What Is a Semi-Structured Interview? 2026 Guide
Qualitati Research Team · 2026-09-19 · 8 min read
Short answer: A semi-structured interview is a qualitative research interview that follows a prepared guide of open-ended questions but lets the interviewer change the order, skip what is already answered, and ask follow-up probes. It sits between a fixed questionnaire and an unstructured conversation: comparable across participants, yet flexible enough to follow what each person actually says. It is the default format in UX, market, and academic qualitative research.
If you have ever run a user interview with a discussion guide, you have probably run a semi-structured interview. This guide defines the method, compares it with structured and unstructured interviews, and explains what changes when an AI moderator conducts it. It includes an original scorecard for checking whether a guide is ready to field.
Last updated: September 19, 2026
Key takeaways
- A semi-structured interview combines a fixed set of core questions with flexible order and unscripted follow-up probes.
- The interview guide is the method's backbone. A widely cited review by Kallio et al. (2016) describes five phases for developing one, including pilot testing.
- Probing is what separates a semi-structured interview from a spoken survey. Without follow-ups, you collect short, surface-level answers.
- AI moderators can now run semi-structured interviews at scale, but published 2025–2026 studies show probing depth is the main weakness to check.
- Use it when you need comparable depth across a modest number of participants; avoid it when you need statistical estimates or pure open exploration.
What is a semi-structured interview?
A semi-structured interview is organized around a set of predetermined open-ended questions, with other questions emerging from the dialogue between interviewer and participant (DiCicco-Bloom & Crabtree, 2006, Medical Education). The researcher decides in advance which topics every participant must cover, but not the exact wording, order, or follow-ups used in each session.
Three elements define it:
- Core questions: open-ended questions tied to research objectives and asked of everyone.
- Probes: follow-ups such as "Can you walk me through the last time that happened?" that deepen an answer.
- Interviewer discretion: permission to reorder, skip, or explore an unexpected but relevant thread.
Structured vs semi-structured vs unstructured interviews
The three formats differ mainly in how much the interviewer can deviate from the script. The right choice depends on whether you need comparability, depth, or discovery.
| Criterion | Structured | Semi-structured | Unstructured |
|---|
| Question wording | Fixed | Core intent fixed, wording can adapt | Emergent |
| Question order | Fixed | Flexible | None planned |
| Follow-up probes | Rare or scripted | Expected | Constant |
| Comparability across participants | High | Moderate to high | Low |
| Depth | Low | High | High |
| Analysis | Often quantitative | Thematic or framework analysis | Narrative or grounded approaches |
| Typical use | Screening, standardized measures | User research, customer discovery, academic studies | Early exploration, life histories |
How to design a semi-structured interview guide
Kallio and colleagues reviewed methodological literature and identified five phases for building a guide: identifying the prerequisites for using semi-structured interviews, retrieving and using previous knowledge, formulating a preliminary guide, pilot testing it, and presenting the complete guide (Kallio et al., 2016, Journal of Advanced Nursing). In practice, that translates into four working steps.
1. Start from research questions, not interview questions
Write two to five research questions first. Each core interview question should map to at least one of them. A question that maps to none is a candidate for deletion.
2. Write open, concrete, single-issue questions
Ask about specific past behavior ("Tell me about the last time you exported a report") rather than hypotheticals or abstract opinions. Avoid leading wording and double-barreled questions.
3. Pre-write probes for every core question
Probes should not be left to chance. Draft two or three per question: an elaboration probe ("What happened next?"), a clarification probe ("What do you mean by 'messy'?"), and an example probe ("Can you give me an example?").
4. Pilot, then freeze the core
Run one or two pilot interviews. Fix confusing wording and timing, then freeze core questions for the rest of the study so findings stay comparable. Our discussion guide template for AI moderators walks through this in more detail.
