How to Write User Interview Questions (2026 Guide)
Qualitati Research Team · 2026-07-21 · 11 min read
Last updated: July 21, 2026
Short answer
User interview questions should be open-ended, neutral, and grounded in past behavior rather than hypothetical opinion. Write questions that start with "tell me about a time…" or "walk me through…", avoid yes/no and leading phrasing, and treat the guide as a flexible outline you probe against — not a script you read aloud. A good guide has 5–8 core questions with follow-up probes, sequenced from broad to specific.
Key takeaways
- The most common failure in user research is not sample size — it is leading questions that plant the answer inside the question.
- Ask about concrete past behavior ("the last time you…"), not predicted future behavior ("would you use…"). People are poor forecasters of their own actions.
- Every core question needs planned probes. The insight almost always lives in the follow-up, not the first answer.
- Sequence matters: open broad, narrow gradually, save sensitive or specific questions for after rapport is built.
- Use the Interview Question Quality Rubric below to score each question before you field it.
- AI-moderated interviewing enforces good probing at scale — but only if the underlying question design is sound. Bad questions asked consistently are still bad questions.
Why question wording decides your data quality
In interview-based research, the question is the instrument. A biased survey item contaminates one column; a leading interview question contaminates the entire conversation, because the participant now knows what you want to hear and helpfully supplies it. The Nielsen Norman Group is blunt on this point: leading questions "plant ideas" and produce data that confirms what the researcher already believed, per its guidance on avoiding leading questions (NN/g).
This matters more now, not less. As AI-moderated interviews make it cheap to run dozens of sessions in parallel, the leverage of your question design multiplies. Field a leading question to 60 participants and you have simply automated your own confirmation bias. The discipline of writing neutral, behavior-anchored questions is the highest-return skill in qualitative research.
The five rules of a good interview question
1. Open-ended, not yes/no
Closed questions ("Did you find that useful?") collapse a rich experience into a single word and hand the interpretation to the participant. Open questions ("What did you make of that feature?") return the experience in the participant's own frame. Rephrase every direct question into an invitation to describe, as Maze recommends in its guide to avoiding leading questions.
2. Behavior over opinion
"Would you pay for this?" measures politeness, not willingness to pay. "Walk me through the last time you paid for a tool like this" measures a real decision with real stakes. Anchor questions in specific, recent, remembered episodes. The past is data; the hypothetical future is speculation.
3. Neutral framing
A leading question smuggles in an assumption or a valence. "How frustrating was the checkout?" presumes frustration. "Tell me about your checkout experience" does not. Strip evaluative adjectives ("easy", "annoying", "helpful") out of the question and let the participant supply the valence.
4. One idea per question
Double-barrelled questions ("How was onboarding and did it help you get set up?") force the participant to pick which half to answer, and you never know which. Split them.
5. Probe-ready
The first answer is a headline. The probe is the article. Plan at least one follow-up per core question: "What do you mean by that?", "Can you give me an example?", "What happened next?" NN/g and UserTesting both stress that the moderator's job is to listen more than talk and follow the participant's lead.
Question types to prefer — and to avoid
| Instead of… | Ask… | Why |
|---|
| "Do you like the dashboard?" | "Walk me through what you do when you open the dashboard." | Behavior, not approval-seeking |
| "Would you use a feature that…?" | "Tell me about the last time you needed to do that." | Real episode, not forecast |
| "Was the process confusing?" | "Describe how the process went for you." | Removes the planted valence |
| "Don't you think X is better?" | "How do X and Y compare for your work?" | Neutral, no steer |
| "How often and why do you use it?" | Two separate questions | Single idea per question |
Original asset: the Interview Question Quality Rubric
Score every question in your guide on these six dimensions before fielding. Any question scoring "No" on rows 1–3 should be rewritten. This is a Qualitati-owned checklist you can copy into your guide doc.
| # | Check | Pass condition |
|---|
| 1 | Open | Cannot be answered with yes/no or a single word |
| 2 | Neutral | Contains no assumption, valence, or preferred answer |
| 3 | Single | Asks exactly one thing |
| 4 | Behavioral | Points to a real past episode, not a hypothetical |
| 5 | Probe-ready | Has at least one planned follow-up |
| 6 | On-question | Maps to a stated research question, not just curiosity |
A reusable interview guide template
Structure a 30–45 minute interview in five moves. This sequence builds rapport before depth and keeps the funnel broad-to-narrow.
