Libraries have long been champions of expanding access to books, and digital platforms like Hoopla Digital, Libby, Audible, Boundless, or Scribd help make that possible. With a few taps, readers can borrow e-books and audiobooks from their local libraries, making literature more accessible than ever. With the emergence of artificial intelligence (AI), libraries have found ways to use technology to expand their offerings efficiently. However, AI also presents a new issue that some librarians might not yet be aware of—audiobooks completely generated by artificial intelligence. While there are hundreds of books narrated by AI, and most of those have some kind of disclaimer saying they are, a wave of audiobooks completely written by large-language models are now flooding reader app databases. 

With the rapid growth of AI-generated content, some library staff may not even be aware that they are adding AI-written books with AI narrators to their collections. These titles can be mixed in with traditional books in large e-resource catalogs, making it difficult to track which books were created by human authors and which were generated by AI. This lack of transparency can pose challenges for libraries striving to maintain quality standards and ensure their collections align with patron expectations.

With some AI-generated books, the addition of fake profiles, websites, and social media pages of the “authors” may make it seem like human writers were responsible for the content.

Some might argue that AI-written books can be produced quickly and at a lower cost, which could help expand digital collections with unique content, particularly in niche genres or underrepresented topics.

Karen McElfresh, E-Read Texas program manager at the Texas State Library and Archives Commission (TSLAC), had this to say. “While AI-written books could be more affordable, they likely don’t meet the criteria we define in our collection development policies, such as author expertise, publisher reputation, and reviews in professional library publications. And adding vast amounts of content like this can make it more difficult for library users to find the content they are looking for if they have to sift through a glut of AI-generated titles in their search results.”

Why is this such a problem? Let’s get into some of the other cons of AI generated and narrated material.

1. Quality Concerns

AI-generated books may lack depth, creativity, and the nuanced storytelling that human authors bring. Readers might find the prose formulaic or lacking emotional depth. Similarly, AI narrators often struggle with tone, pacing, and the expressive nuances that make human narration engaging.

2. Ethical and Copyright Issues

The rise of AI-written content raises questions about originality and intellectual property. If an AI-generated book is based on existing works, is it truly original? Likewise, voice actors have expressed concerns about AI-generated narration taking jobs and using their voices without consent.

3. Reader and Author Preferences

Many readers and authors value the craftsmanship of traditional writing and narration. AI-generated books may not resonate as deeply with readers, and some authors may be hesitant to have their works narrated by synthetic voices.

It might seem obvious that an AI written book might not generate the same interest from readers, but what about a book written by a human author and narrated by AI?

TSLAC’s Talking Book Program (TBP) provides free library services to qualifying Texans with visual, physical, or reading disabilities. TBP Program Director Ann Minner said, “The Texas Talking Book Program has added over 2600 audiobooks to the national collection. We use volunteers to narrate, monitor, and review our recordings. Many of our patrons filter their book selections by voice attribute. They are incredibly selective. Some want only women narrators; some want expressive readers; some want more bass or more treble in the voice. I don’t trust that AI can give the level of nuance that my patrons are used to. And if it can, if AI is learning from existing voices, how are those root voices being compensated? That needs to be addressed. An accurate narration free of pronunciation mistakes and ambient noise can take quite some time. It may sound similar, but I can’t imagine that AI can replicate the genuine artistic and emotional quality recordings created by humans who care about their work and their readers.” 

AI can be a powerful tool in supporting libraries’ efforts to provide many valuable services, like chatbots, event planning, enhanced information retrieval, or budget allocations. But when it comes to audiobooks, it’s clear that human authors and narrators remain irreplaceable in delivering truly compelling stories.

What Can Libraries Do?

If you are concerned about inadvertently purchasing AI generated content for your library, check with the vendor you use to ask if they sell AI-generated content, and if so, how is it identified in their system? For instance, in Baker & Taylor’s TitleSource 360 system, AI-narrated audiobooks usually have “Digitally Narrated Using a Synthesized Voice” in the title of the audiobook. However, there are no hard and fast rules as to how AI content needs to be identified. And if you participate in a subscription e-book package where content gets added automatically by the vendor, it’s important to ask that vendor how they are evaluating content.

Do you have any thoughts or comments on this issue? Let us know at LDO@tsl.texas.gov.