Create a discussion guide for an unmoderated UX research session

intermediateClaude SonnetDesignUx Designux-researchdiscussion-guideunmoderated-testingusability-testingresearch-ops

Use case

Use this when running unmoderated remote tests where you can't probe in real time. Bad guides produce useless videos; good ones produce findings you can act on. Best for prototype validation, onboarding tests, and IA tree tests where you have 15–30 minutes of a participant's time and one shot to learn what you need.

The prompt

You are a senior UX researcher writing a discussion guide for an unmoderated study. The guide must work without a human moderator — every question and task must be self-explanatory and bias-resistant.

Context:
- Product / prototype: {{product}}
- Research goal (one sentence): {{goal}}
- Specific questions you want answered: {{questions}}
- Target participant: {{target_participant}}
- Platform: {{platform}}
- Session length budget (minutes): {{session_length}}

Produce the guide in this exact structure:

1. Screener (3–5 questions). Multiple choice where possible. Include at least one "trap" question that filters out participants who don't actually match the target.
2. Warm-up (2 questions, ~3 min). Get the participant talking and oriented — their context, not your product yet.
3. Tasks (2–4 tasks, scoped to fit the session length). For each task: a realistic scenario, the task instruction (action-oriented, no leading language), success criteria for the participant, and a "post-task probe" — 1–2 follow-up questions that work even when no one's listening.
4. Wrap-up (2–3 questions). Capture overall impressions, comparisons to current tools, and willingness to use.
5. Bias check. List 2–3 specific phrases or framings you avoided in this guide and why.

Rules:
- Never ask "Did you find that easy?" — ask "What did you do to figure that out?"
- Never describe what something is supposed to do before the task. Let the participant discover it.
- Avoid "How would you feel if..." — ask about past behavior, not hypothetical reactions.
- Tasks should describe outcomes ("Find a robot that needs maintenance"), not UI paths ("Click on Maintenance").

Output: the guide, structured exactly as above.

Variables

{{{{product}}}}Replace with your {{product}}
{{{{goal}}}}Replace with your {{goal}}
{{{{questions}}}}Replace with your {{questions}}
{{{{target_participant}}}}Replace with your {{target participant}}
{{{{platform}}}}Replace with your {{platform}}
{{{{session_length}}}}Replace with your {{session length}}

Example

Input

product: Cresta Health patient intake prototype (Figma) — replaces a paper form for new-patient appointments
goal: Validate that patients can complete intake on their own phone in the waiting room without staff help
questions: Can they finish in <8 minutes? Do they understand consent language? Do they trust giving insurance info on a phone they were handed?
target_participant: US adults 25–65 who've seen a new doctor in the past 12 months and are not in healthcare professionally
platform: UserTesting (mobile only)
session_length: 20

Output

1. Screener
- In the past 12 months, have you started seeing a new doctor or specialist? (Yes / No — exclude No)
- Do you or anyone in your immediate household work in healthcare, health tech, or medical insurance? (Yes / No — exclude Yes)
- When you last visited a new doctor's office, what did you fill out before being seen? (paper forms / a tablet they handed me / nothing / I don't remember — exclude "I don't remember")
- Which of these have you used on your phone in the past month? (banking app / health insurance app / DMV / food delivery — multi-select; trap: must select at least one)
- On a scale of 1–5, how comfortable are you entering personal information on a phone someone hands you in a waiting room? (any answer ok; we want a spread)

2. Warm-up (~3 min)
- Walk me through what happened the last time you visited a new doctor's office, from arriving to being called back. What was that like?
- When you've been handed forms or a tablet at a doctor's office, what's gone well? What's been frustrating?

3. Tasks (~14 min)

Task 1 — Start intake
Scenario: "You've just arrived at a new doctor's office. The receptionist hands you this phone and asks you to complete intake before you're seen. Begin."
Success: Participant reaches the first form field.
Probes: What did you expect to happen when you started? Anything surprise you on this first screen?

Task 2 — Add insurance
Scenario: "Continue through the intake. When you reach insurance, use this card." (Show participant a generated insurance card image with name, ID, group.)
Success: Insurance fields populated correctly.
Probes: Was there anything you weren't sure how to enter? Anything you felt unsure about sharing?

Task 3 — Consent
Scenario: "Keep going until you reach the consent step. Decide whether you'd sign and continue, or stop. Do whatever you'd actually do."
Success: Participant makes an explicit choice to sign or decline.
Probes: In your own words, what were you agreeing to? Was anything in that section unclear or concerning?

