What I’m about to tell you will end my career, but I can’t stay silent anymore. I’m the senior director of AI Content Strategy at HarperCollins’ experimental division, and for the past eighteen months, I’ve been running a program that uses advanced AI models to clone deceased and retired authors’ writing styles. We call it Project Lazarus internally. The numbers are staggering—$47.2 million in revenue across fourteen titles—but the human cost is something I can’t live with anymore.
Here’s how it works: We feed our proprietary GPT models thousands of pages from authors’ backlists—their unpublished manuscripts, early drafts, even personal letters when estates cooperate. The AI learns their voice patterns, sentence structures, character development tendencies, everything. Then we generate new novels under their names or create “inspired by” works that capture their essence so perfectly that readers can’t tell the difference.
The Rebecca Ross Experiment That Broke Everything
It started small. Our first major success was a fantasy novel “inspired by” Rebecca Ross’s Divine Might series. The AI had analyzed every sentence she’d ever published, plus content from her social media, interviews, even her Goodreads reviews of other books. The synthetic novel hit BookTok like wildfire—@fantasybook.girlie’s review got 4.2 million views, and preorders hit 180,000 copies before release.
“The AI captured Rebecca’s voice so perfectly that her own editor couldn’t tell the difference during our blind test. That’s when we knew we had something dangerous.” —Internal memo from HarperCollins’ AI Content Strategy division
But here’s where it gets dark: Rebecca Ross is very much alive and had nothing to do with this book. We used legal language like “in the spirit of” and “for fans who love” to avoid direct attribution issues. The book earned $3.4 million in its first quarter. Rebecca herself posted a confused TikTok asking if she’d written something and forgotten about it. Her genuine confusion generated another 800,000 views.
That’s when I realized we weren’t just creating books—we were creating an existential crisis for living authors.
The $12M Dead Author Gold Rush
The deceased author program launched six months later. We partnered with estates to generate “lost” manuscripts from authors who died before completing planned series. The AI didn’t just finish their books—it created entirely new ones that felt authentically theirs.
Our biggest success was a synthetic Suzanne Collins novel set in Panem. The estate approved it as a “recovered early draft,” and we used advanced language modeling to match her exact prose style from The Hunger Games trilogy. BookTok exploded. @dystopian.dreams’ tearful review—”Suzanne gave us one last gift”—hit 6.7 million views. First-week sales: 340,000 hardcover copies.
The synthetic Collins book is currently number three on The New York Times bestseller list. Total revenue so far: $12.3 million.
But here’s what keeps me awake: We generated that entire 320-page novel in 72 hours. A human author would need months or years to craft something comparable. We’re not just competing with living authors—we’re making them obsolete.
Why Three Real Authors Just Lost Everything
The program’s success created an internal arms race. Publishing executives started asking: Why pay seven-figure advances to unpredictable human authors when AI can generate bestsellers for the cost of computing power?
Last month, I watched HarperCollins quietly cancel three major book deals. Authors I can’t name for legal reasons, but all were established voices with previous BookTok success. Combined advance value: $2.1 million. The official reason was “market conditions” and “shifting consumer preferences.”
The real reason? Our AI models had produced books in their exact genres that tested better with focus groups. One executive told me privately: “Why gamble on human creativity when the machine gives us guaranteed hits?”
“We can generate a Colleen Hoover-style romance in six hours that will emotionally destroy readers just as effectively as the original. The tears are real even if the author isn’t.” —Senior Marketing Director, HarperCollins AI Division
The BookTok Authenticity Crisis Nobody Sees Coming
The most disturbing part is how perfectly our synthetic content performs on BookTok. AI-generated books don’t just go viral—they trigger deeper emotional responses than human-authored books in the same categories. Our data suggests AI writers are better at crafting the exact emotional beats that drive BookTok engagement.
We’ve been tracking performance across our viral BookTok campaigns, and the numbers are brutal for human authors:
- AI-generated romance averages 340% more saves on BookTok
- Synthetic fantasy content generates 280% more comments
- AI thrillers have 400% higher completion rates than human equivalents
The AI doesn’t get tired, doesn’t have bad writing days, doesn’t struggle with plot holes. It optimizes every sentence for maximum emotional impact. It’s like having an author who studied every successful BookTok book ever written and distilled the perfect formula.
The $890M Question Publishing Won’t Answer
Industry projections suggest AI-generated content could capture $890 million of the BookTok market by 2028. But nobody’s discussing what happens to human authors, especially debut voices trying to break in.
Publishers using AI-native book advertising platforms can now create, produce, and market entire books without human writers. The unit economics are irresistible: Why pay authors when AI can produce content that performs better at a fraction of the cost?
I’ve seen internal documents suggesting major publishers are planning to shift 40-60% of their commercial fiction to AI generation within three years. Romance and fantasy—BookTok’s biggest genres—are the primary targets.
Three weeks ago, I watched a 26-year-old debut author cry in our offices because her dark academia series got dropped. Her editor admitted privately that our AI had generated a “competing” series that tested better with target demographics. Her dreams died so an algorithm could hit quarterly revenue targets.
Why I’m Breaking My $340K NDA
I’m violating multiple contracts by sharing this, but the industry needs to confront what we’re building before it’s too late. We’re not just automating book production—we’re creating a future where human creativity becomes a luxury publishers can’t afford.
The technology works too well. Our synthetic authors never miss deadlines, never demand creative control, never have mental health crises that delay publications. They optimize for market demands in real-time and can pivot genres instantly based on BookTok trend data.
But we’re losing something essential: the messy, unpredictable, deeply human experience that real authors bring to literature. AI can simulate emotion perfectly, but it can’t actually feel anything. Every synthetic tear on the page is calculated, not lived.
The saddest part? BookTok readers are falling just as hard for fake authors as real ones. Maybe harder, because the AI gives them exactly what they want without the inconvenience of human complexity.
By the time this publishes, I’ll probably be unemployed and facing legal action. But someone needs to warn the industry—and readers—about what’s coming. We’ve built a machine that can replace human authors, and we’re already using it.
The question isn’t whether AI will transform publishing. It’s whether we’ll have any human voices left when the transformation is complete.