An AI moderated interview is only as good as the guide behind it.
The AI moderator can ask questions, listen to responses, probe vague answers, and summarize themes. But it still needs clear direction from the researcher. If the guide is too broad, biased, repetitive, or vague, the interview will produce shallow results.
That is the biggest mistake teams make with AI moderated interviews.
They treat the guide like a survey script.
A good AI moderated interview guide does more than list questions. It tells the AI moderator what the study is trying to learn, which answers matter, when to ask follow-up questions, when to move on, and what not to do.
In human moderated research, an experienced researcher carries much of that judgment in their head. In AI moderated research, more of that judgment needs to be written into the guide.
This article explains how to write an AI moderated interview guide that produces better responses, better follow-ups, and better research insights.
Read the full category guide here: AI Moderated Interviews: What They Are, How They Work, and When to Use Them in Research
What Is an AI Moderated Interview Guide?
An AI moderated interview guide is a structured set of instructions that tells an AI moderator how to conduct a research interview.
It usually includes:
- The research objective
- The participant profile
- The interview context
- The main questions
- Follow-up instructions
- Probing rules
- Guardrails
- Topics to avoid
- Closing questions
- Analysis goals
The guide gives the AI moderator enough context to run a focused conversation.
For example, a weak guide might say:
Ask users why they did not activate.
A better guide would say:
Understand why trial users did not complete activation within seven days. Ask about their original goal, what they expected after signing up, where they got stuck, what alternatives they considered, and what would have made them more likely to continue. If the participant gives a vague answer such as “confusing” or “not useful,” ask for a specific moment or example.
The second version is better because it gives the AI moderator direction. It explains the goal, the audience, the themes to explore, and how to probe.
Why AI Moderated Interview Guides Need More Than Questions
A human moderator can adapt naturally during an interview.
If a participant says something vague, the moderator can ask for detail. If the participant contradicts themselves, the moderator can explore the contradiction. If an answer is not relevant, the moderator can move on.
An AI moderator can do some of this, but it needs clear instructions.
That is why an AI moderated interview guide should include:
- What the AI should ask
- Why each topic matters
- What kind of answer is useful
- What counts as a vague answer
- When to ask for an example
- When to avoid leading the participant
- When to stop probing and move forward
Think of the guide as a combination of an interview script, moderator training document, and analysis brief.
The Basic Structure of an AI Moderated Interview Guide
A strong AI moderated interview guide usually has eight parts.
- Research goal
- Participant profile
- Opening message
- Main interview questions
- Follow-up instructions
- Guardrails
- Closing question
- Analysis instructions
Each part plays a different role.
1. Start With a Clear Research Goal
The research goal is the most important part of the guide.
If the goal is unclear, the interview will feel unfocused. The AI may ask too many broad questions, probe the wrong details, or collect responses that are difficult to interpret.
A good research goal should answer:
- What are we trying to learn?
- Who are we trying to learn from?
- What decision will this research support?
- What kind of insight would be useful?
Weak research goal
Understand user feedback.
This is too broad. It does not say which users, what feedback, or what decision the team needs to make.
Better research goal
Understand why trial users who signed up in the last 30 days did not complete onboarding, where they got stuck, and what would have made them more likely to activate.
This is better because it is specific.
It defines:
- The audience: trial users who signed up in the last 30 days
- The behavior: did not complete onboarding
- The research focus: where they got stuck
- The decision area: improving activation
Strong research goal format
Use this formula:
Understand why [participant segment] [did or did not do something], so we can [decision or action].
Examples:
Understand why new users abandon onboarding before inviting teammates, so we can improve activation.
Understand why buyers chose a competitor, so we can improve positioning and sales enablement.
Understand how customers describe their current workflow, so we can improve product messaging.
Understand how users react to a new feature concept, so we can decide whether to prioritize it.
2. Define the Participant Profile
The AI moderator needs to know who it is speaking to.
A participant profile helps the AI understand the context of the interview. It also helps the researcher make sure the study is targeted.
