EXCLUSIVEThe AI books in your libraries from authors that don't actually exist... and how to detect them
Artificial intelligence is rapidly changing the way that every level of humanity exists, advancing to even writing full-length novels.
AI has seeped into multiple facets of media and entertainment, including digital books written by authors that don't exist.
The literary community has recently been plagued with the problem of approaching AI, and deciphering when a real author is behind the words on the page.
Amazon is flooded with books written by AI, but a fake name and image show up on the book's page so readers have no way of knowing that their digital book wasn't written by a human being.
For example, if a reader searches for diet cookbooks on Amazon, they may find titles written by an author named Stewart Brooks.
The catch? There's no online digital footprint for an author named Stewart Brooks other than his profiles on Amazon and Goodreads.
Upon first glance, there's no indication that Brooks doesn't exist despite an author page on Amazon with a biography and a profile picture.
However, according to Sight Engine, an online service that detects AI images, there's a 94 percent chance that his profile picture is AI-generated.
Books authored by AI are showing up in digital libraries and online retailers
Stewart Brooks is an author on Amazon who has written multiple diet cookbooks and has a profile picture with a 94 percent chance of AI
Brooks authors a variety of cookbooks including Bread Machine, Carnivore Diet, and Fatty Liver: Diet Cookbook for Beginners.
Carnivore Diet starts with an introduction to the history of the diet and what it consists of.
The first three paragraphs were determined to be 100 percent written by AI, according to Undetectable AI.
Darla Robson is another author featured on Amazon who is suspected to be a product of AI. AI or Not determined that her profile picture was likely generated by artificial intelligence.
Robson has also authored similar diet content, including a book titled The New Dr. Nowzaradan Meal Plan System: 1800 Days of Tailored Recipes Designed to Simplify Your Journey to Lasting Health and Effortless Weight Loss.
Her author profile was determined to be 100 percent written by AI or ChatGPT, according to the AI detector ZeroGPT.
The description of the book on Amazon came back as 96.08% written by AI/ChatGPT, according to ZeroGPT.
It's difficult to tell whether an author exists or not without running their profile pictures and words through AI detection sites.
Even then, it can be hard to pinpoint with certainty whether a real person exists behind the books or if a real person did write the book but is using an AI-generated image or partially used AI to complete their novel.
Darla Robson also has a profile photo on Amazon which is suspected to be a product of AI
Magda Tangy is an author featured on Amazon, Hoopla and OverDrive. She oddly only has one earring on and the background is a blank gray
Magda Tangy is another example that shows how complex AI detection can be. She has an author profile on Amazon, but it's unclear if she exists or not.
Her profile picture mirrors those of AI-generated images. She oddly only has one earring on and the background is a blank gray.
AI or Not determined that her image was likely human and Sight Engine came back with only 23 percent certainty that her photo was AI generated.
However, a spokesperson for Reality Defender, a deepfake detector, told 404 Media that the headshot was 85 percent likely to be a product of AI.
Tangy's author biography also seems to be a product of AI, with ZeroGPT determining 100 percent was likely to be produced by AI GPT.
Tangy does have an online footprint, as there is a Facebook and an Instagram profile associated with her name. However, the only photo of herself is the same headshot used on Amazon, and all of her pictures appear to be AI generated.
Another author suspected to be a product of AI is Bill Tarino. Tarino also has a Facebook profile, but he has no posts or a profile picture.
There are over four pages of books available to purchase digitally for free on Amazon written by Tarino, with titles like Puppy Love Parade, Second Chances: A Single Dad & Nanny Romance, and Whispers in the Walls: A Gripping Psychological Thriller.
Tarino also has multiple trivia books on celebrities including Taylor Swift, Lady Gaga, Elon Musk and Justin Bieber.
The covers of his books are clearly AI-generated, with the photos of Swift and Gaga barely resembling the pop stars.
Bill Tarino has authored several books for sale on Amazon including a trivia series on Lady Gaga and Taylor Swift
Booksby.ai is an online bookshop exclusively dedicated to selling books written by artificial intelligence.
Titles include The Imperfect in the Disaster, The Serious: A Proven Divorce, and Auro-Minds and the Hungers. All of these books are available on Amazon, however, there's no label indicating that their authors don't exist.
Komninos Chatzipapas, a founder at HeraHaven.AI and expert in artificial intelligence, told the DailyMail.com that just two years ago there was one AI model that existed which had biases towards using specific words and phrases.
