# AI-moderated interview discussion guide worksheet

Prepared by User Intuition. Editable planning example, updated 19 September 2026.
Companion: https://www.userintuition.ai/posts/ai-moderated-interview-discussion-guide/

This is a planning document to adapt, not an automatic configuration import or a guarantee of interview quality.

## Decision and audience

- Decision to inform: [Fill]
- Decision owner: [Fill]
- Research question: [Fill]
- Relevant recent participant experience: [Fill]
- Eligibility, sample source, and exclusions: [Fill]
- Language, modality, accessibility, and stimulus requirements: [Fill]
- Consent and recording arrangements: [Fill]
- Guide version and date: [Fill]

## Opening

Explain the study purpose, what the participant will do, and how they can decline a question or end the session.
Warm-up prompt: [A short question establishing relevant context]

## Topic plan

Duplicate this block for each priority topic. Remove low-priority topics if the pilot runs long.

- Topic: [Fill]
- Priority: [Required / useful / optional]
- Evidence needed: [A specific event, sequence, comparison, or trade-off]
- Opening prompt: [One neutral question anchored in experience]
- Follow-up direction: [What to clarify if the participant raises it]
- Conditions for branching: [What in the answer would justify another line of questioning?]
- Stopping condition: [Answered sufficiently, no relevant experience, uncertainty, or discomfort]
- Planning time: [Estimate; verify in pilot]
- What this topic cannot establish: [Fill]

## Worked planning example

Decision: Which part of an urgent replacement purchase should we investigate next?
Topic: Purchase sequence.
Evidence needed: Trigger, product choice, seller choice, and experience after arrival.
Opening: “Think about your last replacement purchase. What prompted it?”
Follow-up direction: Explore one decision at a time. Ask what alternatives were considered, if any.
Avoid: Assuming the participant compared sellers, disliked the old product, or chose on price.
Stop: When the sequence is clear or the participant cannot recall more.
Limit: A retrospective account does not prove why the whole market chooses a seller.

## Moderation boundaries

- Ask one clear question at a time in voice, chat, and video.
- Build follow-ups from the participant's actual answer.
- Do not repeat a question simply to reach a target number of probe levels.
- Do not treat a moderator suggestion as a participant's unprompted thought.
- Do not assume emotion, certainty, dissatisfaction, or a comparison.
- Allow missing experience and uncertainty to remain unresolved.
- For sensitive material, respect refusal and the study's agreed safeguards.

## Pilot review log

For each pilot interview:
- Participant ID / guide version: [Fill]
- Source locations reviewed: [Fill]
- Could the participant answer the research question? [Evidence]
- Was a useful concrete example obtained? [Evidence]
- Any leading, repetitive, or unsupported probe? [Exact location]
- Missing topic or evidence gap: [Fill]
- Duration and signs of fatigue: [Fill]
- Proposed wording change and reason: [Fill]
- What would show the change worked? [Fill]
- Reviewer and decision to continue, revise, or stop: [Fill]

Do not label revised prompts as questions asked in a completed interview.
Keep the old version and record which participants received it before comparing waves.

Inspect the actual W30 phone-replacement conversation used in the companion article:
https://www.userintuition.ai/preview/#interviews
Inspect the example downstream presentation, including its limitations:
https://www.userintuition.ai/preview/#deck