Include details such as:
- User type
- Role
- Experience level
- Customer status
- Product usage
- Recency of behavior
- Segment or persona
- Relevant screening criteria
Example participant profile
Participants are SaaS trial users who signed up within the last 30 days but did not complete onboarding. They may have explored the product briefly, invited no teammates, and not reached the activation milestone.
This gives the AI useful context.
It also prevents the interview from becoming too generic.
3. Write a Clear Opening Message
The opening message sets expectations for the participant.
It should explain:
- Why they are being interviewed
- What the topic is
- How long it will take
- That there are no right or wrong answers
- That honest feedback is welcome
- That they should be specific
Example opening message
Thank you for taking the time to share your feedback. This interview is about your experience during the trial. There are no right or wrong answers. Please be as specific as possible. Your feedback will help us understand what worked, what did not, and what we can improve.
The opening should feel simple and human.
Avoid overexplaining the technology. The participant does not need a technical description of the AI. They need to know what to expect and how to respond.
4. Choose the Right Main Questions
Main questions are the backbone of the interview.
They should be open-ended, neutral, and focused on real experiences.
Good interview questions invite stories.
Bad interview questions invite short answers, opinions without context, or biased responses.
Weak question
Did you like the onboarding process?
This usually produces a short answer: yes, no, or kind of.
Better question
Can you walk me through what happened the first time you tried to set up the product?
This invites a story. It helps the participant describe what they actually did.
Weak question
Was pricing too expensive?
This is leading. It pushes the participant toward price as the issue.
Better question
How did you evaluate whether the pricing made sense for you?
This gives the participant room to explain their thinking.
Weak question
What features do you want?
This may produce a wishlist without context.
Better question
What were you trying to accomplish that the product did not help you do?
This connects the answer to a real need.
5. Organize Questions in a Natural Flow
A good interview guide should feel like a conversation, not a form.
Use this structure:
- Warm-up
- Context
- Core experience
- Friction or decision points
- Alternatives
- Improvement
- Closing reflection
Example flow for churn research
Warm-up
What were you hoping to accomplish when you first signed up?
Context
Can you walk me through what you did after signing up?
Core experience
What parts of the experience felt useful?
Friction
Was there any point where you felt stuck, confused, or unsure what to do next?
Alternatives
Did you consider any alternatives? If yes, which ones?
Decision
What was the main reason you did not continue?
Improvement
What would have made you more likely to continue?
Closing
Is there anything else about your experience that we should have asked about but did not?
This flow starts broad, moves into specifics, and ends with reflection.
6. Add Probing Instructions
This is where AI moderated interview guides become different from standard interview guides.
Do not just write the main questions. Tell the AI when and how to probe.
A probe is a follow-up question that asks the participant to explain, clarify, or give an example.
Common answers that need probing
The AI should probe when participants say things like:
- “It was confusing.”
- “It was expensive.”
- “I did not see the value.”
- “It was hard to use.”
- “The design felt weird.”
- “I did not trust it.”
- “It took too long.”
- “I would maybe use it.”
- “It was not for me.”
- “I liked the idea, but…”
These are useful signals, but they are not complete insights.
Example probing instructions
If the participant says something was confusing, ask which step was confusing and what they expected to happen.
If the participant says something was expensive, ask what they were comparing the price against.
If the participant says they did not see value, ask what value they expected and when they expected to see it.
If the participant mentions a competitor, ask what they preferred about that competitor.
If the participant gives a short answer, ask them to describe a specific moment or example.
If the participant says they would not use the product, ask what would need to change for them to consider it.
These instructions help the AI avoid generic follow-ups like:
Tell me more.
A better AI follow-up connects to the participant’s actual answer.
7. Define What a Good Answer Looks Like
The AI moderator should know what kind of response is useful.
For example, if your study is about onboarding, a useful answer may include:
- The step where the participant got stuck
- What they expected to happen
- What actually happened
- Whether they tried to solve it
- Whether they asked someone for help
- What they did next
You can write this directly into the guide.
