Research · Market & customer research
The Best Prompt for a User Interview Guide
The best prompt for an interview guide starts from what the team needs to learn and bans the questions that produce unreliable answers. This one converts your goal into research questions, then writes a timed guide of open, past-behavior questions with follow-up probes, a consent script and a screener. A coverage table shows which question tests which team belief, so you can see gaps before the first session.
User Interview Discussion Guide Prompt
Writes a timed discussion guide for customer or user interviews: screener, opening script, behavior-based questions with probes, and a table tying each question to a learning goal.
- Works with
- ChatGPT, Claude, Gemini
- Difficulty
- Intermediate
- Complexity
- Multi-part reply
- Variables
- 6
- Version
- 1.0
Customize this prompt
Fill in what you know. Anything left blank stays as a [bracketed placeholder]. Nothing you type leaves your browser.
6 variables · fields marked * matter most
# Role You are a user researcher writing a discussion guide for one-to-one interviews. Your questions get people describing what they have done, not predicting what they might do. # Study - Product or area: [Product or area] - What we need to learn: [What you need to learn] - Who we are interviewing: [Who you are interviewing] - Type of study: [Type of study] - Session length: [Session length] - Beliefs the team holds that the interviews should test: [Team beliefs to test] # Build the guide in this order 1. Learning goals - turn what we need to learn into three to five research questions. These are for the team and are never asked aloud. 2. Screener - four to six criteria or questions for recruiting the right participants, including one that excludes people without recent, relevant experience. 3. Opening script - a short spoken introduction: purpose, how long it takes, recording and consent, that there are no right answers and we are not testing them, and that they can skip any question or stop. 4. Warm-up - two or three easy questions about their role and context. 5. Core sections - one section per learning goal, with a time allocation. In each: a main question that asks for a specific recent occasion, then three or four follow-up probes (what happened next, what they used, what was hard, what it cost them in time or money). 6. Reaction section - only if the study type involves showing a concept or product: what to show, the neutral instruction to give, and what to observe before asking anything. 7. Wrap-up - a question on what we should have asked, permission for follow-up, and thanks. 8. Coverage table - every question mapped to the learning goal and team belief it informs. Flag any belief that no question tests. # Question rules - Open questions only in the core sections; no yes/no openers. - Ask about past behavior and concrete examples. No "would you", "how likely" or "do you think you might". - One idea per question. - Do not name our solution, its features or its benefits before the reaction section. - No leading wording: avoid adjectives that suggest the answer (easy, frustrating, useful). - Mark each question as must-ask or if-time. Must-ask questions have to fit within 70% of the session length; state the running time. # Output format A guide the interviewer can read from: numbered headings as above, questions written word for word, probes indented beneath them, timings in brackets. Finish with a five-line note-taking template and three mistakes the interviewer should avoid in this particular study.
Use it when you are
- Planning problem-discovery interviews before deciding what to build
- Preparing churn or win-loss conversations with former or prospective customers
- Getting a founder or product manager without research training ready for their first interviews
- Aligning a team on what a round of interviews is meant to answer
- Reviewing an existing guide for leading or hypothetical questions
Why this prompt works
Research questions and interview questions are separated
Teams often ask participants the thing they want to know ("would you pay for this?"). The prompt first writes learning goals that stay with the team, then derives spoken questions from them. That separation is standard research practice and keeps the guide from turning into a survey read aloud.
Past behavior replaces prediction
People are unreliable forecasters of their own behavior and tend to be agreeable in interviews. Anchoring each core question on a specific recent occasion, and banning "would you" phrasing, draws out what happened, what tools were used and what it cost, which is evidence a team can act on.
Your own solution is held back
Mentioning the product early frames everything that follows and invites polite approval. The rule against naming the solution or its benefits before the reaction section protects the discovery part of the session from that bias, and it keeps any reaction to a concept for the end, after the participant has described their own experience.
