AI is reshaping European translation, but human translators are not done

TL;DR:

• A 2025 UK survey found that 84% of translators expected lower demand for human translation work; a French ADAGP/SGDL survey found 79% believed AI poses a threat of replacing all or part of their work.
• Berlin-based translator Laura Radosh saw monthly job offers fall from four to one in 2024, with much of the remainder shifted into lower-paid “post-editing” work — typically two to eight euros per page versus a quarter of the equivalent rate for translating from scratch.
• Resultsense view: this is the most concrete European labour-market case study of LLM-driven white-collar disruption to date — surveys plus rate-card data plus publisher pilots, not anecdotes. UK translators are explicitly affected, and the comparative resilience of literary translation offers a counter-pattern worth tracking for other creative fields.

What the data shows

Among French translators surveyed by ADAGP and the Société des Gens de Lettres, 79% said AI threatened to replace some or all of their work. The 84% UK figure is from a separate 2025 survey of British translators expecting lower demand and lower pay. The German VdÜ association’s own data put average pre-tax income for literary translators at as little as €20,363 (£17,200) per year — already precarious before LLMs arrived.

Rate compression is the operational signal: Radosh said she had been offered work at 60 cents (51p) per line, when 80 cents (68p) had previously been the floor she had ever encountered. In Germany, post-editing work — correcting a machine-translated draft — is paid hourly at rates publishers describe as “unacceptable” given the time required.

The literary exception

Literary translation, traditionally the lowest-paid end of the market, is proving more resilient. In Germany, books in translation reached 15% of new output in 2024 (8,765 titles) — a historic high — and authors are increasingly inserting clauses in contracts that prohibit AI use during translation. Translators interviewed by The Guardian say machine translation still struggles with dialogue, character voice and embodied understanding. Katy Derbyshire, a Berlin-based translator, told the paper: “I understand what someone might scream when they hit their toe on the bed frame — an algorithm doesn’t.”

Where the line moves

The HarperCollins-owned imprint Harlequin France is piloting AI-translated, human-post-edited work, currently confined to mass-market romance titles. Springer Nature, the German-British academic publisher, offers free auto-translation followed by professional human review — a model that has produced visible errors when context fails. DeepL’s accuracy on specific stylistic problems has measurably improved between 2022 and 2026, according to one translator’s repeated experiment cited in the source.

UK context

The 84% UK survey figure makes this a UK labour-market story, not just a continental European one. The Translators Association, the Society of Authors and the CIOL all face the same regulatory question: how to value, certify and price human translation in a sector where machine translation is competent enough to set the entry-level rate. Authors and translators in the UK have been pushing for stronger contractual protections; this week’s data adds weight to that effort.

Looking forward

Expect UK and EU sector bodies to push harder for transparency rules around AI in translation contracts and for differentiated certifications that distinguish human-led from machine-led work. The deeper question — whether literary translation’s resilience generalises to other creative-judgement professions — will play out across the next two to three publishing cycles, not in months.