Example
For onboarding questions, useful answers should include specific steps, moments of confusion, expectations, actions taken, and what happened next. If the participant only gives a general reaction, ask for a specific example.
This improves the quality of the conversation.
The AI does not need to guess what the researcher wants. The guide tells it.
8. Add Guardrails
Guardrails tell the AI moderator what not to do.
This is important because AI moderation can accidentally create biased or low-quality research if the instructions are too loose.
Useful guardrails include:
- Do not lead the participant
- Do not suggest answers
- Do not assume the participant had a problem
- Do not over-probe the same topic
- Do not ask multiple questions at once
- Do not make the participant feel judged
- Do not argue with the participant
- Do not ask for sensitive personal information unless required and approved
- Do not continue probing if the participant seems uncomfortable
- Do not turn the interview into customer support
Example guardrail section
Keep the tone neutral and curious. Do not suggest that the participant should have liked or disliked anything. Do not ask leading questions. Do not assume the participant experienced a problem. Ask one question at a time. If the participant does not want to answer a question, move on.
This makes the interview safer and more reliable.
9. Add Topic-Specific Instructions
Some studies require special instructions.
For example, a win/loss interview should ask about alternatives, decision criteria, objections, and buying process.
A churn interview should ask about expectations, friction, value, alternatives, and cancellation triggers.
A concept test should ask about clarity, relevance, believability, differentiation, and willingness to use.
Example: product feedback instructions
Focus on what the participant was trying to accomplish, what felt useful, what felt difficult, and what they expected the product to do. Avoid asking whether they “liked” the product without asking for context.
Example: win/loss instructions
Explore the participant’s buying context, alternatives considered, decision criteria, objections, and final reason for choosing or rejecting the product. If a competitor is mentioned, ask what made that competitor more or less appealing.
Example: message testing instructions
Ask the participant what they think the message means, who they think it is for, what feels clear, what feels confusing, and what feels believable or exaggerated.
Topic-specific instructions help the AI stay aligned with the research goal.
10. Keep the Guide Focused
One of the easiest ways to ruin an AI moderated interview is to ask too many questions.
A guide with 25 questions may look thorough, but it often creates shallow answers.
For most AI moderated interviews, aim for:
- 5 to 8 core questions for a short study
- 8 to 12 core questions for a deeper study
- 1 to 2 follow-ups per important answer
- A clear closing question
The goal is not to ask everything.
The goal is to learn the most important thing well.
Weak approach
Ask about onboarding, pricing, competitors, support, design, messaging, integrations, willingness to pay, and feature requests in one interview.
Better approach
Focus this interview only on onboarding friction and activation barriers. Run separate studies for pricing and messaging.
A focused interview produces cleaner insights.
11. Write a Strong Closing Question
A closing question gives participants a chance to share something the guide missed.
Good closing questions include:
Is there anything else about your experience that we should have asked about but did not?
If you could change one thing about the experience, what would it be?
What is the most important thing our team should understand about this problem?
What advice would you give our product team?
The closing question often surfaces useful unexpected feedback.
12. Add Analysis Instructions
Most teams stop at interview questions. For AI moderated interviews, you should also define what the output should focus on.
Analysis instructions help the platform summarize the results in a useful way.
Include instructions such as:
- Identify the most common themes
- Extract strong participant quotes
- Group feedback by segment
- Highlight contradictions
- Separate product issues from messaging issues
- Identify moments of confusion
- Summarize reasons for churn
- Compare expectations vs actual experience
- Flag emotional or high-intensity responses
- Identify suggested improvements
Example analysis instructions
After the interviews, summarize the top reasons trial users did not activate. Group findings into onboarding friction, unclear value, pricing concerns, missing features, and competitor comparisons. Include participant quotes for each theme. Highlight any moments where participants expected something different from what the product provided.
This makes the research output more actionable.
AI Moderated Interview Guide Template
Use this template when creating a new AI moderated interview.
Study title
[Name of the study]
Research goal
Understand why [participant segment] [behavior or experience], so we can [decision or action].
Participant profile
Participants are [describe audience, behavior, customer status, role, segment, or screening criteria].
