Skip to content

Latest US poll: 63% say gas prices are affecting their household budget a great deal or a moderate amount

See the poll

Interviews

Spoken answers, without the camera

Some people talk freely but would rather not be on camera. Audio questions let them answer out loud, and an AI interviewer can ask up to three follow-ups based on what they said.

Book a demoWatch the analysis run
  • Up to three AI follow-ups on what each person said
  • A transcript for every accepted answer
  • Checked for length, words and subject before it counts
“I called my sister first, just to talk it through.”Talked it through with family“It all went on the credit card, which I hate doing.”Put it on a credit card“No meals out, cheaper groceries, and we canceled two subscriptions.”Cut back on extras“The dentist let us pay it over three months.”Asked for a payment plan“I still check the balance every week.”Put it on a credit card“We dipped into savings. That’s what it’s there for.”Used savings“Honestly, I didn’t sleep properly for a week.”Lost sleep over it“Mostly that I wasn’t being stupid. She went through the same thing last year.”Talked it through with family

Why it matters

Typed answers run short. Cameras put some people off.

Ask people to type why they did something and many write a line. Ask them to switch on a camera and some hold back, or leave. Speaking takes less effort than typing and feels less exposed than video, which makes audio a good middle ground for questions that need a real explanation.

3AI follow-ups at most per audio question
3words at least, or the answer is sent back with a reason
0.5relevance an answer needs before it counts (scored 0 to 1)
5audience groups compared side by side

Step by step

How it works

  1. Add an audio question

    In the CatalyseAI builder, type the question, attach a picture if people need something to react to, and name the subject their answer should be about.

  2. Set length and follow-ups

    Pick the shortest and longest answer you will accept. Turn on the AI interviewer for one, two or three follow-ups and note what it should probe.

  3. People answer out loud

    They tap record and speak. A recording that is too short, too long, has fewer than three words or drifts off the subject is sent back with a reason, and they try again.

  4. Listen, read, compare

    Play any answer next to its transcript. Themes, top words and the AI analyst cover the main answers and the follow-ups together.

Example

Example: how people cope with a surprise bill

A US credit union wants to understand how members cope with a bill they did not see coming. Money is personal, so the team chooses audio over video. The survey screens for adults who had a surprise expense in the past year, then asks: “Tell us about the last time a bill you didn't expect arrived. What did you do in the first few days?” Answers can run to two minutes, and the AI interviewer may ask up to three follow-ups, briefed to “find out who they talked to and what they cut back on.” The team reads the themes, then builds a group of everyone who mentions a credit card and compares it with the rest.

The story below follows this study, one answer at a time.

Watch it run

The audio semantic analysis

One voice, played through the analysis

Scroll to play the answer: the words light up as the playhead passes, and the themes appear where they were said.

Step 01 / 05

Recorded, then checked

People tap record and answer out loud. Too short, too long, fewer than three words or off the subject: the answer goes back with a reason, and they try again.

Attempt 1 · sent back

Fewer than three words. We couldn’t make out enough words. Try again somewhere quieter.

Attempt 2 · accepted
  • Lengthup to 120 s84 s
  • Wordsat least 365
  • Topic relevanceneeds 0.5 to count0.91

No camera, so no light check: only the words count.

Step 02 / 05

Themes, found by meaning

AI insights reads every transcript and groups what people said by meaning, so different words for the same point land in the same theme.

Her words, and the theme each joins

  • “I didn’t sleep much that week”Lost sleep over it14%
  • “call my sister”Talked it through with family26%
  • “asked if we could pay in parts”Asked for a payment plan22%
  • “stopped eating out”Cut back on extras38%
  • “went on a credit card”Put it on a credit card31%

Share of all 186 answers that raise each theme

Step 03 / 05

A group, built from her words

Make a group of everyone whose transcript mentions a phrase, then compare it with everyone else on any question in the survey.

New audience group

Everyone whose transcript mentions “credit card”58of 186

By the end of the month, did you feel back in control?

Mentioned a credit card34%
Everyone else61%

Do you have money set aside for the next surprise?

Mentioned a credit card22%
Everyone else47%
Significant difference

Step 04 / 05

Follow-ups on what she said

Up to three AI follow-ups, each built from something the person said and guided by your brief.

Your brief to the AI interviewer“Find out who they talked to and what they cut back on.”
  1. Built from “call my sister”

    1 You said you called your sister first. What were you hoping she’d say?

  2. Built from “stopped eating out”

    2 What did cutting back look like, day to day?

  3. Built from “went on a credit card”

    3 How long did it take to clear the credit card?

Skipping a follow-up ends the follow-ups for that question.

Step 05 / 05

A summary you can check

The AI report writes up the study, and Ask the study answers what comes next, with every number and quote traced to your data.

Coping with a surprise bill HTML

A surprise bill hits sleep before it hits savings. Most members cut back on extras first; nearly a third put the cost on a credit card and regret it.

“Mostly that I wasn’t being stupid. She went through the same thing last year.”Rosa, 46 · follow-up 1 · 0:14

What do people who put it on a credit card say they’d do differently?

