Authors' sentiment on generative AI in the book industry

<p dir="ltr">The rise of generative AI presents both opportunities and challenges for the book industry, particularly in relation to intellectual property and compensation. As AI models are frequently trained on large datasets that may include copyrighted works, concerns about unauthorised use have emerged. While legal and user perspectives have received some attention, limited research has explored how authors themselves view these developments. This study addresses that gap by examining authors’ perspectives on AI in the book industry, focusing on two main questions: first, whether authors are willing to allow their works to be used for AI training, and second, if so, what forms of compensation they expect for both existing and future works, as well as the factors influencing these decisions.</p><p dir="ltr">The study draws on data from an online survey conducted via Qualtrics, targeting members of the Australian Society of Authors. The analysis employs Seemingly Unrelated Bivariate Probit models and a Seemingly Unrelated Regression model to identify key determinants of authors’ decisions regarding AI training and compensation expectations. The results indicate that perceived fairness, the purpose of AI use, and the belief that one’s work has already been used in AI training significantly influence authors’ willingness to allow such use. These effects vary across subgroups defined by AI-related knowledge, professional standing and book genre. Additionally, fairness concerns, expected impact on income, and perceived replacement risk are significant factors shaping compensation expectations. The findings contribute new empirical evidence to debates at the intersection of AI and copyright, offering valuable insights for authors, publishers, AI developers, and policymakers.</p>