AI Moderation Through the Eyes of a Human Moderator
Losing a project to an AI moderator is not something we ever thought would end up on our market research bingo card ten years ago, but here we are.
In the past six months, we’ve started seeing proposals where AI moderation is positioned as an alternative to traditional qualitative interviews. The pitch is fairly straightforward: AI can run a large number of interviews in a short period of time and at a much lower cost.
To be fair, that isn’t an empty claim. AI moderation can do those things, which makes the appeal pretty obvious for research teams under pressure to move quickly or stretch a limited budget. If you can conduct 50 interviews instead of 20 and do it in a fraction of the time, the economics of research start to look very different.
As human moderators, this moment feels a little strange. On one hand, it’s natural to question what this means for the skill we’ve developed. Moderating qualitative research has always been deeply human work, requiring careful listening, sensitivity to nuance, and an ability to sense when something important is sitting just below the surface of what a participant is saying.
At the same time, resisting new tools has never been a particularly productive stance. Technology has always changed how research gets done, and good researchers adapt. The goal isn’t to pretend AI moderation doesn’t exist, but rather to understand where it genuinely makes sense and where it may fall short.
We’re also starting to see the shift show up for respondents. After participating in AI-moderated interviews, some physicians now pause to ask whether they are speaking with a person or an AI moderator.
These experiences have led us to think less about whether AI moderation works and more about when it actually makes sense to use it. Like most tools in research, its value depends heavily on the job you’re asking it to do.
There are certainly situations where AI moderation can work well. Projects with a clear and defined set of questions are a good example. If the goal is to ask the same questions across a large group of participants and collect consistent feedback, AI can handle that kind of work efficiently. In these cases, the moderator isn’t necessarily adapting the conversation dramatically from person to person. The objective is consistency and volume, and AI can deliver that.
AI moderation may also enable research that otherwise wouldn’t happen. When timelines are extremely tight or budgets are constrained, the ability to conduct interviews quickly and at lower cost can allow teams to gather at least some directional insight rather than none at all.
But there are also situations where human moderation remains essential.
Complex research is one of them. Many qualitative projects involve messy decision processes, competing influences, and ideas that participants are still working through as they talk. Human moderators spend a lot of time listening for subtle signals such as hesitation, contradictions, or comments that don’t quite line up with what was said earlier. When those moments arise, we adjust the conversation in real time, probing differently and following threads that may not have been anticipated in the original guide.
That kind of adaptive judgment is unscriptable.
Exploratory research presents another challenge. Some of the most valuable qualitative insights emerge when a moderator notices something unexpected and decides to pursue it. The discussion guide evolves, the conversation shifts, and themes begin to emerge organically across interviews. While AI systems are becoming increasingly capable of generating follow-up questions, true exploration relies heavily on curiosity and intuition. It involves recognizing when something interesting is happening and knowing how to dig deeper in that moment.
Then there are projects involving highly sensitive topics. In healthcare research especially, participants are often sharing deeply personal experiences related to diagnoses, fears, difficult treatment decisions, and the impact of illness on their lives and families. People tend to open up differently when they feel heard and understood by another person, and that sense of trust can be difficult to replicate with technology.
None of this means AI moderation doesn’t have a role in qualitative research. In fact, it almost certainly will. Used thoughtfully, it may become a valuable tool for structured interviewing at scale, early-stage hypothesis testing, or gathering rapid directional feedback.
The most interesting future probably isn’t AI versus human moderators, but rather understanding which approach fits the question you’re trying to answer. And it’s worth acknowledging that technology is evolving quickly, which means this conversation will likely look different even a few years from now.
For now, though, the heart of qualitative research hasn’t changed. Understanding people still requires curiosity, judgment, and the ability to listen for the things that aren’t being said outright, skills human moderators have been refining for a long time and that still matter today.