The first 3 interviews are free · Start free or see the prices

Multilingual Interviews: AI Interviews in More Than One Language

An interview in a second language is not the same guide with translated words. These rules cover translation, local examples, forms of address, screening and analysis across languages.

Overview

An AI moderator makes it practical to interview people in their own language without a moderator for each language. That removes a large cost. It does not remove the work of adapting the guide and checking the results.

In User Evaluation, you set one language for each study. The moderator asks and listens in that language. Participants see the interview room in that language where a translation exists. The help docs list more than 40 interview languages.

Because the language is a study setting, a project in three languages is three studies with the same structure. This also keeps quotas and reports separate for each language.

Rules

  1. 01

    Run one study for each language.

    Each study runs in one language. Separate studies also let you set a separate target and screener for each market.

  2. 02

    Have a native speaker review the translated guide.

    A literal translation can change the meaning of a question or make it sound strange. Idioms are the most common problem.

    Do not

    Was the setup a breeze? (translated word for word)

    Do

    How long did the setup take, and what was the hardest part?

  3. 03

    Replace local examples, not only words.

    Payment methods, brands, holidays and units differ by market. An example that is normal in one market can confuse people in another.

    Do not

    Think about the last time you paid a friend back with Venmo.

    Do

    Think about the last time you paid a friend back for something.

  4. 04

    Tell the moderator which form of address to use.

    Many languages have a formal and an informal "you". The wrong form can sound rude or too familiar for your audience.

    Do not

    No instruction about formality.

    Do

    Note for the German study: use "Sie", not "du".

  5. 05

    Keep product and feature names in their original form.

    Participants see the names in the product. A translated feature name in the interview can make them think of a different thing.

  6. 06

    Screen for the language that the interview uses.

    A participant who lives in a country may not be fluent in its main language. Add an open screener question in the interview language and check the answer.

    Do not

    Do you speak Spanish? Yes / No

    Do

    En pocas palabras, ¿qué usa para organizar sus tareas en el trabajo?

  7. 07

    Pilot each language version.

    A guide that works in English can produce strange follow-ups in another language. Ask a native speaker to take the pilot and read the transcript.

  8. 08

    Analyze each language first, then compare.

    Answer length and directness differ between cultures. A shorter answer is not a weaker opinion. Code each study on its own, then compare themes, not word counts.

  9. 09

    Check the translation of every quote you put in a report.

    A translated quote can lose a qualifier, such as "a little" or "sometimes". Ask a native speaker to check the quotes that support your main claims.

Checklist before launch

  • There is one study for each language, with the same structure.
  • A native speaker reviewed each translated guide.
  • Local examples replace examples from your home market.
  • Probe notes state the form of address for each language.
  • Product names stay in their original form.
  • Each screener checks the interview language with an open question.
  • A native speaker took a pilot in each language.
  • You plan to code each language before you compare them.

Questions

Yes. In User Evaluation you set the language for each study, and the moderator asks and listens in that language.

A machine translation is a good first draft. A native speaker must review it for meaning, tone and local examples before launch.

Yes, at the level of themes. Do not compare answer length or the strength of words directly, because those differ by culture.

Ask real people your question

The AI moderator interviews screened people by voice or text. Each claim in the study report links to its clip. The first 3 interviews are free.

Start free