How to run AI-moderated interviews: A step-by-step guide

AI-moderated interviews help teams collect rich, conversational feedback without manually running every session.

But the quality of your results depends on how well you set up the study. In this chapter, we look at the steps to run an AI-moderated interview workflow in Maze, including how to:

Here’s how to set up your AI-moderated study from start to finish.

Running AI‑moderated interviews follows the same core logic and methodology as any good qualitative research. You still define the goal, design the guide, recruit participants, run sessions, and synthesize insights.

The difference is that an AI moderator handles the live conversation while you are in control of the setup and review. Unlike a general-purpose chatbot, it follows the goals, questions, and probing rules you provide. This makes it possible to run live interviews without assigning a moderator to every session.

Research goals as short statements that describe what you want to understand, like “Understand why new users drop off after signup” or “Learn how teams currently run user interviews.”

Research questions then break those goals into specific, answerable questions, such as “What do users find confusing in the onboarding flow?” or “Which part of the interview process feels most time‑consuming today?”

Once you’ve outlined a few clear research goals and turned them into specific questions, you’re ready to set them up for your AI-moderated study.

Log in to your Maze account, or sign up if you haven’t created one yet. From your workspace, create a new project (or open an existing one), click ‘New study’, then select ‘AI‑moderated study,’ and give it a name your team will recognize.

You’ll see two options for how to set it up:

Here, you’ll need to turn your brief into a short list of specific research goals first.

A discussion guide is the outline that keeps your interview focused on what matters. Start by turning your goals and research questions into a simple structure:

You can also add simple rules or branches to your guide (for example, “If they mention onboarding, ask these follow‑ups”) so the AI can adapt the conversation while still following your structure.

In Maze, you can pick a discussion style by choosing ‘Structured’ or ‘Freeform’. Structured keeps questions in a fixed order for comparability, while Freeform lets the AI moderator adapt to the flow of the conversation.

Next, you need to set ‘Discussion depth’ to decide how long the AI should stay on this goal. For example:

Then add discussion instructions. Here, you tell the AI moderator how to guide the conversation for this goal. You can write your own guidance or select a template to get started quickly.

Recruiting the right participants means finding people who match your target users, so the insights you collect are relevant to the user experience you’re studying.

Start by defining who you want to talk to:

Then choose your recruitment channels, such as your product’s user base, email lists, social media, or external panels and tools. For market research, external panels can help you reach people beyond your existing customer base.

If you’re using Maze, go to the ‘Recruit’ tab to decide who should take part in your AI‑moderated interviews. You can either share a ‘study link’ through your own channels or create an order with panel participants to reach a targeted audience from Maze panel.

If you’re using panel participants, add a screener so only the right people qualify for your study. Click ‘Add screener’, then set up one or more multiple-choice questions, choose single-select or multi-select, and mark which choices qualify or disqualify participants so only your target audience continues to the study.

Start with a small pilot to make sure your AI‑moderated interview flows as expected and collects the insights you need. Share your study link with a few internal stakeholders or a handful of trusted participants, then review how the AI moderator handles your goals, questions, and discussion instructions.

Spot unclear wording, missing probes, or goals that need a different format, style, or depth, and make those adjustments before you open recruitment to your entire participant sample size. This protects the participant experience and improves your chances of collecting high-quality responses at scale.

Once your pilot looks good, move your study to live and start collecting responses from your recruited participants. In the Recruit tab, check that your sources are ready (study link and panel order), then share the link or confirm the order so new participants can start joining sessions.

Since participants can join on their own time, you can reach people across time zones without scheduling human moderators for every session.

Although moderation is automated, AI-led interviews still require the same safeguards used in traditional methods. Skipping those foundations can weaken the quality of the research process and the decisions based on it.

You’ve now seen how to plan, set up, and launch AI‑moderated interviews. In the next chapter, you’ll learn how to review sessions, work with AI‑generated themes and highlights, and turn your interviews into analysis and reports your team can trust.