AI voice interviews that ask the follow-up out loud.
Participants open your study link in a browser and talk. The AI interviewer asks your questions, listens to each answer and follows up on what they said. Every interview is transcribed. Voice runs in 25 languages and costs about 5 cents a minute on the default setup.
Walk me through the last time you tried to export a report. What were you trying to do?
Honestly, I gave up and took a screenshot. It was faster than finding the right format.
You said a screenshot was faster. Where did you look first?
Hear a follow-up before you build anything.
The interviewer asks what you are looking for in a research tool, listens to your answer and follows up out loud on the part you glossed over.
- About 3 minutes
- Uses your microphone
- No sign-up needed
A recorded answer can’t be asked why.
A recorded voice question works like voicemail. The participant hears a fixed question, records an answer, and the study moves on, whatever they said. A vague answer stays vague.
An AI voice interview is a conversation. The interviewer hears the answer, asks about the part that needs more detail, and keeps going until your topics are covered or the time limit is up. QuestionPunk has both: an audio recording question for one recorded answer, and an AI interview set to Voice when you want the follow-ups.
Brief it. Send the link. Read the transcripts.
You write the goal and set the time limit. The AI runs the conversation.
Brief
Set the goal and the time limit.
Add an AI interview to your study and set it to Voice. Write the interview goal, the greeting and any topics to avoid, then pick a time limit from 1 minute to 4 hours.
Field
Participants talk in their browser.
They open the study link on a phone or computer and allow the microphone. There is nothing to install.
Read
A transcript for every interview.
Each interview is transcribed when it ends. The AI report groups answers into themes with quotes, and any quote that doesn’t match a response is dropped.
Start from a voice interview guide.
Discovery, churn and win-loss guides set up as voice interviews. Preview the guide and adapt it to your goal.
How the interviewer decides what to ask next.
It works from your interview goal and bases each follow-up on the participant’s own answers. When it isn’t sure the interview is finished, it asks one more follow-up rather than ending early. You can read the interviewer’s prompt and add your own sections to it.
Voice interviews run on a time limit you set. The interviewer ends once your topics are covered and the participant has nothing to add, or when time is up. On longer interviews it starts wrapping up about a minute before the limit. It closes the current thread and asks at most one more question.
- Time limit from 1 minute to 4 hours, 2 minutes by default
- 11 conversation models from Anthropic, Google and OpenAI, Claude Sonnet 5 by default (Opus models need Pro)
- A cough or a quick “mm-hm” doesn’t cut the interviewer off
- Voice and text questions can share one study
About 5 cents a minute.
Voice interviews are paid in AI credits, and 1,000 credits are $1. The charge follows the length of the interview, with a 30-second minimum. The default setup (separate speech-to-text, conversation model and voice) uses about 54 credits a minute, so a 10-minute interview costs about 54 cents.
The rate changes with the voice and model you choose, and the editor shows it before you publish. Ultra-low latency, which uses a single realtime model, costs 80 credits a minute with Google or 300 with OpenAI.
- Free: 3,000 credits a month, about 55 minutes if you spend them all on voice
- Pro: 25,000 credits a month, about 460 minutes
- Buy more credits any time, and bought credits don’t expire
- Unlimited surveys and responses on every plan, under fair use
Voice in 25 languages, text in 130+.
Voice interviews run in 25 languages, including Spanish, French, German, Portuguese, Japanese, Korean, Chinese, Arabic and Hindi. Translate the study, and each participant picks a language. The AI follows up in the same one.
- Choose the voice and the speech-to-text provider
- Audio enhancement isolates the speaker’s voice in noisy rooms
- Optional animated avatar interviewer from Tavus or LemonSlice
Transcripts, recordings and a report.
Transcripts
Every voice interview is transcribed when it ends, so you can read exactly what was asked and what was said.
Recordings, if you want them
Choose whether to record audio, video or nothing, and ask for screen share when you need it.
Themes and a report
The AI report groups answers into themes with checked quotes. Export it as PDF or Word, or build a PowerPoint deck.
For UX, product and founding teams.
Concept Testing
Concept testing questions, and the tool that runs them.
Read moreUser Interviews
A user interview platform that does the interviewing.
Read moreUsability Testing
Moderated usability testing, run by an AI moderator.
Read moreCandidate Experience
A candidate experience survey you can send this afternoon.
Read moreRun your first voice interview.Start on the Free plan.
Free includes 3,000 AI credits a month, about 55 minutes of voice at the default rate, and needs no credit card.