People are more honest with an AI moderator than a human one, and they're most honest on the questions that decide launches: what consumers will actually pay, why they quietly switched, and what disappointed them about a product they stopped buying.
The reason?
There is no person on the other side to impress. No social bias. No comparison.
In Keplar interviews, we see people talk to a voice moderator about their emotions, their price sensitivity, their socioeconomic situation, and the real reasons behind a purchase with an openness that is hard to get at scale from human-run interviews.
Two things are driving it.
- No fear of judgment: people know the interviewer is a system, not someone with a life and a status of their own to be measured against, so the instinct to adjust how they come across dissipates.
- Trust forms faster: With a human, the opening minutes go to building up a rapport before anyone tells a real story. With an AI moderator, people open up from the third or fourth question, and we see it consistently in the call recordings.
Are people more honest with AI moderators than human ones?
None of this is surprising. Survey methodologists worked it out decades ago: people answer more honestly when no human is in the room. Put an interviewer in front of someone and the answers drift toward what sounds good. Let them answer in private and they become more honest when talking about sensitive topics (Kreuter et al.). Researchers have watched the why of it directly. When people believe they're talking to a computer rather than a person, their fear of being judged drops and they stop managing what impression they want to leave (Lucas et al.).
Honesty used to cost you depth
Here's what the old methods never figured out. The only reliable way to take the human out was a survey, and the survey took the interviewer out with it. You got the honesty but gave up depth in the process. A form can't hear someone hesitate, can't ask why, can't dig deeper.
So teams were stuck choosing between the honest-but-shallow survey and the deep-but-performed interview.
An AI voice moderator is the first method that removes the human without removing the interviewer. No person to perform for, so participants feel like they can open up, completely judgement free, but keeping an interviewer in the room, so the depth stays. Because it runs with hundreds of people at once, you get both at a scale you would normally get for a quant study.
This isnât unfounded either, evidence for this is starting to land. A set of LSE studies by Geiecke and Jaravel benchmarked AI-led interviews against trained sociologists running them face to face. The AI voice interviews scored higher on quality than the human ones, 3.93 versus 3.51 on an expert scale. Between 42% and 55% of people preferred the AI, and many said they felt less judged and less pressured, and more willing to share a sensitive view. Voice mattered more than any other factor in the study: moving people from typing to talking produced the single biggest jump in answer quality. The interviews also pulled far more material than just an open-text survey box, a 148% jump in word count (London School of Economics).
The honest objections
The obvious one is that a skilled human moderator can read the room. They can catch small gestures, or tone switches, that a machine just canât. That's real, and it's worth something. But it goes both ways. The same human connection that builds a relationship is the presence that makes people feel like they have to perform with their responses.
The second objection is that people rush a machine and give surface-level answers. That's a risk with anything self-administered, which is why the format matters. A static AI survey would earn that complaint. A voice moderator that asks a follow-up when you hint at something is the opposite of a form, and the jump from text to voice in the LSE work is the evidence that the conversation is what is pulling the depth.
However, none of this is universal. The honesty gain is most important, on those sensitive, identity-loaded questions, and smaller on the mundane ones. That tradeoff is okay. The sensitive questions are the ones that sink launches.
Whatâs human, and always will be
Despite all the benefits of conversational research in gathering both depth and scale, often faster and more free from bias than humans, there is one crucial drawback. That is when teams want to run true qualitative work. True qualitative work is aimed at helping an insights or marketing team build empathy for their consumer through immersion. True qual if often 60 minutes or more in length, and doesnât actually need to be done at scale. This kind of work offers the human team strategic value in addition to the insights. So it remains a deeply human activity, and likely always will be.
What this changes for a CPG insights teams
The questions you most need the truth on are the ones a voice AI interviewer can get unbiased signals on at scale:
âWhatâs the value you perceive this product to offer?â
âOn which occasions would you consider it?â
âWhat would you really pay, why did you stop buying from us, what let you down about our product?â
Keplar runs AI-moderated voice interviews with hundreds of real consumers at once and codes them automatically, so you can put those questions to a category-sized sample and get qualitative answers at quantitative scale. On price sensitivity, that is the difference between a number you can take to a pricing committee and one that flatters you into the wrong move.
The deeply human part of true qualitative work isnât going anywhere, but voice AI enables a new standard of depth and scale, with increasing evidence showing that people open up more easily and quickly with AI moderators than humans.
FAQ
Are participants more honest with an AI moderator than a human one? Yes, especially on sensitive topics like price, switching, and product disappointment. Taking the human out of the room removes the fear of being judged, which is the main thing that pushes interview and survey answers toward what sounds good instead of what they are actually feeling.
Do people open up to an AI interviewer, or hold back? They tend to open up, often faster than with a human. With no person to build rapport with or impress, people share emotions, price sensitivities, and the real reasons behind a purchase from the first question. An AI voice moderator that asks follow-ups then gets depth a static survey never could.
When should you use an AI moderator instead of a human one? Use an AI moderator when you need honest answers on sensitive or identity-loaded topics at scale, like price, switching, and socioeconomic questions, where a human's presence biases the response. Keeping a human in the loop when reading subtle nonverbal cues or handling a delicate live dynamic is the whole point of the session.
Curious about what AI-native research looks like? Try Keplar here or book a meeting with us here.