AI moderation is often framed as a question of replacement: can an AI interviewer do what a human moderator does, and if so, when should teams use one instead of the other? That framing misses the nuances and context of a study.
AI interviewers introduce another way to conduct research, one that sits somewhere between the scale of unmoderated research and the adaptability of a semi-structured moderated session. An AI interviewer can lead a research conversation on its own, asking questions, following up in real time, and adapting to what a participant says, without a human moderator present. Where that approach fits depends on the research question, the participants, the subject matter, and the stage of the product or idea.
To understand how people who do research are actually thinking about and using AI moderation, and what they see as its value and limitations, we surveyed 80 researchers and product managers across the US, Europe, the UK, and Canada.
What you'll learn
- How teams are actually splitting moderated vs. unmoderated work today
- Which research methods are most common, and where AI moderation shows up in that mix
- What teams expect from an AI interviewer in terms of oversight and control
- Where AI moderation isn't the right fit and where it starts to behave like its own hybrid method
What research teams are actually running in 2026
Before asking where AI interviewers fit, it's worth looking at what teams are running in 2026. Among the teams we surveyed:
- 90% Interviews
- 85% Usability testing
- 75% Surveys
- 71.3% Concept testing
- 40% Card sorting / tree testing
- 31.3% Diary studies
Interviews and usability testing are the most common methods in our data. Both allow you to understand not just what participants do, but the reasoning and context behind their behavior. Choosing a research approach means connecting the question you want to answer with the kind of data you need to answer it, and understanding how that data will lead you to the answer. It also depends on the time and resources you have.
The tradeoff between depth and scale
Half of the teams we surveyed lean to moderated research, while 38.8% run an approximately 50/50 split between moderated and unmoderated research. Notably, zero participants said they run fully unmoderated research. A live moderator gives you the ability to respond, ask why, probe deeper when a participant says something unexpected or interesting, describe a problem you didn't anticipate, and spend more time on an answer that deserves further exploration. But it also comes with practical constraints. You need to recruit participants, coordinate calendars, manage time zones, attend sessions, and find time to review them afterward. The number of sessions a team can conduct is ultimately limited by the team’s availability and capacity.
Unmoderated research removes much of that friction. A study can scale. Teams can reach more participants and run studies in parallel. AI moderation sits between these approaches. It introduces the ability to respond to participants without requiring a researcher to be present for every conversation.
A new approach for gathering qualitative data: AI-moderated research
AI-moderated research uses an AI interviewer to conduct a research conversation without a human moderator present. Instead of presenting participants with a fixed set of questions, the AI can ask a question, interpret the response, and decide what to ask next based on the participant's answer and the goals you set. For instance, you want to understand how users solve a particular problem. One participant might use a third-party tool, another might have developed a manual workaround, while someone else might have a different workflow altogether. The AI can conduct a semi-structured interview, following up differently with each participant, exploring what they do, why they do it, and what they're trying to accomplish so you can really grasp their motivations and behaviors based on your research goals.
60% of participants said they see AI interviewers as a way to conduct significantly more research. That doesn't mean replacing human moderators with an AI interviewer for every study. It means making more sessions possible with the right people in situations where a researcher wouldn't otherwise have the time or resources to moderate every conversation.
Beyond scale, AI interviewers also reduce researcher bias by providing a more consistent interviewing experience. In a pilot study of Optimal’s AI Interviewer, participants also noted the AI interview is easier to talk to than a traditional interview and the on-screen text was helpful for accessibility. Curious what we learned? Read the full study write-up.
Setting parameters and control while using an AI Interviewer
AI interviewers don't automatically add value to a study or help your team gather meaningful insights on their own. The quality of those conversations depends on how well you configure the AI interviewer, the context they provide, and the goals they define.
To get the most out of an AI interviewer, you should begin by providing clear context about the study, the company, its participants, and the product being researched. You can define specific goals at different levels, including the overall study, individual sections, and specific questions. These goals guide the AI in determining which follow-up questions to ask and where to probe more deeply. You can also adjust how actively the AI interviewer probes participants, allowing them to control the depth of follow-up questions when responses are unexpected, unclear, or lack detail.
Limitations of AI Interviewers
Even with thoughtful setup and clear parameters, AI interviewers aren't the right fit for every study. Human moderators will always play an essential role, especially in cases like:
- Sensitive topics or research with vulnerable populations which requires trauma-informed training
- Trust-building conversations and client engagement where the value of the conversation extends beyond data collection and insights
- Early exploratory research, where you don't yet know what you're looking for (AI moderation works best when the research team can provide clear goals and context).
AI interviewers are best understood as an expansion of researchers' capabilities and a new approach to gathering qualitative insight, not a universal replacement for existing methods and approaches.
Insight doesn’t come from AI alone
One of the clearest themes from our research is that people who do research want visibility into how AI-generated research is produced. Teams want to scrutinize the data themselves. They want access to conversations, raw recordings, transcripts, and the ability to sense-check findings against what participants actually said. That matters because AI moderation changes the role of the researcher. Instead of being present for every conversation, researchers can focus on identifying the right people to talk to, developing research questions, organizing priorities, engaging with and aligning cross-functional teams, weighing insights and seeing what requires further exploration, and crafting the bigger story.
AI can conduct more conversations and accelerate the time to insight.
Research teams still need to determine what those insights mean within the context of the market, the business, and everything that surrounds a decision.
AI moderation makes more possible
AI interviewer changes the tradeoff between depth, scale, speed, and the team involved. It is still an emerging approach, and teams are figuring out where it belongs. Its value comes from what it makes possible like more opportunities to ask, follow up, and understand.
More than any single feature, what we're building toward with Optimal Intelligence is a research experience that uplifts human expertise. Technology can accelerate the work. It can't weigh nuanced risk, define ethics, or build relationships. That's still on us.
AI moderation is a new methodology, not a replacement for the ones that came before it. It earns its place in your workflow because of what it makes possible. Discover what our AI Interviewer taught us about conversational AI.

.webp)
.jpeg)