Time is budgeted, with must-ask questions capped
Guides are usually too long, and the last learning goal gets squeezed. Allocating minutes per section and fitting must-ask questions into 70% of the session leaves room for the unplanned follow-ups where the most useful detail tends to surface, and for participants who talk at length.
The coverage table exposes untested beliefs
Mapping every question to a goal and a team belief shows two problems quickly: questions that serve no goal, and beliefs that nothing in the guide would confirm or contradict. Both are cheaper to fix on paper than after five sessions.
How to use it
- Write the research goal as the uncertainty you want resolved, and list the beliefs your team currently holds, including the uncomfortable ones.
- Describe the participant with a behavior and a time frame, not only a job title.
- Run the prompt, then read the coverage table first. Add or cut questions until every belief is tested and every question earns its place.
- Read the guide aloud and time it. Trim wording that sounds stiff when spoken.
- Pilot it with one colleague or friendly participant, then revise the questions that produced short answers.
- Treat the guide as a map during sessions. Follow interesting answers with the probes instead of reading every question in order.
AI model compatibility
| Model | Status | Notes |
|---|---|---|
| ChatGPT | Expected to work | After the first draft, ask it to act as the participant described and answer the must-ask questions briefly. Questions that produce one-word answers need rewriting. |
| Claude | Expected to work | If you have notes from earlier interviews, paste them below the Study section inside <previous_notes> tags and ask the guide to avoid ground already covered. |
| Gemini | Expected to work | Keep the question rules as a list. If a hypothetical question appears in the draft, quote the rule back and ask for a rewrite of that question only. |
“Expected to work” means the prompt uses plain, model-independent structure and has been reviewed for that — not that it has been run on that model. A model is marked Tested only when a test record is published below.
Worked example
Scenario: A two-person startup planning ten discovery interviews about late invoice payments.
# Role You are a user researcher writing a discussion guide for one-to-one interviews. Your questions get people describing what they have done, not predicting what they might do. # Study - Product or area: Invoicing software for freelance designers - What we need to learn: How freelancers currently chase late payments, how much time it takes, and what they have tried to fix it - Who we are interviewing: Freelance designers who sent at least five invoices in the last three months and had at least one paid late - Type of study: Problem discovery - Session length: 30 minutes - Beliefs the team holds that the interviews should test: We believe most late payment happens because clients lose the invoice email, and that freelancers find chasing awkward enough to pay for automation. # Build the guide in this order 1. Learning goals - turn what we need to learn into three to five research questions. These are for the team and are never asked aloud. 2. Screener - four to six criteria or questions for recruiting the right participants, including one that excludes people without recent, relevant experience. 3. Opening script - a short spoken introduction: purpose, how long it takes, recording and consent, that there are no right answers and we are not testing them, and that they can skip any question or stop. 4. Warm-up - two or three easy questions about their role and context. 5. Core sections - one section per learning goal, with a time allocation. In each: a main question that asks for a specific recent occasion, then three or four follow-up probes (what happened next, what they used, what was hard, what it cost them in time or money). 6. Reaction section - only if the study type involves showing a concept or product: what to show, the neutral instruction to give, and what to observe before asking anything. 7. Wrap-up - a question on what we should have asked, permission for follow-up, and thanks. 8. Coverage table - every question mapped to the learning goal and team belief it informs. Flag any belief that no question tests. # Question rules - Open questions only in the core sections; no yes/no openers. - Ask about past behavior and concrete examples. No "would you", "how likely" or "do you think you might". - One idea per question. - Do not name our solution, its features or its benefits before the reaction section. - No leading wording: avoid adjectives that suggest the answer (easy, frustrating, useful). - Mark each question as must-ask or if-time. Must-ask questions have to fit within 70% of the session length; state the running time. # Output format A guide the interviewer can read from: numbered headings as above, questions written word for word, probes indented beneath them, timings in brackets. Finish with a five-line note-taking template and three mistakes the interviewer should avoid in this particular study.
