Europe’s book markets benefit from AI but face rising English-language dependence

11.04.2026

You see an error in the text - select the fragment and press Ctrl + Enter

Artificial intelligence has already taken a key role within the European book market, mostly for translation, audiobook creation, academic publishing, and internal editorial workflows. At the same time, the active use of AI has contributed to increased dependence on English as an intermediary language for Europe’s small-language markets. Conclusions on the topic were given by the authors of the section “AI, Publishing, and Small-Language Markets” in the report “Books in Translation: Trends and Transformations in the European Publishing Market.”

The report emphasizes that the impact of AI is often more important in small publishing markets than in large ones. In places where publishers combine a cultural mission with the need to survive economically, the automation of publishing processes is not merely an optional extra, but one of the factors determining whether a book will reach its audience at all. At the same time, smaller markets must adopt such solutions at a slower and more cautious pace because of high implementation costs and the need for human control.

According to data by Eurostat, Denmark leads Europe in the adoption of AI in business: 27.6% of companies use AI technologies. In Sweden, this figure stands at 25.1%. The report attributes this, in particular, to government initiatives, including the development of Danish-language models and the efforts of the Center for Artificial Intelligence in society. Estonia is developing a national AI strategy for working with the Estonian language, while Slovenia is participating in European initiatives like EuroHPC and building out its computing infrastructure to enable and support the development of Slovenian-language AI.

These changes are especially evident in the translation field:

Taylor & Francis works with more than 30 languages, using machine translation while insisting on high-quality editorial review;
Springer Nature offers authors machine translation so they can submit manuscripts in their language;
the Dutch publisher Veen Bosch & Keuning is testing a limited AI translation program for fewer than ten commercial novels from Dutch to English;
Nuanxed has announced over 900 translations created using AI-assisted processes followed with post-editing by human translators;
the Greenlandic media outlet Sermitsiaq, for which the Danish startup MediaCatch created a translation tool between Danish and Kalaallisut (the official language of Greenland), trained on 15 years’ worth of newspapers.

AI is also quickly gaining ground in the audio book market segment. Audible is expanding its catalog with AI-generated narration, offering over 100 synthetic voices in various languages and accents. Spotify is partnering with ElevenLabs, allowing audiobooks to be narrated in 29 languages. Storytel in Sweden and Poland has launched the Voice Switcher feature with three AI voices; one of them is a synthetic version of Swedish actor Stefan Sauk’s voice. According to Storytel, 89% of listeners have abandoned a book at least once because of the narrator’s voice, not the text. During tests, nine out of ten people were unable to tell AI narration apart from human. In Slovenia, Mladinska knjiga Založba is already using AI for select audiobooks, and on the MK+ platform, some of them are labeled as “machine reading.” For small markets, this model means lower costs and the ability to release niche books that wouldn’t be profitable under a conventional approach.

To date, the most detailed AI guidelines have been developed by academic publishers. Amsterdam University Press and Central European University Press have already introduced specific rules for authors, editors, and reviewers, permitting the use of AI for brainstorming, proofreading, and structural guidance, but require that the use of AI be disclosed and subject to human control. As the report’s authors note, publishers Elsevier and Springer Nature have effectively established an industry standard: AI cannot be the author of a text, but it can perform technical tasks.

The report also highlights that commercial publishers have adopted a stricter approach to AI: Penguin Random House has banned the use of its books for training AI models, while the French publisher Les Nouveaux Éditeurs has banned the publication of books written by AI.

The study specifically describes an experiment involving AI translation. It focused on Slovenian and Zulu — languages that lack a long-standing tradition of mutual translation, dictionaries, university departments, and direct cultural contact. ChatGPT described English as a “bridge” between the languages for translation. On simpler texts, the language model sometimes approached the level of an “average human translator,” but lost details: it translated “book-like contents” as “books” and added “podcasts,” which were not in the original. The report’s authors also cite examples of human errors: Florence Nightingale was confused with an opera singer from Florence, and “iron pills” were translated as “iron fillings.” Their conclusion is that AI will likely replace weak and average translators, but not good editors.

In the translation of Andrej Rozman-Roza’s poem from Slovenian to English, the AI system preserved the main meaning and some of the rhymes, but sacrificed some of the cultural and political nuance. In the Slovenian-English-Zulu translation, the original ironic, politically charged text was replaced with a different, less impactful poem. The study’s authors believe that the system can reproduce “average text, but not the complex linguistic and cultural features of a literary work.”

Main image: Pexels

Copyediting: Ben Angel