongest evidence in this entire analysis is not an opinion. It is a fact: a social-media giant already tried this.

ByteDance, the parent company of TikTok, launched a publishing imprint called 8th Note Press in 2023. Its logic was identical to the Meta idea. TikTok’s BookTok community is the most powerful book-discovery force in the industry — authors with large BookTok followings sold 41 million print books in 2024, up 23% in a year. If any platform could convert discovery into a publishing business, it was this one. 8th Note acquired more than 30 novels, signed a print-distribution partnership, and launched its first physical books in early 2025.

Within about two years, ByteDance shut it down. Authors’ rights were returned; the imprint’s online presence was quietly deleted. Reporting on the closure points to a consistent cause: the platform could generate attention, but it could not do the unglamorous work of publishing — editing, author support, long-term career care.

“Distribution is not the same as demand.” — ChatGPT, the sharpest sentence of the panel

8th Note suggests a harder version: distribution is not the same as publishing, either. And ByteDance attempted it with no AI-training lawsuits, no AI-content stigma, and the best book-discovery engine on earth. Meta would be attempting it with all three disadvantages.

The other experiment — the one that lasted sixteen years

There is a second real-world test of this idea, older and quieter — and it came from Apple.

In 2002, Apple added a feature to its iPhoto software: arrange your own photographs, press a button, and Apple would deliver a professionally bound hardcover book, with the printing contracted out to outside printers. The service later grew to include albums and calendars, and the people who used it loved the quality.

Apple never became a publisher. There was no imprint, no editorial staff, no bookstore — just a button inside the software where the photos already lived, and fulfillment handled by others. And it was never a business worth mentioning: Apple ran it for sixteen years without its revenue ever becoming significant enough to report, and the service never even made the jump from the Mac to the iPhone, where everyone’s photos had moved. In 2018, Apple shut it down and pointed customers to third-party printing apps such as Shutterfly and Mimeo. Its parting move was to replace its own service with an interface that lets outside printers plug directly into the Photos app. In the end, even the button was outsourced.

The reason it lasted sixteen years is the same reason it never mattered: it skipped everything hard about publishing. The books were personal keepsakes made from the buyer’s own photographs — nothing that had to be sold to strangers, no rights to dispute, no audience to find.

Set the two experiments side by side and they frame the entire question. The platform that tried to be a publisher was gone in about two years. The platform that offered only a button lasted sixteen — as a beloved feature that earned nothing worth reporting.

The door all five left open

Not one AI said the underlying idea was worthless. All five converged on the same smaller shape:

• Print-on-demand only, fulfilled by partners. No printing presses, no inventory, no warehouses. This model is already being proven in public: in February 2026, Spotify began selling physical books inside its app through Bookshop.org, which handles pricing, inventory, and shipping while Spotify keeps an affiliate fee of about 10%. Spotify entered books without becoming a publisher.
• An export, not an imprint. The most-repeated proposal — from Claude, Gemini, and Meta AI alike, in nearly the same words — is a “send to print” or export feature inside the AI tools: finish your work in Meta AI, send it to print-on-demand or to existing stores like Amazon’s, with the AI-use disclosure attached automatically. Gemini’s version adds the twist that fits Meta best: authors already spend heavily advertising their books on Instagram and Facebook. Meta could serve the publishing world profitably by owning the promotion layer, without publishing anything.
• Honest labels as the price of entry. Where the AIs went beyond Amazon’s standard, they agreed Meta should too: AI-generated versus AI-assisted disclosure shown to readers before purchase, not just reported privately to the platform. That direction is already becoming law — the European Union’s AI Act transparency rules, including labeling of AI-generated content, became applicable on August 2, 2026.

Too early, or never?

The fairest reading of all five analyses, plus the 8th Note evidence, is that timing explains part of the verdict — but not all of it.

So the prediction this research supports is modest and specific: if Meta ever enters publishing, it will not announce a publishing division. One day there will simply be a button inside Meta AI — make this a book — with printing and storefronts handled quietly behind it, honest labels attached, and Meta earning its money from the tools and the promotion rather than the books. It will look obvious in hindsight. The five AIs, asked blind, all pointed at that same button. None of them pointed at a publishing house.