Amazon Is Full of AI-Written Novels That Don’t Make Sense

Now that artificial intelligence can spit out nearly anything you want, people are desperate to use the technology to produce content they can monetize. Books are no exception. Earlier this week, Amazon’s Best Sellers list was rife with fiction clearly written by the likes of ChatGPT. Amazon has since removed the titles from the list. However, the so-called books are still available for purchase, raising questions about AI’s effects on retail transparency and intellectual property.

Authors and customers were shocked Monday morning to find that out of 100 books on Amazon’s “Teen & Young Adult Contemporary Romance eBooks” Best Sellers list, only 19 were “legitimate” books written by actual human authors. AI produced the other 81 books. How could people tell? The differences were evident: The text made no sense, the manuscripts were messily formatted, and the covers had been slapped together, usually with a single background graphic and an unstylized title. The listings were also suspicious, with zero information (like synopses, review quotes, highlights, or graphics) to help readers decide whether to dig in or pass. At most, they featured two or three random sentences in the description section of the listing. 

Amazon’s Kindle Unlimited Best Sellers list had also gone awry. “This brother-in-law is really too ambitious, this person is the master of the Shadow Demon Palace, any sect master has an extremely terrifying strength, If not, how can I sit on the position of the head of the sect in the eyes of everyone’s submission,” one such “best-selling” book, Fanatical Leader, says on page one. The author’s name is simply Nhu. 

Credit: Amazon's Best Sellers list as of Monday, June 26.

Amazon removed the offending books from its Best Sellers lists sometime between Tuesday afternoon and Wednesday morning. Still, the alarming lists have writers and readers concerned that books written by AI could quickly cause real authors (and their labors of love) to suffer.

Clare Pollard, a novelist and poet whose books appear to have been published by small presses such as Fig Tree and Bloodaxe, tweeted Wednesday that what happened this week will “break” Amazon self-publishing. “This is very interesting for those concerned about AI novels,” she wrote. “There seems to be something going on whereby bots read books by bots to game the charts.” 

Of course, AI chatbots like ChatGPT and Bard aren’t producing novels alone. Humans (likely those unwilling to put in the time or effort necessary to write a book themselves) are telling them to do so, taking the resulting butchered text and self-publishing it online. This raises important questions about AI and self-publishing that Amazon and other major book retailers will probably face in the coming months. Should retailers implement quality checks before a self-published text goes live? While it’s unlikely that retailers will prohibit the publishing or sale of AI-produced works, should they mandate labels that clarify which works weren’t written by a human? If people begin publishing AI-produced non-fiction, should there be potential misinformation disclaimers? And how should recognition—like that received by landing on a Best Sellers list—be addressed?