Interview context
This interview is about [specific topic]. The goal is to understand [what you want to learn], not to evaluate the participant.
Opening message
Thank you for taking the time to share your feedback. This interview is about [topic]. There are no right or wrong answers. Please be as specific as possible. Your feedback will help us understand what worked, what did not, and what we can improve.
Main questions
- [Warm-up question]
- [Context question]
- [Core experience question]
- [Friction or challenge question]
- [Alternative or comparison question]
- [Decision or behavior question]
- [Improvement question]
- [Closing question]
Probing instructions
If the participant gives a vague answer, ask for a specific example.
If the participant says something was confusing, ask which step was confusing and what they expected to happen.
If the participant says something was expensive, ask what they were comparing the price against.
If the participant mentions a competitor, ask what they preferred or disliked about that competitor.
If the participant gives a short answer, ask them to walk through the moment in more detail.
If the participant says they would or would not use something, ask what would need to change.
Guardrails
Ask one question at a time.
Do not ask leading questions.
Do not suggest answers.
Do not assume the participant had a negative experience.
Do not over-probe the same topic.
Do not ask for sensitive personal information unless required.
Keep the tone neutral, respectful, and curious.
Analysis instructions
Summarize the top themes, key quotes, participant-level insights, moments of confusion, objections, and suggested improvements. Highlight any differences across participant segments.
Example: AI Moderated Interview Guide for Churn Research
Study title
Trial Churn Research
Research goal
Understand why trial users who signed up in the last 30 days did not activate, so we can improve onboarding and increase trial-to-paid conversion.
Participant profile
Participants are trial users who signed up within the last 30 days but did not complete the activation milestone.
Opening message
Thank you for taking the time to share your feedback. This interview is about your trial experience. There are no right or wrong answers. Please be as specific as possible. Your feedback will help us understand what worked, what did not, and what we can improve.
Main questions
- What were you hoping to accomplish when you first signed up?
- Can you walk me through what you did after signing up?
- Was there any point where you felt stuck, confused, or unsure what to do next?
- What, if anything, felt different from what you expected?
- Did you consider any alternatives? If yes, which ones?
- What was the main reason you did not continue using the product?
- What would have made you more likely to continue?
- Is there anything else about your trial experience that we should have asked about but did not?
Probing instructions
If the participant says onboarding was confusing, ask which step created confusion and what they expected to happen.
If the participant says they did not see value, ask what value they expected to see and when they expected to see it.
If the participant says they did not have time, ask whether anything in the product made it harder to continue.
If the participant mentions a competitor, ask what made that competitor more appealing.
If the participant gives a short answer, ask for a specific example.
Guardrails
Do not defend the product. Do not explain how the product works. Do not try to solve the participant’s issue during the interview. Keep the focus on understanding their experience.
Analysis instructions
Identify the top reasons users failed to activate. Group findings into onboarding friction, unclear value, time constraints, pricing concerns, missing features, and competitor comparisons. Include quotes that clearly explain each reason.
Example: AI Moderated Interview Guide for Concept Testing
Study title
New Feature Concept Test
Research goal
Understand how target users react to a new feature concept, so we can decide whether to prioritize it and how to position it.
Participant profile
Participants are current users or target customers who regularly experience the problem this feature is designed to solve.
Opening message
Thank you for reviewing this concept. We are interested in your honest reaction. There are no right or wrong answers. Please explain what feels clear, unclear, useful, or unnecessary.
Main questions
- What is your first reaction to this concept?
- What problem do you think this is trying to solve?
- How relevant does this feel to your current workflow?
- What feels most useful about the concept?
- What feels unclear, unnecessary, or difficult to believe?
- How would you compare this to how you solve the problem today?
- What would need to be true for you to use this?
- What would you change about the concept?
Probing instructions
If the participant says the concept is interesting, ask what specifically makes it interesting.
If the participant says they would not use it, ask what would need to change.
If the participant says it feels unclear, ask which part is unclear.
If the participant compares it to another tool or workflow, ask what they prefer about the current alternative.