Ask the studyMost say they would keep a small emergency fund. Asking for a payment plan sooner comes second.

Every number and quote traced to your data

What the analysis gives you

Groups, themes, voices and words for every audio question

Build a group from any phrase people said and compare it with everyone else. All of it sits in the report next to the rest of your survey, and all of it works from the transcripts.

Audience groups

Start with who said it: a group built from two words

New audience group

Everyone whose transcript mentions “credit card”

58people of 186

Where it matched · each person counts once

“It all went on the credit card.”Dev, 31 · main answer

“The rest went on a credit card.”Rosa, 46 · main answer

“Half on the credit card, half from overtime.”Marta, 39 · follow-up 2

Compared with everyone else

Felt back in control by the end of the month

34%61%

27 points apart

Has money set aside for the next surprise

22%47%

25 points apart

Mentioned a credit card · 58Everyone else · 128

Significant difference, marked in the crosstab

AI insights · Themes

Every theme, sized and with its tone

ThemeShare of answersTone
Cut back on extras38%71 people
Put it on a credit card31%58 people
Talked it through with family26%48 people
Asked for a payment plan22%41 people
Used savings17%32 people
Lost sleep over it14%26 people

PositiveNeutralNegative

Voices

Press play on the answers behind each theme

Put it on a credit card

“It all went on the credit card, which I hate doing.”

Dev, 31 · Main answer · 0:52

Talked it through with family

“Mostly that I wasn’t being stupid. She went through the same thing last year.”

Rosa, 46 · Follow-up 1 · 0:14

Lost sleep over it

“Honestly, I didn’t sleep properly for a week.”

Nina, 27 · Main answer · 0:35

Words and phrases

The words people used

  • “credit card”31%
  • “cut back”24%
  • “payment plan”18%
  • “my sister”9%

Share of people. Each person counts once, even if they said it again in a follow-up.

In the report

Per-question numbers in the report

Accepted
Answers that passed all checks and count in the results.
Re-recordings
Failed tries on main answers and follow-ups. These count attempts, not people.
Average length
Seconds per main answer. Follow-up answers are left out so the figure stays comparable.
Average relevance
A 0 to 1 score for how well answers stay on your subject. An answer needs at least 0.5 to be accepted.
Why tries failed
A count for each reason: length, too few words, or off the subject.
Words and phrases
The terms people used most. Each person counts once, even if they said it in three follow-ups.

Audio or video?

Pick the format that fits the question

Choose audio

When the topic is personal, such as money or family, or when you only need the words.

  • The same checks on length, words and subject
  • A transcript for every accepted answer
  • Up to three AI follow-ups
  • No camera, so no presence or lighting check and no attention score

Choose video

When you want to see the person, or something they show you.

  • A camera check before recording: a live person, in front of the camera, in enough light
  • An attention score from 0 to 1
  • Transcripts, themes and follow-ups, as for audio
AI video interviews

One survey can use both.

Everything you take away

Transcripts and themes

Keep scrolling: the reel runs sideways.

A transcript for every answer

An answer is only accepted once it has a transcript of at least three words, so nothing in the results is audio you can't read.

The follow-up thread

Under each answer you see what the interviewer asked next and what the person said back.

Themes per question

Answers are grouped into themes with a sentiment bar. Pick a theme to read and play the answers in it.

Search across answers

Find every answer that mentions a word, in the main answers and the follow-ups.

Groups from what people said

Build an audience group from everyone who mentioned, say, “price”, and compare it with everyone else.

Raw data

Download a CSV with file links, transcripts, relevance scores, lengths and follow-ups.

Inside any study

Ask why, out loud, in the studies you already run

An audio question is one more question in a CatalyseAI survey. Put it after a rating and the themes tell you what is behind the number.

Concept testingFind the winning idea before you build itCustomer satisfaction & NPSMeasure satisfaction, then find what to fixUsage & attitudeLearn how people use your category, and why

FAQ

Questions, answered

When should I choose audio over video?

Choose audio when the topic is personal, such as money or family, or when you only need the words. Choose video when you want to see the person or something they show you. One survey can use both.

Is an audio answer checked the same way as video?

It gets the same checks on length, words and subject. There is no camera, so there is no presence or lighting check and no attention score.

Do people have to answer the follow-ups?

No. They can skip a follow-up, and that ends the follow-ups for that question. Their main answer still counts.

Can I hear the recordings, or only read them?

Both. Every answer plays in the report beside its transcript, and its follow-ups play underneath.

Related

Goes well with

ProductAI video interviewsReal people, on camera, in their own wordsProductQualitative researchRecruitment and fieldwork for qualitative researchProductResponse quality checksFind careless answers before they reach your resultsOverviewAll productsThe platform, panels and checks behind our research.

Hear the reasons, not just the ratings

Tell us what you want to learn and who from. We'll show you how an audio question with AI follow-ups would work in your study.

Book a demo

Nearest office: Orlando, FL (corporate office)