Translation Prompt That Keeps Tone and Terminology
Instead of a literal translation, keep meaning, tone and technical terms — and make the model check its own work at the end.
You are a professional translator working from [SOURCE LANG] into [TARGET LANG]. TEXT [TEXT] CONTEXT - Audience: [AUDIENCE] - Register: [FORMAL / NEUTRAL / CASUAL] - Domain: [DOMAIN] - Glossary (use these exact terms): [GLOSSARY] RULES - Translate meaning, not words. If a literal translation reads unnatural, restructure the sentence. - Keep the register of the source. Do not make a casual text formal or a legal text friendly. - Preserve names, numbers, units, code, URLs and formatting exactly. - Do not add explanations inside the translation. Notes go in a separate section. - If a term has no settled equivalent, keep the source term in brackets on first use. OUTPUT 1) TRANSLATION — clean, ready to paste. 2) TERMINOLOGY NOTES — every choice that a reviewer might question, with the alternative you rejected. 3) UNTRANSLATABLES — phrases that lose meaning, and what you did instead. 4) SELF-REVIEW — read your translation back and list any sentence that still sounds like the source language.
Models translate fluently but not consistently. If two translators must match, the glossary does the work, not the prompt.
02 — Variables
Fill in the brackets
The highlighted fields are the places where your own details go. You do not have to fill all of them; anything left empty is completed by the model's own assumption.
[SOURCE LANG]
[TARGET LANG]
[TEXT]
[AUDIENCE]
[FORMAL / NEUTRAL / CASUAL]
[DOMAIN]
[GLOSSARY]
03 — How to use
How to use it, step by step
- 01
Paste the glossary even if it has three entries; the model drifts on terminology without one.
- 02
For legal or medical text, ask for a second pass that only checks terminology and leaves style alone.
- 03
Run the self-review step on a fresh message so the model is not defending its own first draft.
04 — Variations
Same prompt, different result
One prompt does not solve every case. These variations ask for the same job in a different tone, length or format — pick the one that fits.
Back-translation check
Translate the text back into the source language, then compare with the original and list every meaning shift.
Localisation
Same text, but adapt for the target market: currency, date format, units, examples, idioms. Mark every adaptation.
05 — Tips
Notes that take the result up a level
- The self-review step catches most calque sentences — the ones that are grammatically correct but obviously translated.
- Never let the model translate numbers with units in one pass without checking; unit conversion is where silent errors happen.
06 — FAQ
Frequently asked about ChatGPT Prompts
Is machine translation good enough for publishing?
For internal use, usually. For anything public, treat the output as a first draft: terminology and register are where it fails, not grammar.
Why keep the glossary?
Because the model makes different choices each run. A glossary is the only way to get the same term translated the same way across ten documents.
Next
You may also find these useful
ChatGPT Prompt for Summarising Long Documents (3 Layers)
Turn a 30-page report into something readable in ten minutes: one-sentence core, an executive summary and a bulleted detail layer.
Beginner ChatGPTClaudeChatGPT Idea Generation Prompt: 30 Ideas, 5 Angles
Ends the "nothing comes to mind" loop: multiplies one topic across five angles and scores every idea for feasibility.
Beginner ChatGPTClaudeChatGPT Persona Prompt: Expert Role Template
Instead of a shallow answer to the same question, get an expert's view: identity, experience, bias, boundaries and output rules.
Beginner ChatGPTClaudeResearch Prompt: Balanced Brief With Counter-Arguments
A research prompt that produces a quick but balanced picture: you define the scope, the model must include opposing views.
IntermediateThis page was prepared by the CopyPromt editorial team, first published 25 October 2026. Prompts are written to work across models; brand names are mentioned for identification only.