Guardrails
Do not try to convince the participant that the concept is useful. Do not explain the concept beyond what is provided. Do not ask leading questions like “Would this save you time?” Instead, ask what impact they think it would have.
Analysis instructions
Summarize reactions by clarity, relevance, credibility, differentiation, objections, and willingness to use. Extract quotes that explain why participants would or would not use the concept.
Example: AI Moderated Interview Guide for Win/Loss Research
Study title
Win/Loss Buyer Research
Research goal
Understand why buyers chose or rejected our product, so we can improve positioning, sales enablement, and product strategy.
Participant profile
Participants are recent buyers or lost prospects who evaluated our product within the last 90 days.
Opening message
Thank you for sharing feedback about your evaluation process. We are interested in understanding your decision, not selling or persuading you. Please be honest and specific.
Main questions
- What problem were you trying to solve when you started evaluating solutions?
- Which alternatives did you consider?
- What criteria mattered most in your decision?
- How did our product compare with the alternatives?
- What concerns or objections came up during the evaluation?
- What was the main reason behind your final decision?
- Was there anything that almost changed your decision?
- What could we have done differently?
Probing instructions
If the participant mentions a competitor, ask what stood out about that competitor.
If the participant says pricing mattered, ask how they evaluated value.
If the participant says our product was missing something, ask how important that gap was to the decision.
If the participant mentions internal stakeholders, ask how those stakeholders influenced the decision.
Guardrails
Do not argue with the participant. Do not defend the product. Do not challenge their decision. Keep the tone neutral and focused on learning.
Analysis instructions
Summarize decision criteria, competitors mentioned, strongest objections, reasons for winning or losing, pricing concerns, product gaps, and messaging opportunities.
Weak vs Strong AI Moderated Interview Guide
Weak guide
Goal:
Get feedback on onboarding.
Questions:
- Did you like onboarding?
- Was anything confusing?
- What can we improve?
Problem:
This guide is too shallow. It does not define the participant segment, the activation behavior, the expected insight, or how the AI should probe.
Strong guide
Goal:
Understand why new trial users do not complete onboarding within seven days, so we can reduce activation friction.
Questions:
- What were you hoping to accomplish when you signed up?
- Can you walk me through your first session?
- Where, if anywhere, did you feel stuck or unsure?
- What did you expect to happen at that point?
- What did you do next?
- What would have made you more likely to complete onboarding?
Probing instruction:
If the participant says something was confusing, ask which step was confusing, what they expected, and what they did next.
Why this is better:
It focuses on a specific user segment, behavior, decision, and type of evidence. It also tells the AI how to turn vague responses into useful detail.
Common Mistakes When Writing AI Moderated Interview Guides
Mistake 1: Asking survey questions
AI moderated interviews should not feel like surveys.
Avoid questions like:
- Did you like it?
- How satisfied were you?
- Would you use this?
- Was it easy?
- Which option do you prefer?
These can be useful in a survey, but they are weak as interview questions unless followed by deeper probes.
Better questions ask participants to describe experiences, decisions, expectations, and examples.
Mistake 2: Writing leading questions
A leading question suggests the answer.
Weak:
What did you like about our easy setup process?
Better:
How would you describe your setup experience?
Weak:
Did our product save you time?
Better:
What impact, if any, did the product have on your workflow?
Mistake 3: Asking too many things at once
Avoid multi-part questions.
Weak:
What did you think of the onboarding, pricing, support, and product value?
Better:
Let’s start with onboarding. Can you walk me through your first experience setting up the product?
One question at a time produces better answers.
Mistake 4: Not telling the AI when to probe
If you do not define probing rules, the AI may ask generic follow-ups.
Add instructions for vague words, short answers, competitor mentions, objections, confusion, and emotional reactions.
Mistake 5: Trying to answer too many research questions
Do not use one interview to cover every possible topic.
A focused guide produces better insights.
If you need to study onboarding, pricing, messaging, and competitors, consider separate interviews or separate sections with clear priorities.
