Artificial Intelligence (AI) is playing an increasingly prominent role in the publishing industry with AI-generated books, illustrations and translations becoming more common in library collections. The Collection Development Unit (CDU) has already encountered AI-generated titles being purchased by libraries, highlighting the need for critical evaluation.
AI offers several benefits. It can generate stories rapidly, assist with translations and make new content available to readers more quickly than traditional publishing processes. However, these advantages are accompanied by significant concerns.
One of the most important is bias. AI systems are trained on vast collections of books, websites, and other digital sources that often reflect historical inequalities and stereotypes (K.P., 2024: 2-3). As a result, AI-generated fiction may unintentionally reproduce biased representations of race, culture, gender, and other identities found in the classical fictional materials from which large language models (LLMs) learn from (Alteri, Birch & Huet, 2018; Cressman Library, n.d.; Chartier, 2026). This raises important questions about authenticity, representation and inclusion.
Quality is another key consideration. While some AI-generated works are engaging, others may contain repetitive storylines, factual inaccuracies, weak character development or poor-quality illustrations (Saleh, 2025; Tanzi, 2025). Librarians must continue applying rigorous selection criteria to maintain collection standards.
Ethical concerns extend beyond literary quality. Reports suggest that companies such as Anthropic have used physical books, including rare and out-of-print titles, for AI training, sometimes through destructive scanning processes (Chartier, 2026). This raises concerns about preservation and access to historical knowledge (Thatcher, 2026). More concerningly, the destruction of scanned books for AI training to can be compared to George Orwells “1984”, where original records are eliminated, leaving only altered versions behind (Chartier, 2026).
Copyright and intellectual property issues remain unresolved. Debates continue over whether copyrighted works have been used for AI training without permission from creators and rights holders, and whether AI-generated content blurs distinctions between inspiration, plagiarism, and originality (Saleh, 2025).
AI-assisted translation presents both opportunities and risks. Literary translation requires having the skills to correctly translate cultural norms, humour, tone, and context of words being used (Shahmerdanova, 2025: 62-63). While AI can assist with translation, it often struggles to capture the nuances that make stories meaningful and culturally relevant (Serra, 2025). CDU has encountered AI-translated children’s books, particularly in isiXhosa books which are of poor quality. IsiXhosa speaking CDU staff identified concerns regarding mixed languages (Zulu and Xhosa words), literal translations and inappropriate diction found within these texts. Although technically correct, such translations may fail to communicate meaning effectively or reflect cultural context accurately (Horodyska, 2025; Bradford, 2026).
AI also raises broader questions about creativity and authorship. Research suggests that LLMs often produce predictable content and may reinforce dominant cultural perspectives. However, studies indicate that AI and human writers possess complementary strengths – with humans excelling in originality and atmosphere and AI performing well in suspense and character development (Boumous, 2024: 10-13).
By applying established collection development principles, including quality, relevance, diversity, representation and community value, librarians should evaluate AI-generated materials responsibly while remaining open to innovation.
Join the Conversation
AI-generated books and translations are becoming more common, prompting libraries to consider how these materials should be evaluated and included in collections. Librarians play a key role in assessing AI-generated books, whether it is the images, translations or texts that are AI generated. We, at the CDU, encourage librarians to share their experiences, questions and approaches to collection development in this evolving area. Through collaboration and discussion, libraries can develop best practices for managing AI-generated content. As librarians have always adapted to technological change, ongoing dialogue will help ensure collections remain relevant, inclusive and responsive to the communities they serve.
By Masturah Adams
Assistant Professional Officer: Acquisitions
Collection Development Unit (CDU)