A large study published in September 2026 by the Frankfurter Buchmesse and Dosdoce, a Spanish firm specializing in the digital transformation of cultural industries, reveals that 85% of professionals surveyed already use AI tools at least at one stage of their audio book production. The study, based on responses from 53 publishers, studios, and distributors, as well as contributions from 32 other professionals, indicates that AI is being used not only for synthetic voice generation but also for tasks such as text preparation, quality control, and metadata generation. The report highlights that the immediate interest of AI lies less in interpretation than in the backstage of production. According to the study, AI tools are being used for text preparation, pronunciation guides, editorial analysis, quality control, and alignment between text and recording. Claude and ChatGPT are cited for analysis, text adaptation, casting preparation, or pronunciations. Pozotron is used for error detection and text-audio control, while ElevenLabs is more focused on voice generation and cloning. The economic savings from AI integration are real but limited, with the study suggesting that savings could be between 20 and 50% from the second and third year of use, far from the 85 to 95% reductions sometimes announced. Quality control alone represents a significant amount of work, with the authors of the report estimating a verification pass to be about 1.2 times the final duration of the audio book. The study suggests that AI could serve less to replace existing productions than to produce audio books that, until now, would never have been recorded: old funds, specialized essays, academic works, small markets, or translations into languages for which commercial potential does not justify a studio passage. The report envisions a production at three speeds: premium human narration for major new releases, synthetic production for funds or niche works, and a hybrid model combining human voice and AI tools. The professionals surveyed judge non-fiction, technical, or academic works more easily automatable, while fiction remains much more dependent on human interpretation. In France, Laure Saget, general director of Audiolib at Hachette, explains that the company has already adapted its verification processes to integrate AI tools but insists on maintaining human supervision at each stage and on transparency between publishers, authors, interpreters, and the public. The main obstacles to AI deployment are not technical, with half of the respondents citing limitations imposed by literary agents, authors, or editors refusing certain uses, as well as platform restrictions. Only about 10% mention technical difficulties of integrating AI into their existing tools. When addressing voice cloning, 90% demand more transparency towards authors and listeners on the use of AI, and 60% insist on consent for cloned voices, legally constituted data sets, and fair compensation for interpreters. In Europe, since August 2, 2026, the AI Act imposes the detectable marking of audio content generated artificially. Providers of general-purpose AI models must also respect European copyright, including rights reservations, and publish a summary of the content used for their training. In France, a bill adopted by the Senate in April and by the Culture Committee of the National Assembly in June provides that in case of civil dispute, the use of a work protected by an AI system may be presumed as soon as an indicator makes it plausible, unless the provider provides proof to the contrary. In the United States, audio books narrated by AI would have represented only about 0.03% of the audio market's revenue in 2025, according to the Audio Publishers Association (APA). Its survey on sales, conducted by Toluna, estimates the American audio book market at $2.43 billion in 2025, up 9%. Another APA survey, conducted by Edison Research among consumers, indicates that only 16% of audio book listeners have already listened to a title read by a synthetic voice. In France, the issue is no longer marginal: 10 million people listened to at least one digital audio book in 2025, according to the Médiamétrie-Sofia-SNE-SGDL barometer. The digital audio book listener is on average 31 years old and 48% of these listeners are subscribed to at least one platform allowing the reading or listening of books. The arrival of less costly productions could feed a market that already has a large audience, but more titles also mean more competition to be discovered. The study also identifies the risk of "AI slop," a term used by Michael Tamblyn, CEO of Rakuten Kobo, to describe the mass of low-cost generated content that risks drowning the worked-on works and degrading the trust of readers. Despite these challenges, the professionals surveyed remain optimistic about growth, with 44% anticipating a strongly disruptive effect of AI on the audio market in the next three years and 42% foreseeing a more moderate growth. The report also highlights the importance of the Apostolic Library, which preserves a collection that far exceeds religious texts, including about 180,000 manuscripts and archival documents, 2 million printed books, and other items. The library also preserves about forty complete or nearly complete manuscripts of The Divine Comedy, including the famous Dante Urbinate. The Apostolic Archives preserve documents produced or received by the government of the Church and the Holy See over nearly twelve centuries. Researchers can currently consult the documents up to the end of the pontificate of Pius XII, in October 1958.