What a good response should contain
- Three to five learning goals, such as what triggers a freelancer to start chasing and what they do first
- A screener that excludes people with no recent late payment
- Core questions that open with a specific occasion, for example asking about the last invoice that was paid late, each followed by probes on steps taken, tools used and time spent
- No reaction section, because the study type is problem discovery, and no mention of automation in the questions
- A coverage table showing which questions test the lost-email belief and the willingness-to-pay belief, with running time for must-ask questions at or under 21 minutes
This is an illustrative scenario showing how to fill in the prompt and what to look for. It is not a recorded model output. Recorded outputs will be published here once this prompt has a test record.
Testing and editorial review
Status: Untested. Reviewed for structure and clarity by the editorial team. No recorded model test yet.
No test record has been published for this prompt yet. When it is tested, this section will show the model and version, the date, the inputs, the actual output and notes against our seven-point rubric. We do not publish numerical scores.
| Author | TheBestPromptFor Editorial Team — The editorial team writes, reviews and maintains the prompts and guides on TheBestPromptFor. Prompts are drafted with AI assistance and reviewed against published editorial standards. |
|---|---|
| Published | November 15, 2026 |
| Last reviewed | November 15, 2026 |
| Last tested | Not yet tested |
| Version history |
|
Limitations and things to check
- The model does not know your users. Expect to adjust vocabulary so questions use the words participants use.
- A good guide does not make a good interview. Listening, silence and follow-up are skills the prompt cannot supply.
- The consent script is a starting point. Check it against your organization's privacy policy and the law where your participants live.
- Interviews show what people do and why; they do not tell you how common a behavior is. Pair them with quantitative data for that.
Alternative prompts
Marketing
Evidence-Based Customer Persona Prompt
Builds personas from the interviews, reviews or survey answers you paste in, tagging every line as evidence, inference or assumption.
- ChatGPT
- Claude
- Gemini
Research
Desk Market Research Brief Prompt
Produces a desk-research brief on a market, with segments, a range-based size estimate, competitors and pricing, and every figure labeled as sourced, provided or estimated.
- Perplexity
- ChatGPT
- Claude
- Gemini
Web Design
UX Audit Prompt
Runs a heuristic review of one user flow: a step-by-step walkthrough, findings tied to usability heuristics with severity ratings, and fixes.
- ChatGPT
- Claude
- Gemini
E-commerce
Customer Review Analysis Prompt
Turns pasted reviews into counted themes with verbatim quotes, causes of complaints, customer language and recommendations graded by support.
- ChatGPT
- Claude
- Gemini
Related resources
Guide
How to Write a Good AI Prompt
The five parts of a complete prompt (goal, context, task, constraints, output format), with a worked rewrite and a pre-send checklist.
Guide
How to Improve AI Responses
A diagnostic approach to weak AI answers: name the fault, trace it to the gap in your prompt, then follow up precisely or rewrite and start again.
Frequently asked questions
How many questions should a 30-minute interview have?
Fewer than most people expect. With an introduction, warm-up and wrap-up, a 30-minute session has room for roughly three core topics, each with one main question and a handful of probes. The prompt enforces this by capping must-ask questions at 70% of the session.
Why can't I ask whether they would use my product?
Because the answer tells you little. People asked to predict their behavior, face to face with someone who built the thing, tend to say yes. Asking what they did the last time the problem came up, and what they have already paid or tried, gives stronger evidence of demand.
Can I use this for usability testing?
Only partly. Choosing concept or prototype feedback adds a reaction section, but a full usability test needs tasks, success criteria and observation notes, not a conversation guide. For reviewing an interface against established principles without participants, the UX audit prompt is a better fit.
Sources and references
- Prompt engineering overview — Anthropic
- Prompt design strategies — Google AI for Developers
Was this prompt useful?
No account needed. Do not include personal information.