Now, there are dozens of AI models with advanced training, so it can be almost impossible to detect AI with 100 percent certainty.
'Amazon enabled international distribution for new writers, making it easy for anyone to write and publish any work for free,' Chatzipapas said.
'And as with anything, progress is exponential. It took over 500 years from the invention of the printing press to get to the internet. And in just 25 more years, AI has allowed anyone to write any book in a very accessible manner.'
Catzipapas added that the disadvantage to the accessibility of writing and selling novels is that quality can dip.
'We have content guidelines governing which books can be listed for sale, and we have proactive and reactive methods that help us detect content that violates our guidelines, whether AI-generated or not,' Tim Gilman, a spokesperson for Amazon told the Dailymail.com.
'We invest significant time and resources to ensure our guidelines are followed, and remove books that do not adhere to those guidelines.'
Amazon said that they don't offer a disclaimer for AI-generated content, however they do require publishers to disclose whether they used AI.
Digital e-books seemingly written by artificial intelligence are popping up for sale on Amazon and digital library databases
Tarino has several fact books on celebrities that are seemingly AI-generated for sale on Amazon
AI content in digital library databases
Potentially AI-generated eBooks have also made it onto digital library databases like OneDrive and Hoopla.
Tangy's books were available in both library systems, so consumers could use their library cards to check them out. Tarino's titles are only available on Hoopla.
With Hoopla, librarians subscribe to their entire literary database, while OverDrive allows librarians to select content that meets their standard.
Jennie Pu, the director of Hoboken Public Library in New Jersey said that her library uses Hoopla for their Online Public Access Catalogs, also known as OPACs.
Pu, a librarian for over 17 years, told the DailyMail.com that the rise of AI content has been on librarians' radars for years.
'We're very hyper aware of this because we want to make sure that we're providing quality human generated content,' she said.
Pu said librarians are trying to be proactive with spotting AI-generated content, flagging the titles the minute they see it.
'We do have a very strong information network and we are able to advocate and tell the public. This is our work and it's a priority for us,' Pu said.
She said that her and 77 other libraries in North New Jersey started finding books blatantly written by AI in their OPACs.
They raised the concern with Hoopla, who did eventually pull some of the titles and has since said they are working to develop a plan to identify AI content.
Librarians take great care in curating their content and pride themselves on making sure libraries stay trusted institutions with high standards.
'Libraries have earned this position of respect in the community because we've been around for hundreds of years and we take our work very seriously,' Pu said.
Steve Potash, the CEO of OverDrive told the DailyMail.com that the company began implementing rigorous standards to combat AI years ago.
OverDrive works with over 200 librarians, each one has the ability to choose whether they want to include AI content in their database or not.
'I'm not dismissing the catalogs of retail booksellers. But when we have institutional buyers in city, county, state, or national roles they have a higher degree of responsibility in selecting materials, because these are materials that will go into their public library,' Potash said.
OverDrive's catalogs get fed into 40,000 discoverable OPACs globally, meaning librarians around the world can subscribe to their digital content.
Potash told the DailyMail.com that industry standards for terminology are still in development as AI evolves.
'Let's face it. We're all consumers. What's the first thing you do when you see a new product in the dairy case or a new Greek yogurt. You pick up that label, you turn it around. Is this yogurt or Greek yogurt? Does it have probiotics or not?' Potash explained.
'So we have to work with the industry to help create identifiers, how this identifier needs to be properly communicated, and how we standardize.
'What does it mean if it says, "portions may be contributed by generative AI?" What does that mean?'
Sifting through a sea of content online may be difficult for consumers without proper identifiers, especially when AI-generated work is disguised so well.
IRB Media, a publisher with 182 titles in OverDrive's catalog, and 190 results on Hoopla, creates 'summaries' of novels through AI.
For someone looking to read David Grann's novel Killers of the Flower Moon: The Osage Murders, which was later developed into a film starring Leonardo DiCaprio, they may be met with a summary produced by IRB Media.
Pu said that her library has come across AI-generated summaries frequently and are working to educate the public on the difference between human generated content and AI.
Potash suggested that those looking to avoid AI-generated summaries can use the advanced search section to exclude these titles.
He added that readers can also avoid AI content by checking for traditional publishers like HarperCollins or Penguin Random House.
Even though industry professionals are working to develop standards to label AI, the content has continued to flood retailers and online databases.
Some titles, like the novels featured on Booksby.ai and written by Tarino, are easy to detect because the covers are more clearly AI-generated.
