AI-Moderated Voice Interview

Participants browse your product while an AI voice moderator asks questions, probes friction points, and adapts in real time. You get prioritized findings — not raw transcripts.

yourproduct.com
CANDOR
AI MODERATOR

Walk me through what you're seeing on this page. What stands out first?

PARTICIPANT

The pricing section is confusing — I can't tell which plan includes...

Automated

Candor does this so you don't have to

🔄

Adapts in real time

The AI moderator listens to participant responses and dynamically adjusts follow-up questions to probe deeper into friction points and moments of delight.

🚫

Never asks leading questions

Built-in guardrails prevent the moderator from introducing bias. Questions are open-ended and neutral — the kind a trained researcher would ask.

🔒

Enforces privacy (no PII)

If a participant shares personal information, the AI gently redirects. Names, emails, and other PII are automatically redacted from transcripts.

📋

Custom interview guides

Provide your own discussion guide with topics, tasks, and rating scales. The moderator follows your structure while adapting naturally to participant responses.

01 Under the hood
1

Consent screen + mic access

Participants see a clear consent screen explaining the study. Once accepted, the browser requests microphone access and the session begins automatically.

🎙
Allow microphone
2

Product loads in iframe

Your product URL loads in a sandboxed iframe alongside the moderator panel. Participants interact with the real product while the AI observes and asks questions.

<iframe>
+
AI
3

AI asks questions following your guide

The moderator follows your interview script — or generates questions from your topic list. It probes deeper on interesting responses and gracefully handles off-topic tangents.

Topic 1: First impressions
Topic 2: Navigation
Topic 3: Pricing
4

Interactive overlays

The moderator can surface task cards, rating scales, and progress bars during the session. Participants complete structured activities while continuing the conversation.

TASK
RATING
PROGRESS
5

Findings synthesized automatically

After each session, Candor synthesizes the conversation into prioritized findings. As more sessions complete, cross-session patterns emerge and findings are re-ranked automatically.

P0
02 What you get back

Prioritized Findings

8 sessions analyzed
P0

Pricing page confusion

6 of 8 participants could not determine which plan included the feature they needed. Most expected a comparison table but found only paragraph descriptions.

"I've been going back and forth for two minutes and I still can't tell if the Pro plan includes API access."

P1

Onboarding flow drop-off

4 of 8 participants hesitated at the account creation step, citing the number of required fields. Two mentioned they would have abandoned in a real scenario.

"Do I really need to fill out my company size just to try the product?"

P2

Navigation label ambiguity

3 of 8 participants misinterpreted 'Workspace' as a collaboration feature rather than a project container. This caused brief confusion but was self-corrected.

"Oh wait, Workspace isn't like a shared space? It's just my projects?"

Cross-session learning: Findings are re-prioritized as more sessions complete. Early patterns strengthen or fade based on new evidence.
03 Best for

Ready to run your first voice interview?