Key Takeaways
- Literary translator Yoann Gentric found early AI translations clumsy but noted improvement by 2026, suggesting rapid progress in machine‑translation quality.
- Surveys show 79 % of European translators view AI as a threat to their livelihood, with many expecting lower demand and pay.
- Human translators increasingly receive “post‑editing” jobs that are as time‑consuming as original translation but paid far less—often a quarter of the rate for translating from scratch.
- Despite speed gains, AI still struggles with context, idiomatic nuance, creativity, and dialogue, areas where human embodiment and cultural intuition remain decisive.
- Literary translation is comparatively safer than technical translation, yet fears persist that clients may overestimate AI’s capabilities and undervalue human expertise.
Yoann Gentric’s 2022 AI Experiment
In February 2022, while rendering Dana Spiotta’s novel Wayward into French, literary translator Yoann Gentric tested whether AI could replace him. He fed the short, non‑verbal sentence “Bright, sharp night air, bracing” into DeepL, which returned: L’air de la nuit, vif et vif, était vivifiant (“The night air, lively and lively, was enlivening”). Gentric noted the AI had captured the meaning but produced an absurd repetition, far inferior to his own published translation: L’air pur et piquant de la nuit, vivifiant.
Re‑testing AI in 2026
Repeating the experiment this spring, Gentric observed a different outcome: DeepL suggested L’air nocturne était vif, pur et vivifiant. Although the engine added a verb and lost the original’s stylistic brevity, it employed three distinct words with a “musical ring.” Gentric remarked, “I don’t know if it’s just chance or a fine‑tuned algorithm at work, but nocturne and pur is not bad,” indicating noticeable progress in AI’s handling of nuance over four years.
AI’s Rapid Infiltration of Translation
Chatbots powered by large language models (LLMs) are spreading through work and leisure, but few sectors feel the impact as acutely as Europe’s translation industry, home to over 200 languages and a thriving tech scene. A joint survey by the French authors’ societies ADAGP and the Société des Gens de Lettres found that 79 % of translators believe AI “poses a threat of replacing all or part of their work.” In Britain, a 2025 poll showed 84 % anticipated lower demand for human translation, translating into reduced earnings.
Shifting Work Patterns and Pay Cuts
Laura Radosh, a Berlin‑based German‑to‑English translator, used to receive about four job requests per month from universities, professors, and museums; last year that fell to one per month. Many offers were “post‑editing” assignments—correcting texts already run through a machine‑translation engine. Radosh said, “Post-editing took me as much time as translating from scratch,” yet such work is usually paid hourly, not per page or book, and deemed “at unacceptable rates” by the French translators’ association. In Germany, publishers offer two to eight euros per page for post‑editing, roughly a quarter of the average rate for translating a page from scratch. Even regular technical translation rates have slid; Radosh recalled being offered “60 cents a line,” whereas previously 80 cents was the floor.
The Precarious Economics of Human Translation
Before LLMs, translation was already unstable: a German translators association (VdÜ) survey reported the average annual income for literary translators at roughly €20,363 before tax. Marco Trombetti, co‑founder and CEO of machine‑translation firm Translated, argued that the cost of human translation has historically been limited by the brain’s neuronal capacity—about 100 billion neurons enabling roughly 3,000 words per day for an average translator, 6,000 for the best. He contended, “If we change that, then we change the unit economics of translation,” hinting that AI could shift the cost baseline dramatically.
AI’s Continuing Weakness with Context
Despite speed gains, machine translators still falter on context. Springer Nature offers free auto‑translation of authors’ books, promising subsequent “human checks.” In 2024, an English‑language volume titled Capital in the East: Reflections on Marx was machine‑translated into German; DeepL rendered the chapter heading “capital” as Hauptstadt (“capital city”) instead of Kapital. A Springer Nature spokesperson acknowledged the error, stating, “Our AI‑supported translation is human‑led and reviewed by professional editors. Errors like this are rare and regrettable, and this instance relates to a limited pilot that has since ended.”
The Creativity Gap
Jörn Cambreleng, director of Atlas—a French organisation promoting literary translation—observed, “Machine translation is not creative. These systems are built to produce sentences that are generic, sentences that have been said before or sound like they have been said before. Whereas good human translators strive to put into words something that has never been said before.” This highlights the irreplaceable role of human ingenuity in crafting novel expressions.
Literary Translation’s Relative Safety
Ironically, literary translation appears comparatively safer than its technical counterpart. Harlequin France, owned by HarperCollins, is experimenting with AI‑generated translations post‑edited by humans, though trials remain limited to “pulpier” titles such as A Mistress’s Confession and The Embrace of a Prince. In Germany, where overall new‑book output declines, literature in translation held steady at 8,765 titles in 2024—15 % of total publications. Many authors now contractually prohibit AI use in translation, noted Marieke Heimburger, a Danish‑to‑German translator who chairs VdÜ.
Dialogue and Embodied Understanding
Katy Derbyshire, a Berlin‑based translator of works by Clemens Meyer and Christa Wolf, stressed AI’s limits with dialogue: “When you are translating from scratch, you learn to understand the characters and their motivations, and you’re constantly adjusting them in your head – to individual situations, but also to genre. The dialogue that AI came up with just didn’t suit the character description at all.” She added, “My body has experienced all the pain and the joy that literature strives to convey. I understand what someone might scream when they hit their toe on the bed frame – an algorithm doesn’t.” This embodied insight remains beyond current AI reach.
Enrollment Trends and Training Adaptations
Fernando Prieto Ramos of the University of Geneva’s faculty of translation and interpreting noted a dip in applications to translation courses three years ago, coinciding with the generative‑AI hype. “But the trend is gradually reverting again with a more diversified training offer,” he said, suggesting that institutions are responding by broadening curricula to include AI literacy alongside traditional skills.
Idiomatic Limits Acknowledged by AI Developers
Even creators of machine‑translation tools admit lingering shortcomings. Trombetti illustrated with an Italian phrase: “Solo tre parole: non sei solo.” A literal word‑for‑word render yields “Just three words: you are not alone”—four words, not three, breaking the original’s concise structure. He conceded, “That’s something that machine translation still struggles with.”
Fear of Misunderstanding AI’s Role
Heimburger captured the translators’ prevailing anxiety: “I am not really scared of AI, because I know it cannot do what I can do. What I am afraid of is the people who think that AI can do my job.” The concern is less about the technology itself and more about clients and employers overestimating its capabilities, leading to degraded work conditions and undervaluation of human expertise.
In sum, while AI translation has advanced—from producing repetitive, awkward phrasing to generating smoother, more varied output—it still falls short on context, creativity, idiomatic fidelity, and the deep, embodied understanding that literary translators bring to their craft. The industry’s economics are shifting, with many professionals facing lower pay and more post‑editing work, yet literary translation retains a niche where human skill remains indispensable. The ongoing challenge lies in ensuring that AI’s role is seen as a supplement, not a substitute, for the nuanced art of translation.
https://www.theguardian.com/technology/2026/may/08/being-human-helps-despite-rise-of-ai-is-there-still-hope-for-europes-translators