AI Usage Policy

• Understanding your documents — extracting role, skills and experience from the JD and your (redacted) resume.
• Planning the interview — choosing topics and difficulty for your level and focus skills.
• Asking questions — generating each question and adaptive follow-up from your previous answer.
• Evaluating answers — scoring, quoting your words, diagnosing, and drafting a stronger rewrite.
• Voice mode (optional) — transcribing your spoken answers and reading questions aloud; audio is processed in memory and never stored.

InterviewIQ is built on a routing layer that can use different model providers (e.g. OpenAI, DeepSeek, and other OpenAI-compatible endpoints). We pick capable models per task to balance quality and cost, and can swap providers as the field evolves.

Whatever the provider, the privacy contract is the same: only de-identified text is sent — direct identifiers are redacted from your documents and your interview answers before any downstream model call.

• Documents (JD + resume) are redacted before parsing — direct identifiers like email, phone and links are stripped first.
• Your interview answers are used to generate follow-ups and your scorecard; we redact direct identifiers from answers before those downstream model calls too.
• Your original files never reach any model and are deleted after parsing. Your original answers are kept only for your own transcript.
• If you use voice mode (optional), your spoken answer is sent to OpenAI for transcription and is processed in memory, never stored as audio — only the resulting text is kept, and direct identifiers are redacted before evaluation.
• We do not use your data to train models.
• Full detail in our privacy policy.

You can answer out loud and have questions read aloud. Speech-to-text and text-to-speech are handled server-side by OpenAI: your microphone audio is uploaded, transcribed, and immediately discarded — we never write the audio to disk. You always see the transcript and can edit it before submitting.

Voice is entirely optional — every question can be typed, and you can switch between voice and typing at any time.

Our prompts are designed to avoid generic praise. Feedback must quote your actual words, give a specific diagnosis, and propose a concrete stronger version. If a piece of feedback could apply to any candidate, we consider it a defect — not a feature.

We notice tab-switches and disable pasting to keep practice honest, and these feed an integrity score on your report. They are signals shown to you — not a webcam, not a hard block, and never used to silently fail you.

InterviewIQ is built to help you prepare before the real interview. It is explicitly not a tool for use during a live interview, and there is no real-time copilot, answer feed, or stealth overlay.

We believe the honest path — practising, getting evidence-backed feedback, and improving — is also the one that actually lands the job. Using AI to deceive a real interviewer is against the spirit (and terms) of InterviewIQ.

• AI can make mistakes, misread context, or score imperfectly.
• A high score is practice feedback, not a guarantee of a real-world outcome.
• Treat the report as coaching input, not a hiring decision or professional advice.
• Use your own judgment — especially on rewrites you choose to reuse.

You choose what to upload, you can end the interview at any time, and you can delete all your data afterward from the contact & data requests page.