What it feels like to be interviewed by an AI: what 200 participants actually said

We ask every participant the same thing at the end of a Nava interview. Not about the product they just discussed. About the conversation itself. What was that like for you?

I read those answers obsessively. Across about 200 participants, the most consistent thread was one I will admit I did not fully expect to be the loudest one. Around 60% told us they felt less judged talking to an AI moderator than they would have with a human interviewer. Less judged. Not "more efficient," not "faster," not "more convenient," though we heard those too. The thing people kept naming was the absence of someone sizing them up.

I want to walk through what participants actually report, because it changed how I think about what we are building.

The thing people did not expect to feel

The single most common sentiment was relief at not being judged. When you talk to another person, even a kind and professional one, part of your attention goes to managing the impression you are making. Are they bored? Did that sound stupid? Should I have a more interesting opinion about this? That background process never fully shuts off in a human conversation, and it costs you. It is the same social desirability bias that quietly bends survey and focus group data, just felt from the inside.

With an AI moderator, a lot of participants describe that process going quiet. There is no face subtly reacting, no person to perform competence for. So they say the unglamorous true thing instead of the polished acceptable thing. One pattern we see again and again: people admit the boring or slightly embarrassing version of their behavior, the snack they actually eat, the feature they never use, the workaround they are a little ashamed of, in a way they themselves seem mildly surprised to be admitting.

The most common thing participants told us was not that it was fast. It was that they felt less judged.

That candor is the whole point. We did not build voice interviews to be a novelty. We built them to get the answer people actually believe, in their own words, and it turns out removing the human audience is a large part of how you get there.

"I didn't expect it to feel like a conversation"

The second recurring theme is surprise at how natural it feels. People come in with a picture in their head of an automated phone tree, the press-one-for-billing kind of experience, flat and scripted and deaf. That is the expectation we are working against.

What they describe instead is being listened to. The AI moderator hears what they said, responds to it, and asks a real follow-up about the specific thing they just mentioned, not the next item on a fixed list. Each of our 20+ languages has its own named AI moderator with a natural native voice, so it is not a translated script read out loud, it is a conversation in your own language with someone who sounds like they belong in it. Participants notice. The comments often have a small note of "wait, that was actually pleasant" underneath them.

A piece of that is the absence of a particular human awkwardness. The AI moderator never sighs, never checks the clock, never makes you feel like you are talking too long or not getting to the point fast enough. Which leads directly to the next thing people mention.

Permission to take your time

A lot of participants describe feeling free to think. In a human interview there is a pace you feel obligated to keep. Silence gets uncomfortable. You rush to fill it, and you give the first answer that arrives rather than the truer one underneath it.

The AI moderator does not rush you. It waits. People tell us they appreciated being able to pause, reconsider, and circle back to something they said three minutes ago because they realized they had not said it quite right. That is where the good material lives, in the second pass, the correction, the "actually, now that I think about it." A patient listener with no schedule of its own gives people room to get there. Several participants essentially said the quiet part out loud: they felt they could be more honest because nobody was waiting on them.

Being voice-only matters here too. No camera means no one is watching your face while you think, no managing your expression on top of your words. You can do this from your couch, look out the window while you gather a thought, and just talk. For a lot of people that is the difference between a guarded answer and an open one.

Skepticism, then comfort

I do not want to make it sound like everyone arrives delighted. They do not. A real share of participants start out skeptical, and some say so directly at the top. Talking to an AI about something that matters to them feels strange before it feels normal. A few expect it to be cold, or to misunderstand them, or to be vaguely pointless.

The pattern worth reporting is what happens next. The skepticism tends to fade within the first couple of exchanges, usually the moment the moderator asks a follow-up that proves it was actually listening. Once a participant realizes the thing in front of them is responsive rather than scripted, the posture changes. The arc from "I'm not sure about this" to "oh, this is fine, I can just talk" is one of the most common shapes in our end-of-interview feedback, and watching it repeat is one of the more satisfying parts of this job.

I take the skeptics seriously, and I do not want us to ever wave them away. Their initial hesitation is a fair instinct. The fact that it so often softens into genuine comfort is not something to take for granted. It is something to keep earning, interview after interview, by making the experience good enough that the strangeness gives way to candor.

What 200 voices taught us

Here is the honest summary. We set out to make qualitative research faster and far less expensive, and it is both. But the finding that has stayed with me is not about cost or speed. It is that, for the better part of these participants, a machine that does not judge them turned out to be a setting they could be more honest in than a room with a person in it.

That is a slightly uncomfortable thing to sit with, and I think it is also the most important thing we have learned. It says less about the cleverness of the technology and more about how much of our everyday candor we quietly spend on managing what other people think of us. Take that audience away, give someone a patient voice and as much time as they need, and a lot of people will simply tell you the truth. Our job is to keep listening, in their own words, and to keep deserving the honesty they offer.