The sentence structure often at times lacks a human voice and sometimes just simply doesn't make sense.
Bill Tarino has authored a slew of books for sale on Amazon that are suspected to be AI-generated
For example, The Imperfect in the Disaster by Barreast Wolf, an author identified as AI on Booksby.ai, has a description on Amazon that's difficult to decipher.
The first sentence in the description reads, 'This years best known for us in modern million copies. Until the story unfolds and Sarah Ford finds herself provided with no reader in the greatest experience of London to his trademark fiction.'
However, other novels suspected to be products of AI are more coherently written. The first two sentences of Brooks's Carnivore Diet reads, 'The carnivore diet is a low-carbohydrate and high-protein dietary regiment that focuses on animal products while eliminating plant-based foods.
'This means that you should eat meat, fish, eggs, and animal fats to avoid fruits, vegetables, grains, legumes, nuts and seeds. The concept of the carnivore diet has been around for centuries.'
AI-generated content can be unclear, as some novels are entirely composed of artificial intelligence, while some advertising that AI only 'partially' contributed.
Analyzing the author's profile picture, their digital footprint, and how they formulate sentences can help, but even that can bring readers to a dead end.
Books featured on Booksby.ai that are authored entirely by artificial intelligence are on sale on Amazon
AI stealing from authors
Not only is AI-generated content showing up in library databases and retailers, but some authors have shared that their names have been associated with AI content.
Jane Friedman, an author and journalist featured in The New York Times, The Atlantic, and NPR, among other publications, previously shared that her name was attached to fake books on Amazon and Goodreads.
She shared screenshots of her name under books she didn't write for sale on Amazon and available to review on Goodreads.
Friedman speculated in a blog post that the books were written by AI, writing, 'I’ve been blogging since 2009 - there’s a lot of my content publicly available for training AI models. As soon as I read the first pages of these fake books, it was like reading ChatGPT responses I had generated myself.'
Friedman reached out to Amazon, and a representative replied asking for the trademark reservation. When she said she didn't have one, she received a response that the case wouldn't be removed from sale.
Goodreads, however, did take action and removed the fake titles from her author profile.
As an author whose dealt with the potential downfalls of AI, Friedman told the DailyMail.com: 'In a best-case scenario, AI-generated works are so low quality and unremarkable that they simply sink to the bottom of searches, never get reviewed, and never appear to readers. If I had to guess, that's the fate of the large majority of AI-generated work.'
Algorithms have an impact on whether users even see the AI-generated content among the thousands of titles available.
Potash told the DailyMail.com that OverDrive doesn't employ an algorithm to push certain content.
AI-generated content has presented debates on how books are vetted and published online
Books authored by Stewart Brooks are on sale on Amazon
Authors suing AI companies
Although AI has advanced significantly, the model still lacks the level of creativity of a human being.
AI can only pull from existing data, so a user can ask it to produce content that mimics their favorite author, but it can't come up with entirely new information.
The concept can become a slippery slope when authors have copyrighted works that AI pulls from without permission.
Under US Copyright Law, work is copyrightable if it is original and possesses a degree of creativity, is fixed in a tangible medium, and is a product of human authorship.
Since AI is not a product of human authorship, work created by it cannot be copyrighted.
However, AI has in the past pulled from human copyrighted work. The Atlantic obtained a data set of more than 170,000 books that were pirated to train AI models.
The database, Books3, pulls content from books without the author's permission to help generative AI systems create advanced content.
A slew of lawsuits have transpired against Meta, Bloomberg and OpenAI for using copyrighted content.
In September 2023, a group of writers with the Authors Guild filed a lawsuit against OpenAI for 'systematic theft'.
'It is imperative that we stop this theft in its tracks or we will destroy our incredible literary culture, which feeds many other creative industries in the US,' the CEO of Authors Guild, Mary Rasenberger, said in a statement at the time.
'Great books are generally written by those who spend their careers and, indeed, their lives, learning and perfecting their crafts. To preserve our literature, authors must have the ability to control if and how their works are used by generative AI.'
Last February, a California judge dismissed parts of a similar lawsuit on the grounds that the work developed by AI wasn't close enough to constitute copyright.
OpenAI, Meta and other tech companies have argued that they're allowed to use copyrighted work to train their models under the fair use doctrine, which allows limited use of copyrighted material without the owner's permission.
As AI models begin to advance with the help of thousands of authored works, it will only make it increasingly difficult to decipher whether a real writer is behind books showing up on library databases and retail booksellers.
