Skip to main content
2.6

The Translator

Estimated time: 25 minTool: Claude Pro
After this drill, you can:

After this drill, you can use AI for translation, tone adaptation, and cross-cultural localization — with quality checks built in.

Why this matters

Language work is one of the highest-value, most immediate applications of AI for non-English speakers and international professionals. But raw translation is only the beginning. Tone adaptation (the same message for different registers), localization (adjusting cultural references), and back-translation (verifying accuracy) are the professional-grade skills this drill covers. For Icelandic speakers especially: AI translation of technical content has improved dramatically but still needs human review for nuance.

How to do it

  1. 1

    Choose a real piece of text to translate or localize

    A professional document, email, or web copy you need in another language. Or: content in your first language that needs to work in English.

  2. 2

    Use the quality translation prompt — specify register and audience

    Register matters as much as accuracy. A legal document translated in casual register is wrong even if every word is correct.

  3. 3

    Run the back-translation check

    Translate the AI output back to the original language in a new conversation (so Claude hasn't seen the original). Compare to your original. Differences reveal translation decisions that need your judgment.

  4. 4

    Apply one tone adaptation

    Take the translated content and request a version for a different register: formal → informal, technical → accessible, professional → conversational.

The prompt

PROMPT — Quality TranslationModel: Claude Pro
Translate the following text [FROM LANGUAGE] to [TARGET LANGUAGE].

Register: [formal / professional / casual / technical]
Audience: [who will read this — e.g. legal professionals, general public, young adults]
Purpose: [what this text needs to accomplish — inform, persuade, instruct]

Important: Do not translate literally if a more natural expression exists in [TARGET LANGUAGE]. Flag any culturally-specific references that may not translate directly.

Text to translate:
[PASTE YOUR TEXT]
PROMPT — Back-Translation Quality CheckModel: Claude Pro
Please translate the following text back to [ORIGINAL LANGUAGE]. Do not try to recognize or reconstruct the original — translate fresh from what you see:

[PASTE THE AI-TRANSLATED VERSION]

Success criteria

  • ✓You produced a translation with explicit register and audience specification
  • ✓You ran the back-translation check and identified at least one difference
  • ✓You produced one tone adaptation of the translated content

Common mistakes

Not specifying register and audience

→ "Translate this to Icelandic" produces an acceptable default. "Translate this to formal Icelandic for a government audience" produces professional content.

Treating the back-translation as pass/fail

→ The back-translation will always differ somewhat — that's expected. The question is: do the differences represent acceptable translation choices or mistranslations?