370 AI Novels, 21 Pen Names: How generative AI is overrunning publishing

One fifth of new self-published books may be at least 25% AI-written, a Stony Brook study finds, as scandals, withdrawn deals and a $1.5bn settlement reshape publishing.
By Sophie Lindqvist
4 min readAt a glance
- One novelist has published 370 AI-assisted books under 21 pen names since 2025, earning high six figures
- A Stony Brook study estimates ~20% of recent self-published books contain at least 25% AI-generated text
- Anthropic paid a $1.5bn settlement in a class action over ~500,000 titles used without consent
- Hachette withdrew Shy Girl in March; agents quit a $2m deal with Jerry Falade in July over AI accusations
- Pangram detection is estimated at 95%+ accuracy with near-0% false positives
By Sophie Lindqvist
One romance novelist has written 370 novels under 21 pen names since 2025 using AI, earning what she calls "high six figures" on Kindle Unlimited. Coral Hart, who began at Mills & Boon before turning to self-publishing in 2012, now teaches courses on producing books with AI. Her output illustrates the economics driving the shift: Kindle Unlimited's algorithm rewards authors who publish at least once a month, and payment comes per page read.
"A bad writer or a bad storyteller is not going to get a good book out of AI," Hart said. "You still need to understand craft. It's not taking the human out of the equation. It's taking the friction and the time and the business blocks out of the way."
How much of the market is already machine-written?
A study awaiting peer review, led by Tuhin Chakrabarty of Stony Brook University, estimates around one fifth of recently self-published books contain at least 25% AI-generated text. That figure, if confirmed, would mean roughly two of every ten new self-published titles are substantially machine-produced.
The training data is contested ground. A class action involving around 500,000 titles won a $1.5 billion settlement from Anthropic last year for alleged unconsented use of writers' work. Similar lawsuits are ongoing. Seventy prominent authors signed an open letter demanding the big five publishers — Penguin Random House, Hachette, HarperCollins, Simon & Schuster and Macmillan — pledge to "never release books that were created by a machine." None did.
What happened in the scandals?
One literary agent described mainstream publishing's atmosphere on AI as "the wild west." The record supports that framing.
- In March, Hachette withdrew Mia Ballard's romance Shy Girl over evidence of AI use. Ballard told the New York Times an editor she hired may have inserted the AI text.
- Commonwealth short story prize winner Jamir Nazir was accused after readers flagged odd phrases and ran his story through Pangram. He denies it; Pangram later concluded four regional winners since 2025 likely used AI.
- In July, agents withdrew from debut novelist Jerry Falade's $2m (£1.5m) deal after AI accusations. He denied using AI beyond research, and has since signed with a new agent.
- Last month, Thélyson Orélien's novel was struck off the Goncourt longlist after social media users ran it through detection software. He denied AI use but admitted plagiarism in earlier writing; his publisher Boréal dropped him in late October.
Every writer at the centre of these scandals has been an author of colour. Falade wrote that non-native and "culturally unfamiliar" English draws suspicion, adding: "apparently that level of scrutiny is reserved for a Black boy from Nigeria." By contrast, white author Steven Rosenbaum's book, which contained fabricated AI-inserted quotes he disclosed, was not withdrawn in May.
Can detectors settle it?
Academic research estimates Pangram is at least 95% accurate with a near-0% false positive rate, implying genuine human writing is almost never mislabelled. Yet other evidence suggests non-native and neurodivergent writers more often raise false suspicion. Hannah Bowman, who chairs the Association of American Literary Agents' AI special committee, said agents disagree on routine detection versus trust-based relationships. "I worry for authors when they don't have a chance to make a case in defence of their work," she said.
Agents now advise writers to avoid PDFs and save multiple drafts as proof of process. The anxiety produces its own artefacts: one agent received a manuscript from a long-time client full of "bizarre" wording — not from AI use, but from panicked attempts to sound non-AI.
Why won't publishers set rules?
Thad McIllroy, author of The AI Book Revolution, says publishers must be "brave" and publish AI policies. Hachette edited Shy Girl considerably after acquisition, he notes, so internal AI concerns going unraised strains belief. Agent Tess Callero of Europa Content suspects publishers want to keep the AI door open: "Is it because they see a future in which potentially these books could be monetised?"
US copyright law adds pressure: books must be "human authored," and AI-generated text counts as public domain. "If it's AI, they have no protection, they can't have exclusive rights," said agent Natalie Lakosil, who expects the principle to face court challenges but doubts the human-authored requirement will change. "AI is not an artist," she said.
Stephen Marche's 2023 AI novella Death of an Author could only be released as an audiobook over copyright concerns. He argues for transparency as the "very clear moral line." A second Chakrabarty study, also unreviewed, found MFA students in blind tests tended to prefer fine-tuned AI passages over those written by their MFA colleagues. France launched a prize for AI fiction, the Prix Pascal de Littérature IA, last month.
Meanwhile, Hart advises authors not to disclose AI use — disclosure, she says, brings death threats, as she experienced after speaking to the New York Times. "People in publishing are delusional in thinking that it's not going to overtake our industry," she said. With publishers still unpublished-policied, detection imperfect and output scaling fast, the industry's reckoning is arriving faster than its rules.
Source: theguardian.com
Sophie Lindqvist
Show full bio
Correspondent covering people, launches and day-to-day practice in Social platform news and policy at Reach Brief.
Related articles
Net Influencer Maps AI's Role Across the Creator Economy for 2026
High earners face more AI pressure and heavier output demands
Cloudflare Says AI Bots Scrape Thousands of Pages Per Reader Sent
CreatorIQ's 7th Annual Study: Record ROI for Creator Marketing
Creators Are Softening on AI — Again, Digiday Reports