Original framework: the Semi-Structured Interview Readiness Scorecard
Score each item 0 (no), 1 (partly), or 2 (yes) before fielding. A total of 12 or more out of 16 suggests the guide is ready; below 10, revise first.
| # | Check | What "yes" looks like |
|---|
| 1 | Objective mapping | Every core question maps to a named research question |
| 2 | Openness | No core question can be answered with yes or no |
| 3 | Concreteness | Questions ask about specific past episodes, not hypotheticals |
| 4 | Neutrality | No wording implies a preferred answer |
| 5 | Probe coverage | Each core question has at least two pre-written probes |
| 6 | Priority marking | Must-ask questions are marked, so a short session still covers them |
| 7 | Timing | A pilot confirmed the guide fits the planned session length |
| 8 | Consent and close | Opening covers consent and recording; closing invites anything missed |
This scorecard is a Qualitati planning heuristic, not a validated instrument. Treat the thresholds as a starting point and adapt them to your field's standards.
Semi-structured interviews with AI moderators
Bottom line: AI moderators can follow a semi-structured guide and reach far more participants, but you should verify that they actually probe rather than just move to the next question.
Recent research is encouraging on scale. In a study posted on June 18, 2026, Wuttke, Lang, Klamm, Würschinger, and Kreuter ran AI conversational interviews with 571 respondents and reported that transcripts surfaced considerations a standardized battery missed, while participants rated the AI interview at or above the standardized survey (Wuttke et al., 2026).
Research on probing is more cautious. A 17-participant study of AI-generated follow-up questions positioned the LLM as a complement to human judgment rather than a replacement (Zhang et al., 2025). A later study of a voice interview bot coded 428 conversation turns and found that deepening probes were rare by default (Zhang et al., 2026).
The practical implication: in an AI-moderated study, the probes you pre-write in the guide are not optional. They are the instructions that keep the interview semi-structured instead of a spoken survey.
Where Qualitati fits
Qualitati is an AI user research platform for running and analyzing qualitative studies. For semi-structured interviews, it supports:
- AI-moderated interviews in text and voice that follow your guide and ask follow-up questions, in 10 languages.
- Active Listener mode, where a human interviewer gets real-time prompts and section tracking while running the session themselves.
- ThemeLens thematic analysis across up to 100 transcripts, mapping codes to research questions with participant-anchored quotes.
- QDA Workspace for inductive and deductive coding and codebook generation.
Pricing is published per credit, and new accounts get 30 free credits with no credit card required. See pricing.
Limitations and trade-offs
- Interviewer effects: flexibility means two moderators, human or AI, may probe differently. Review a sample of transcripts for consistency.
- Comparability vs depth: the more sessions diverge, the harder cross-participant comparison becomes.
- Analysis workload: open answers need systematic coding; AI coding speeds this up but still needs human review of themes and quotes.
- Not for estimates: semi-structured interviews explain how and why; they do not measure prevalence.
Methodology note: guidance on AI moderation reflects published studies as of September 2026 and should be reviewed against your discipline's standards and ethics requirements.
Who this is for, and when not to use it
Use it for UX research, customer discovery, jobs-to-be-done work, and academic studies where you need depth and some comparability. Do not use it when you need population estimates (use a survey), when you know too little to write core questions (start unstructured), or when group dynamics matter (consider a focus group). For sample size, see our guide to data saturation.
FAQ
What is the difference between a semi-structured and a structured interview?
A structured interview asks the same questions in the same order with fixed wording. A semi-structured interview keeps the core topics fixed but allows flexible order and unscripted follow-up probes.
How many questions should a semi-structured interview have?
There is no fixed rule. Many hour-long guides use somewhere between six and twelve core questions plus probes; pilot testing is the best way to confirm fit.
How do you analyze semi-structured interviews?
Most teams use thematic analysis or framework analysis: transcribe, code, group codes into themes, and link themes back to research questions with supporting quotes.
Can AI conduct a semi-structured interview?
Yes. Studies from 2025–2026 show LLM interviewers can follow a guide at scale, but default probing is often shallow, so write explicit probes and review transcripts.
Is a semi-structured interview qualitative or quantitative?
It is a qualitative method. Some studies code answers into counts, but the format is designed to explain experiences and reasoning, not to estimate frequencies.
Conclusion
A semi-structured interview works because it balances structure and flexibility: fixed core questions for comparability, probes for depth. Whether a person or an AI moderator runs it, the quality of the guide, and especially its probes, determines the quality of the data. Start free with 30 credits and field your next semi-structured interview study with Qualitati.