In February 2022, while translating Dana Spiotta's novel Wayward into French, literary translator Yoann Gentric decided to test whether AI could replace him. He fed a short sentence into DeepL, a neural-network translation engine. The AI rendered "Bright, sharp night air, bracing" as L'air de la nuit, vif et vif, était vivifiant (The night air, lively and lively, was enlivening)—capturing meaning but missing the absurd repetition. Gentric's own translation, published a year later, was superior: L'air pur et piquant de la nuit, vivifiant.
When Gentric repeated the test in spring 2026, the result was more unsettling. DeepL suggested L'air nocturne était vif, pur et vivifiant. Though it added a verb, it used three different words with a musical ring. "I don't know if it's just chance or a fine-tuned algorithm, but nocturne and pur is not bad," he said.
AI Disruption in Europe's Translation Industry
Chatbots powered by large language models (LLMs) are rapidly infiltrating work and leisure, but few sectors face disruption as fast as translation in Europe, home to over 200 languages and a booming tech sector. A joint survey by French authors' societies ADAGP and SGDL found that 79% of translators believe AI poses a threat to replacing all or part of their work. In Britain, a 2025 survey showed 84% expected lower demand and pay.
For many, the nature of work has already shifted. Laura Radosh, a Berlin-based German-to-English translator, saw monthly job requests drop from four to one. Most were "post-editing" jobs—correcting machine-translated texts. "Post-editing took me as much time as translating from scratch," she said, and it pays less, often by the hour at unacceptable rates. In Germany, publishers offer €2–8 per page for post-editing, a quarter of typical rates for translating from scratch.
Even technical translation rates have fallen. Radosh was offered €0.60 per line, down from a previous low of €0.80. Literary translators, traditionally lower-paid, earn an average of €20,363 annually before tax, according to the German translators association VdÜ. Many find the numbers no longer add up; Radosh recently took a part-time book-keeping job at an NGO.
Human Strengths Persist
Marco Trombetti, CEO of Translated, noted that the human brain produces about 3,000 words of translation per day. "The cost of human translation has been defined by our 100 billion neurons. If we change that, we change the unit economics." Yet AI still struggles with context. In 2024, Springer Nature's machine translation rendered "Capital" in a book title as Hauptstadt (capital city) instead of Kapital. A spokesperson called it a rare error from a limited pilot that has since ended.
Jörn Cambreleng of Atlas, a French literary translation promoter, said: "Machine translation is not creative. It produces generic sentences. Good human translators strive to say something new."
Literary Translation Holds Its Ground
Ironically, literary translation appears safer than technical work. Harlequin France uses AI with human post-editing for pulp fiction, but in Germany, translated literature made up a historic 15% of new books in 2024 (8,765 titles). Authors increasingly contractually forbid AI use. "AI really cannot do dialogue," said translator Katy Derbyshire. "My body has experienced the pain and joy literature conveys. I understand what someone might scream when they hit their toe—an algorithm doesn't."
Fernando Prieto Ramos of the University of Geneva noted a drop in translation course applications three years ago, but the trend is reversing with more diversified training. Even AI developers acknowledge limits. Trombetti gave an example: "In Italian, 'Solo tre parole: non sei solo' translates literally to 'Just three words: you are not alone,' but that's four words. Machine translation still struggles."
Heimburger concluded: "I am not scared of AI because I know it cannot do what I can. I am afraid of people who think AI can do my job."