- Warm-up (2–3 min). Low-stakes context. "To start, tell me a bit about your role and a typical week."
- Grand-tour question (5 min). One broad, behavioral opener. "Walk me through the last time you [did the target activity]."
- Core questions (20–25 min). Five to eight questions from the rubric, each with probes. Follow the participant, not the order on the page.
- Specific / sensitive (5 min). Pricing, frustrations, comparisons — now that rapport exists.
- Wrap-up (2 min). "Is there anything I should have asked but didn't?" — consistently one of the highest-yield questions in the guide.
Where Qualitati fits
Qualitati is an AI user research platform for product, UX, and customer-insights teams. Its AI moderator conducts interviews in text or voice, and is built to enforce exactly the discipline above: it asks open, neutral questions and — critically — probes automatically when an answer is thin, following up rather than moving on. That closes the single biggest gap between a good guide on paper and a good interview in practice.
For teams that want a human in the chair, Active Listener mode feeds the interviewer real-time probe suggestions and section tracking, so the person moderating stays neutral and never forgets a follow-up. When the interviews are done, ThemeLens and the QDA Workspace turn transcripts into coded, quote-anchored themes. Question design is where research quality is won or lost; Qualitati is designed so that a well-written guide is executed the same way in interview 1 and interview 60.
Limitations and trade-offs
No question guide eliminates bias entirely. Three honest caveats:
- Recall is imperfect. Behavioral questions depend on memory, which decays and reconstructs. Recent episodes are more reliable than distant ones.
- Neutral is not the same as directionless. An over-open guide can drift and never reach the research question. Neutrality is about wording; focus is about structure. You need both.
- AI moderation follows its design. An AI interviewer probes consistently, but it will probe a badly designed question just as consistently. Automation raises the floor on execution; it does not fix the guide. Sensitive topics still warrant human review of the moderation behavior. Human review is recommended for any guide touching health, finances, or vulnerable populations.
Frequently asked questions
How many questions should a user interview have?
Aim for 5–8 core questions for a 30–45 minute session, each with planned probes. Fewer, deeper questions beat a long list read at speed — the probes are where the insight is.
What is a leading question in user research?
A leading question contains an assumption or preferred answer that steers the participant, such as "How frustrating was the setup?" It contaminates the response because the participant now knows what you expect. Rephrase to neutral: "Tell me about the setup."
Should I ask users what they want?
Ask about problems and past behavior, not desired solutions. Users reliably report what frustrates them but poorly predict which features they will actually use. Translate wishes back into the underlying need.
Open-ended vs. closed questions — which is better?
Open-ended questions are the backbone of interviews because they surface the participant's own framing. Closed questions have a narrow role for confirming specific facts, but should never carry the main line of inquiry.
Can AI write good interview questions?
AI can draft and pressure-test questions against rules like the rubric above, and AI moderators can execute a guide consistently with reliable probing. Final judgment on framing and research fit should stay with the researcher.
Bottom line
Great user research is mostly great questions asked well. Write open, neutral, single-idea questions anchored in real behavior, plan your probes, and sequence broad-to-narrow. Score each question with the rubric before you field it. Then execute the guide the same way every time — which is exactly what AI-moderated interviewing is for.
Start free with 30 credits — no credit card required. Create an account, build a guide, and run an AI-moderated interview that probes like your best moderator. Or view transparent pricing first.