AI-Moderated Research: A Practical Guide

AI moderation is a research approach where an AI conducts sessions with real participants using a study setup defined by the researcher.

Unlike a fixed survey, AI moderation can adapt the conversation in real time. It listens to each response, asks relevant follow-up questions, and guides the session toward the research goals.

Participants can answer questions, complete product tasks, or navigate a website or prototype. The AI follows the discussion guide while adapting its follow-ups to what each participant says.

AI-moderated research is sometimes confused with other ways AI is used in research. The distinction is important: the participant is still a real person.

It is also different from simply adding conversational wording to a survey. In a survey, the path is usually predefined. In an AI-moderated interview, what the participant says can influence what the moderator asks next.

The research goal can be the same. The major difference is who conducts the live session and how that changes the scale, consistency, and flexibility of the research.

Each method makes a different trade-off. The colors below show the relative strength of each approach on a given dimension — they are not quality scores.

AI moderation works best when the research team can define what the moderator needs to learn before sessions begin.

At Glimma, we see AI moderation as a human-in-the-loop workflow. The AI handles the live conversation, but the researcher stays involved at every important stage: setting up the study, reviewing how moderation works, validating the evidence, and refining the study over time.

This balance is important because product research often requires both flexibility and control. The moderator should adapt to each participant, but it should do so within boundaries defined by the researcher.

The researcher starts by defining the research goal, the product decision the study should inform, and the participant group they want to learn from. At this point, they also choose the right Glimma study type.

Glimma supports several product research workflows, including in-depth user interviews, concept testing, Figma prototype testing, live website testing, mobile app testing, onboarding or journey testing, and AI product or user evals.

Next, the researcher creates the discussion guide and configures how Glimma should behave. This includes the core questions, tasks, probing instructions, tone, structure, and how flexible the moderator should be.

With Glimma, you can specify the follow-up questions for each question in the guide and how the moderator should probe.

Once the study is live, Glimma conducts participant sessions independently. It asks the planned questions, adapts its follow-ups in real time, and responds to what each participant says or does.

Because the sessions are AI-moderated, many participants can complete them in parallel and on their own schedule, which makes it easier to collect richer feedback from a broader group.

After sessions are complete, Glimma analyzes transcripts, recordings, ratings, and participant responses to surface recurring patterns, key findings, and supporting evidence across the study.

Glimma brings together advanced analysis features from both the quantitative and the qualitative research toolbox.

The researcher reviews the findings against the underlying evidence, decides what matters for the product, and checks whether any important perspectives or exceptions were missed.

If new questions emerge, the researcher can refine the guide, change the moderation rules, and run another round.