March 15, 2024
As we feared—and warned—the growing access to AI is driving a new surge of low-quality sham “books” on Amazon. In the last few weeks, we have seen hundreds of examples of how bad actors are using generative AI to produce “books” that deceive customers and drive sales away from legitimate books. It appears that every new anticipated high-profile book has one or more scam books up within a couple days of going on sale. For its part, Amazon has been quick to take down scam books when it receives complaints, but the fact that they are able to get through Amazon’s content filters in the first place suggests that detecting AI-aided scams is presenting a challenge.
Previously, scammers had to at least pay human writers (often based in foreign countries) to produce these low-quality “books” meant to steal sales. Now they can churn out the content instantly at virtually no cost using AI. And while Amazon has implemented measures to clamp down on low-quality and scam books on its services, the speed of creating these books and the time between when a book is uploaded, detected, and taken down is giving bad actors an edge—at least for now. Ultimately, every new book is at risk of having several AI-generated biographies, copycat books, summaries, or workbooks meant to divert sales posted right alongside the book.
Based on the complaints we have received, summaries, workbooks, and guides appear to be the category most commonly exploited by the scammers. Low-quality works like these have long been a problem on Amazon because it is easy to piggyback on the success of legitimate titles and sell “companion” books. Fair use case law has found that “companion” books that have a great deal of analysis and commentary and do not take too much from the original work are fair use, so Amazon allows many of them to remain on the site. A few years ago, the Authors Guild, recognizing the impact these books can have on legitimate sales, convinced Amazon to require sellers of summary books to include a conspicuous disclaimer on the listing page and cover, disclosing that the book is a summary or guide and not a substitute for the original work.
Enter AI. As we expected, listings in this category are exploding, with summaries and guides appearing alongside new release books almost immediately. Recent high profile examples of this include NBC News anchor Savannah Guthrie and bestselling author Lisa Sun, but we have also heard from many members with new books out. Authors Guild Foundation president Marie Arana discovered so-called summaries of her latest book, LatinoLand: A Portrait of America’s Largest and Least Understood Minority, on Amazon the day after its release, with the scam summary books appearing immediately after the listing for the book that Arana spent years researching and writing. Roxana Robinson, former president of the Authors Guild, also found “summaries” of her latest novel Leaving, published on February 13, on Amazon almost immediately after publication. Guild member and legal scholar Michael Graetz’s latest book, The Power to Destroy: How the Antitax Movement Hijacked America, also released on February 13, had similar scam “summaries” appear soon after release.
These AI-generated scam companion books rarely rise to the level of fair use, as they have little to no original analysis or commentary and are meant only to confuse consumers and skim sales off of the real books. Instead, the scam companion books simply regurgitate the key points of the original work in a condensed form, which is clearly infringing.
We are also starting to see scammers use AI to produce unauthorized “biographies” of authors that are simply AI-generated rehashings of their lives, often based on autobiographical works. In a recent Substack post, best-selling author and Authors Guild member Stephanie Land describes finding several “biographies” of her that were lifted from her books. In one instance, Land acquired a 180-page biography of herself—written in the first-person, and supposedly authored and self-published by a person named Brandi L. Sahlfeld—that simply reworded virtually every sentence from Land’s memoir Maid. When the Authors Guild learned of this, we reached out to Amazon and it promptly took down the offending books.
Land’s experience follows that of Kara Swisher, which was recently reported in The Washington Post. Swisher also found a spate of scam “biographies” on Amazon ahead of the release of her new memoir. And there have been other cases of scammers using AI to plagiarize books—often new releases—by slightly modifying the language and selling the competing books under fake author names. In other instances, scammers have been selling AI-generated books on the same subjects as those authored by best-selling and well-known authors, as in the case of veteran jazz writer Ted Goia, who found books on jazz that claimed to be authored by Frank Gioia and Ted Alkyer—a combination of the names of Frank Alyer, editor and publisher of leading jazz magazine Downbeat, and Goia himself. We have also seen AI-generated content spring up after pre-release listings. These examples illustrate the challenges in vetting for and filtering infringing AI-generated text and show that scammers are not limiting themselves to companion guides.
The clear purpose of these scams is to co-opt and exploit the good-will of authors’ names and appropriate their work, regardless of whether the scam involves direct copyright infringement or trademark infringement. These books not only deceive customers, but siphon off from sales of legitimate books, and have no place in a fair, well-functioning marketplace.
We have been in discussions with Amazon about this issue. We understand that they are taking it very seriously and working on solutions to prevent scam books from going up in the first place. The quality of these AI-generated books is, for now, quite poor, so it is in Amazon’s own interest to crack down on them. Amazon has disallowed listings in the companion books category, except with “limited exceptions for guides with positive reader engagement.” We hope and expect that these measures, at least for now, will start to slow the tide of low-quality, AI-generated companion books.
But we need them to do more. With AI getting better and better at generating book-length readable text, the number of AI-generated books trying to steal sales from human-written books will soon explode. We need Amazon to prevent these books from being posted in the first place and to clearly and conspicuously label books that are fully or mostly generated by AI on the images or in the listings.
Last winter we asked Amazon to require that all book posters label AI-generated books so that customers know when a book is created with AI and can make informed purchasing decisions. Amazon now requires Kindle Direct Publishing users to disclose the use of AI when uploading content, but this information is not yet made visible to consumers.
We are also lobbying for laws that would require AI-generated text content to be labeled and identifiable as such. Unlike AI-generated images and videos, which can be watermarked with current technologies, there is no current technological way for AI platforms to include a permanent identifier in text their models generate. (AI companies are in discussions with lawmakers about including permanent watermarks or metadata in AI generated audio, visual, and audiovisual content.) For that reason, the requirement must be put on the publishers of AI-generated text, and we are working with congressional offices on what that legislation would look like and how it would be enforced.
Relatedly, we are also in contact with developers who work on AI-detection technologies, and we hope that current and future solutions will make it easier to detect and identify AI-generated text. In addition, we have been lobbying for a measured, federal right of publicity law that would cover the name and recognizable style of an author, and give authors more tools to fight against scams that exploit their work.
We encourage any readers who purchase scammy AI-generated books unwittingly to request a refund and file a complaint with Amazon. Consumer feedback is one of the main ways Amazon gauges the severity of the problem.
If you are an author finding scammy summaries or other unauthorized books, please share your information with us by emailing staff@authorsguild.org. If you are a current Authors Guild member, you can also submit a legal help request so we can help you get the content taken down.
