1. Where AI is used
Zone of Truth uses AI in three places: (1) an AI interviewer conducts research interviews — it asks the questions, adapts follow-ups to the answers, and decides when the interview is complete; (2) voice answers, when a respondent chooses to record them, are transcribed by an automated speech-to-text service; and (3) an AI generates the structured interview report the research team reads, and powers an insights assistant researchers can ask about their own data.
2. How respondents are informed (EU AI Act, Article 50)
- The invite page tells respondents an AI interviewer runs the conversation, before they start.
- The interview screen carries a persistent “AI interviewer” label on every question.
- If a respondent asks mid-interview whether they are talking to an AI, the interviewer confirms it plainly — it is instructed never to deny or evade being an AI. (What it keeps confidential is the research sponsor’s product idea, to avoid biasing answers.)
- AI-generated reports carry a visible “AI-generated” label and embedded machine-readable provenance metadata (generator, model, timestamp) in the stored report, the export, and webhook payloads.
3. What the AI does — and does not — do
- It classifies research fit: how well a respondent matches the study’s target profile (strong / possible / poor / disqualified), and it may end an interview early when there is clearly no relevant experience to discuss. This affects nothing beyond the interview itself — our Terms prohibit customers from using reports or fit ratings for employment, credit, insurance, or any decision with legal or similarly significant effects.
- It does not analyze voice characteristics, infer emotions from audio, or biometrically identify anyone. Recordings are transcribed to text first; all analysis happens on the text.
- It can be wrong. Reports are grounded in the transcript (quotes are verbatim and link to their source answers), but summaries and classifications are model outputs — the product tells researchers to verify against the transcript, and every report ships with an explanation of how it was produced.
- Respondent data is not used to train AI models — ours or our providers’.
4. Your options as a respondent
- Recording audio is optional — you can always type instead.
- You can stop at any time, and choose to discard your answers — discarding deletes them, including recordings.
- You can ask the organization that invited you for a copy of your interview data and report, an explanation of how the report was produced, correction, or deletion — and we assist them as their processor. If you don’t know how to reach them, write to contact@zoneoftruth.com.
5. Questions
This page supports our transparency duties as an AI-system provider (EU AI Act Article 50) and AI-literacy measures (Article 4). Questions: contact@zoneoftruth.com.