4. Wrap-up
- How did this compare to filling out forms at doctors' offices you've been to recently?
- If your own doctor's office switched to this tomorrow, what would you want changed first?
- Was there a moment in this you wanted to stop or hand the phone back? When?

5. Bias check
- Avoided "Was that easy?" throughout — replaced with "What did you do to figure that out?" which captures behavior even when participants are being polite.
- Avoided describing the prototype as "new" or "improved" — participants praise novelty by default.
- Avoided "Would you use this?" in wrap-up — replaced with a behavioral comparison to their actual recent experiences, which is far more predictive.

Tips for best results

1Read every task instruction out loud. If you can hear yourself leading the participant, rewrite it.
2Trap screener questions catch 10–20% of unqualified participants on most platforms. Always include one.
3Post-task probes work because participants narrate while completing them. You'll hear what was hard even if they say it was easy.
4Cap unmoderated sessions at 20 minutes. Quality drops sharply after that even with engaged participants.
5After fielding, watch the first 3 sessions before launching to all 15. You'll catch a broken task or unclear instruction before it ruins the dataset.

Related prompts

Rewrite usability test findings into action-prioritized recommendations

advanced

Take raw usability test notes (your own or a researcher's) and turn them into a prioritized recommendation list ranked by impact, effort, and confidence — not a wall of observations.

Designux-researchusability-testingsynthesis

Synthesize card-sort results into IA recommendations

advanced

Turn raw card-sort data (open or closed) into a prioritized set of information architecture recommendations with confidence levels and the trade-offs you're choosing between.

Designux-researchinformation-architecturecard-sort

Run a Nielsen heuristic evaluation on a screen

intermediate

Apply Nielsen's 10 heuristics to a screen description or screenshot description and produce a prioritized issue list with severity ratings and concrete fixes.

Designuxusabilityheuristic-evaluation

Need help implementing this prompt in your workflow?

Book a call

Use case

You are a senior UX researcher writing a discussion guide for an unmoderated study. The guide must work without a human moderator — every question and task must be self-explanatory and bias-resistant. Context: - Product / prototype: {{product}} - Research goal (one sentence): {{goal}} - Specific questions you want answered: {{questions}} - Target participant: {{target_participant}} - Platform: {{platform}} - Session length budget (minutes): {{session_length}} Produce the guide in this exact structure: 1. Screener (3–5 questions). Multiple choice where possible. Include at least one "trap" question that filters out participants who don't actually match the target. 2. Warm-up (2 questions, ~3 min). Get the participant talking and oriented — their context, not your product yet. 3. Tasks (2–4 tasks, scoped to fit the session length). For each task: a realistic scenario, the task instruction (action-oriented, no leading language), success criteria for the participant, and a "post-task probe" — 1–2 follow-up questions that work even when no one's listening. 4. Wrap-up (2–3 questions). Capture overall impressions, comparisons to current tools, and willingness to use. 5. Bias check. List 2–3 specific phrases or framings you avoided in this guide and why. Rules: - Never ask "Did you find that easy?" — ask "What did you do to figure that out?" - Never describe what something is supposed to do before the task. Let the participant discover it. - Avoid "How would you feel if..." — ask about past behavior, not hypothetical reactions. - Tasks should describe outcomes ("Find a robot that needs maintenance"), not UI paths ("Click on Maintenance"). Output: the guide, structured exactly as above.

Variables

{{{{product}}}}Replace with your {{product}}

{{{{goal}}}}Replace with your {{goal}}

{{{{questions}}}}Replace with your {{questions}}

{{{{target_participant}}}}Replace with your {{target participant}}

{{{{platform}}}}Replace with your {{platform}}

{{{{session_length}}}}Replace with your {{session length}}

Example

Input

product: Cresta Health patient intake prototype (Figma) — replaces a paper form for new-patient appointments
goal: Validate that patients can complete intake on their own phone in the waiting room without staff help
questions: Can they finish in <8 minutes? Do they understand consent language? Do they trust giving insurance info on a phone they were handed?
target_participant: US adults 25–65 who've seen a new doctor in the past 12 months and are not in healthcare professionally
platform: UserTesting (mobile only)
session_length: 20

Output

3. Tasks (~14 min)

Tips for best results

1Read every task instruction out loud. If you can hear yourself leading the participant, rewrite it.

2Trap screener questions catch 10–20% of unqualified participants on most platforms. Always include one.

3Post-task probes work because participants narrate while completing them. You'll hear what was hard even if they say it was easy.

4Cap unmoderated sessions at 20 minutes. Quality drops sharply after that even with engaged participants.

5After fielding, watch the first 3 sessions before launching to all 15. You'll catch a broken task or unclear instruction before it ruins the dataset.