Frankfurt Book Fair puts human judgment at center of AI debate

Artificial intelligence dominated discussion at the 2026 Frankfurt Book Fair, where publishers, authors and technology companies described it as useful for translation, research, marketing and routine work, while resisting the idea that it could substitute for human authorship, editorial judgment or firsthand expertise. NuiNui, a Swiss publisher, is developing an AI translation agent intended to cut costs, make translations more efficient and ease the international sale of rights. Paolo Biano, an editor at the company, said the technology remained a work in progress and that editors would continue to be needed. He pointed to long nonfiction, such as 350-page nature guides, as a promising use because a near-finished English draft could lessen buyers’ worries over translation costs. That calculation has limits, Biano said. Japanese poetry requires a translator able to understand meanings beyond individual words, making it ill-suited to the tool. Ian Strathcarron, chairman of Unicorn Publishing Group, similarly described AI as a productive research, editing and proofreading aid, but argued that readers should expect an author to have spent time writing a book. The distinction between assistance and replacement also emerged in book discovery. Shimmr AI uses the technology to identify overlooked titles and market them to prospective readers. Its founder, Nadim Sadek, said AI could help people express ideas, but that fully machine-generated writing risked losing the unpredictable individual qualities that distinguish authors. Travel publishing provided a more specific case for the value placed on human knowledge. At an Oct. 7 discussion, Lonely Planet Publishing chief executive Piers Pickard said that misinformation could lead travelers to make poor decisions, making trust central to a publisher’s role. Lonely Planet’s research found that younger readers judged travel information by whether it reflected real experience, met their needs, was independent of commercial sponsorship, truthful, current and verified. Pickard said AI could not provide local knowledge in the same way as travel writers. Christine Brisch of Mairdumont said people who live in, or regularly experience, a destination bring an understanding of local life, culture and context that algorithmically generated material cannot fully replace. Both publishers also made the case for books as a structured product in a fragmented information market. Lonely Planet offers curated itineraries for one, two and three days that readers can adapt, while a Mairdumont representative said travelers need connected plans suited to their time, budget, interests and companions, rather than disconnected recommendations. Pickard said books were best placed for organizing information and planning, social media for inspiring exploration, and apps for quick searches once travelers had arrived. The fair’s debate extended beyond the uses of AI to its consequences for attribution, copyright and employment. More than 120 publishers initially signed the Frankfurter Appell, calling for stronger copyright protection, transparency and fair licensing. Separately, Deutsche Welle reported that more than 140 publishers signed an appeal by the German Publishers and Booksellers Association for stricter AI regulation. The legal and technical questions include whether readers can tell, or prove, that a text used AI assistance. Hannes Bajohr predicted that if a text’s origins could not be reliably established, readers would increasingly judge it by quality rather than by the presence of AI. Nina George took the opposing position that authors should not be required to prove their work was made by a human. The dispute is also reflected in policy. The EU AI Act requires transparency on training data, a copyright policy respecting rights reservations and labels for AI-generated content, although an exception applies where editorial control is exercised. Elizabeth Crossick argued that the rules were weak without meaningful consequences for companies that fail to comply. The Trace project, launched by NISO, STM, Ithaka and Counter, is intended to preserve content-provenance metadata through AI inference. In the United States, the Copyright Office has concluded that existing law can decide copyrightability questions without new legislation. Its report found that using AI to assist human creativity does not remove copyright protection, whereas purely AI-generated material, or material without sufficient human control over expressive elements, is not protected. It said the question of human authorship in AI outputs must be assessed case by case and that prompts alone generally do not provide sufficient control. Employment was another concern. On an Oct. 7 Frankfurt panel, Nigel Newton of Bloomsbury, Sarah Arbuthnot and Suzanne Collier agreed that AI would change publishing jobs rather than eliminate them, though they warned that entry-level posts could disappear and employers should retrain staff. Newton said Bloomsbury had created several AI-related roles and that no employee had lost a job because of AI; Collier highlighted critical thinking, problem-solving and entrepreneurship as valuable skills for the industry.