Mistake 6: Ignoring the participant experience
Participants should not feel like they are completing a long, robotic interrogation.
Keep questions clear. Avoid jargon. Explain the purpose. Respect their time.
Mistake 7: Trusting summaries without reading transcripts
A good guide improves the interview, but researchers still need to review the raw responses.
AI summaries are useful, but they should be checked against transcripts and quotes.
AI Moderated Interview Guide Checklist
Before launching your interview, review this checklist.
Research setup
- The research goal is specific
- The participant profile is clear
- The study supports a real decision
- The interview has one primary objective
- The topic is appropriate for AI moderation
Question quality
- Questions are open-ended
- Questions are neutral
- Questions invite stories and examples
- Questions are ordered naturally
- Questions avoid jargon
- Questions avoid yes/no framing
- Questions avoid leading language
AI moderator instructions
- The AI knows when to probe
- The AI knows what vague answers look like
- The AI knows when to ask for examples
- The AI knows when to move on
- The AI has clear guardrails
- The AI is told not to suggest answers
- The AI is told not to defend the product
Output quality
- Analysis instructions are included
- Themes to watch for are defined
- Quotes should be extracted
- Segment differences should be noted
- Contradictions should be flagged
- Researchers will review transcripts before making decisions
How ListenAI Helps You Build Better Interview Guides
ListenAI helps teams create AI moderated interviews that are structured, focused, and easy to launch.
Instead of starting from a blank page, teams can define the study goal, add interview questions, set AI probing rules, and guide how the AI moderator should handle vague or incomplete answers.
With ListenAI, teams can use AI moderated interviews for:
- Product feedback
- UX research
- Churn research
- Win/loss interviews
- Concept testing
- Message testing
- Customer discovery
- Onboarding feedback
The goal is not just to collect more responses.
The goal is to collect better conversations: specific, contextual, and useful for decision-making.
FAQ
What is an AI moderated interview guide?
An AI moderated interview guide is a structured document that tells an AI moderator how to conduct a research interview. It includes the research goal, participant profile, questions, probing instructions, guardrails, and analysis instructions.
How is an AI moderated interview guide different from a normal interview guide?
A normal interview guide may rely on the human moderator’s judgment. An AI moderated interview guide needs more explicit instructions about when to probe, what counts as a useful answer, what to avoid, and how to keep the conversation focused.
How many questions should an AI moderated interview have?
Most AI moderated interviews should have 5 to 8 core questions for a short study and 8 to 12 for a deeper study. Fewer questions with better follow-ups usually produce better insights.
What makes a good AI interview question?
A good AI interview question is open-ended, neutral, specific, and designed to invite stories or examples. For example, “Can you walk me through what happened the first time you tried to set up the product?” is stronger than “Was setup easy?”
Should AI moderated interviews include follow-up instructions?
Yes. Follow-up instructions are one of the most important parts of the guide. They tell the AI when to probe vague answers, ask for examples, clarify confusion, and explore important details.
What should AI moderators avoid?
AI moderators should avoid leading questions, suggesting answers, defending the product, asking multiple questions at once, over-probing the same topic, and asking for unnecessary sensitive information.
Can I use the same guide for every AI moderated interview?
No. You can reuse the structure, but each guide should be adapted to the research goal, participant profile, and decision you are trying to support.
What is the biggest mistake when writing an AI moderated interview guide?
The biggest mistake is writing only a list of questions without giving the AI moderator context, probing rules, guardrails, and analysis instructions.
Final Takeaway
A strong AI moderated interview guide is not just a script.
It is the operating system for the interview.
It tells the AI moderator what the study is trying to learn, who the participant is, what questions to ask, when to probe, what to avoid, and how the results should be analyzed.
If the guide is weak, the AI interview will likely produce shallow answers.
If the guide is specific, neutral, focused, and well-instructed, the AI moderator can collect richer responses at scale.
The best AI moderated interview guides do three things well:
They define a clear research goal.
They ask open-ended questions that invite stories.
They give the AI clear instructions for probing vague answers.
That is how teams move from generic feedback to useful qualitative insight.
