Choosing the right Maze study type (unmoderated, moderated, and AI‑moderated)

Choosing the right kind of study in Maze starts with what you want to learn, who you need to hear from, and how quickly you need user insights. This chapter helps product teams match their user research goals to the right testing method, whether you’re exploring a problem, validating a Figma prototype, or improving usability in a live product.

You’ll learn how:

Unmoderated research is useful when you need fast quantitative data on tasks, flows, and usability
Moderated research works best when you need deeper context from testers
AI-moderated research gives teams a way to collect open-ended user feedback at scale without losing the structure of a guided conversation

When to use unmoderated vs. moderated vs. AI-moderated research methods

Each method supports a different kind of user experience question:

If you want to…	Unmoderated	Moderated	AI‑moderated
Test a prototype or UX	You need fast, scalable task data.	You expect confusion and want to guide people live.	You want users to talk through flows but can’t attend every session.
Explore a problem	You’re validating known themes with more people.	You’re early and need open, flexible conversations.	You have an interview guide and want many voices at scale.
Get rich qualitative insight	You just need short comments alongside metrics.	You need depth, probing, and rapport.	You want lots of open‑ended answers, auto‑captured and summarized.
Move fast with limited time	You need results in hours/days with minimal setup.	You can only run a few high‑value sessions live.	You need interview-like depth but can’t schedule many calls.

Foundational best practices for running any research

No matter whether you run research unmoderated, moderated, or AI-moderated, a few basics will make your work more reliable and easier to act on. These apply across most UX research methods, from early discovery interviews to usability testing for a new feature or template.

Write a clear study description: Explain in a few sentences what you’re exploring, who it’s about, what participants will do, and how long it will take.
Define 1–2 primary research goals: State what you want to learn and how it will inform decisions. For example, “understand why people abandon checkout so we can improve conversion.”
Recruit the right participants: Decide who you need to hear from, use simple screeners, and avoid whoever is available so your findings reflect your real audience.
Respect time and privacy: Set realistic session lengths, get informed consent, and be transparent about recording, incentives, and data use.
Plan your analysis before you launch: Agree on what data you’ll collect, how you’ll tag or code it, and who needs to see the outcomes so insights turn into decisions.

Those fundamentals apply to any study, but interview studies require a few extra considerations to ensure conversations stay focused, consistent, and useful.

Best practices for running unmoderated research

Unmoderated research only works well when the study you’re conducting is easy to understand and complete without live guidance. These best practices help you reduce confusion, protect data quality, and get results you can trust.

Start with one clear goal: Define what you want to learn in a sentence (for example, “Can people complete checkout without getting stuck?”), and pick 1–2 success metrics like completion rate or time on task. This way the study stays focused and the results are easier to interpret.
Keep tasks realistic and in plain language: Write tasks that mirror real behavior (“Pay an electricity bill”) and avoid giving away the path (“Click the Bills tab, then the green Pay button”). That makes it easier to see whether the experience is genuinely usable.
Give extra context, because you won’t be there: Briefly explain who the participants are pretending to be, what they’re trying to do, and any constraints. This helps them understand the task on their own without a moderator stepping in.
Pilot with 2–3 people first: Run a dry run to check if instructions are confusing, tasks are too long, or tech breaks, then fix those issues before launching at scale. That way, you can fix problems early instead of collecting unreliable data at scale.
Limit the length of the session: Aim for a focused test (for example, 10–20 minutes). Shorter sessions help participants stay engaged and reduce the risk of rushed or low-quality responses.
Include a few targeted follow-up questions: Add short, open-ended questions after key tasks to capture useful context on what happened. This gives you more insight without overwhelming participants with a long survey.
Stick to mature designs and simple flows: Use unmoderated studies for high‑fidelity prototypes or live experiences where flows are reasonably clear. More complex or unfinished flows are harder to evaluate well without a moderator present.

Best practices for running moderated research

Moderated studies give you the chance to ask follow-up questions, explore unexpected behavior, and understand the reasoning behind what participants do. But that only works when the session is well structured, and the moderator knows how to guide the conversation without taking it over.

Prepare a flexible discussion guide, not a script: Outline the key topics, tasks, and must-ask questions before the session. This gives you enough structure to stay focused while still leaving room to follow useful threads as they come up.
Warm up and build genuine rapport: Start with simple, friendly questions and explain what will happen in the session. This helps participants feel more comfortable thinking aloud and being honest, even when they feel stuck or confused.
Facilitate actively but stay neutral: Use open prompts like “What are you thinking here?” or “What would you expect to happen next?” Avoid praising or reacting in ways that hint at the right answer. This helps you learn how participants naturally interpret the experience without steering them toward a particular answer.
Balance observing, probing, and staying on time: Pay attention to hesitation, confusion, and non-verbal cues, ask short follow-up questions when something matters, and keep the session moving. This helps you get deeper insight without losing the most important parts of the discussion.
Coordinate roles behind the scenes: If other teammates are observing, agree in advance on who is moderating, who is taking notes, and how questions will be passed along. This keeps the session organized and makes the experience feel smoother for the participant.

Best practices for running AI-moderated research

AI-moderated research can help teams run more interviews faster, but it only works well when the study is carefully set up. These best practices help you give the AI the right structure, reduce messy responses, and collect actionable insights.

Be explicit about goals and boundaries: Tell the AI moderator exactly what you want to learn, which topics to prioritize, and where not to go. This helps keep follow-up questions relevant and reduces the risk of the interview drifting into less useful territory.
Design a structured but flexible guide: Use clear, neutral, single-focus questions, and mark where the AI should probe deeper versus move on. This gives the conversation enough structure to stay on track while still leaving room for richer responses.
Pilot and watch a few early sessions: Review how the AI handles different kinds of answers, then tighten wording, adjust probing rules, and fix confusing moments before scaling up. This helps you catch weak prompts or awkward interview flows before they affect a larger batch of sessions.
Set clear expectations for participants: Explain that they will be speaking with an AI, how long the session will take, and what to do if something feels unclear. That makes the experience feel more transparent and helps maintain trust and engagement throughout the study.
Use AI for analysis, but keep a human in the loop: Let the system cluster themes and identify quotes, then review a sample of sessions yourself before you share or act on the findings. This gives you the speed benefits of AI without treating its output as automatically final or correct.
Combine AI‑moderated research with other methods: Pair AI-moderated interviews with prototype testing to validate designs, surveys to quantify recurring themes, and a few live interviews to dig deeper with key participants. This gives you a more complete picture than any one method can provide on its own.

Next up: Running moderated and AI-moderated interview studies

Interview studies help you go beyond clicks and task success to understand the why behind people’s behavior. In the next chapter, we’ll dig into how to structure those interviews, what to prepare before each session, and how to use both live and AI moderation to uncover deeper insights from your users.