AI Interview Recording: A Participant Disclosure Checklist

A customer agrees to talk about a skincare purchase. Later, someone on the marketing team finds a useful moment in the recording and drops the clip into a presentation. That could be within the purpose the customer understood. It could also be a use nobody mentioned before the camera came on.

Before recording an AI interview, decide what you will collect, who can use it and when you will delete it, then explain those choices before recording starts.

A participant notice is the plain-language explanation of those arrangements. Consent, where required, involves a separate agreement process. I would prepare both alongside the interview questions, while there is still time to change the study. A notice cannot fix a workflow your team has not thought through.

What should participants know before an AI interview?

Participants should understand that an AI system will conduct the interview, what it will record, why you are asking for their information and who will receive it. They also need to know whether their words or image may appear in clips, how long you plan to keep the material and whom they can contact with questions.

The explanation needs to match the study they are about to join. “We collect information to improve our services” gives someone very little help in deciding whether to discuss a purchase, a health concern or their household budget on camera.

A useful notice lets them make that decision without guessing. Depending on the study and applicable law, you may also need specific consent steps or other disclosures. The worksheet below is a preparation aid, not a universal legal checklist.

Why recording creates extra decisions

A voice or video recording can contain more than the answer you asked for. Someone may mention a family member, show their home or describe a medical issue while explaining why they bought a product. Removing their name from a transcript does not make the original video anonymous.

The material can also travel. A researcher watches a full interview. A colleague pulls out a clip. Someone else exports it for a wider meeting. Each action can make evidence easier to understand, while putting it in front of another audience.

I can see the temptation to use the clearest customer quote wherever it will help an argument. That is exactly why I would set the sharing boundary before finding the quote.

How the recording workflow works

Think of a recorded interview like a parcel that gets unpacked and copied at several stops. The participant provides the original material. The platform may turn it into a transcript, extract clips and organize themes. Your team may then export some of those outputs into another system.

Each copy needs an owner and a use boundary. Deleting the original recording does not necessarily delete an exported clip or the transcript someone saved elsewhere. The actual behavior depends on the platform and your workflow, so you need to confirm it.

Draw the route before launch: participant, platform, research team, other recipients, exports. At each stop, ask what is available, who has access and what happens when the retention period ends. This is often easier than trying to reason from a privacy policy alone.

What Listen Labs explicitly says

Listen Labs describes useful evidence-review features on its Bain page: voice and video AI moderation, transcripts, clips, coded themes and quotes linked to the relevant video moment. Those are vendor-described capabilities. They can help a team revisit the evidence behind a finding rather than rely only on a summary.

Its study privacy policy says studies collect responses that may include audio or video, depending on the study type. It also says a study may be sponsored by a customer, called the Research Organization, and that responses will then be shared with that organization.

Its AI conduct statement says it does not use collected data to train its AI models. Keep that wording narrow. It does not, by itself, establish every retention, deletion, subprocessor or downstream-use condition.

These statements provide a starting point. They do not answer every question about your particular study. Nor does a detail missing from these statements establish that Listen Labs lacks a protection. Check the current terms, relevant documentation and your agreement, and get unresolved points confirmed in writing.

A pre-launch participant disclosure worksheet

For each field below, label the answer vendor-stated, team-decided or awaiting confirmation. Save the source or written confirmation beside it. Some fields will need more than one label: your team might choose a retention period that the vendor must confirm it can support.

An unanswered field is a task, not permission to fill the participant notice with reassuring language.

1. AI moderator disclosure

Fill in: “The interview will be conducted by [system/tool]. [Describe any human observation or review].”

Say directly that participants will interact with an AI moderator. Decide whether anyone will watch live or review the recording later, and confirm what the tool actually supports. Do not imply that a human is present to help if that is not the arrangement.

Put this explanation before the interview begins, rather than leaving someone to discover it through the first question.

2. Recording format

Fill in: “We will collect [typed responses/audio/video] and create [transcripts/clips/other outputs]. Recording begins [when].”

List only the formats you will actually use. Confirm whether video includes audio, whether transcripts are generated and whether participants can turn off their camera or choose another format.

If there is no alternative, say so. A participant should not discover that a camera is mandatory after working through the introduction. Test the flow to check that the notice and any required agreement appear before capture begins.

3. Study purpose

Fill in: “We are asking about [specific experience] so that [team] can decide [specific decision].”

Keep the purpose narrow enough to guide use. Understanding why customers abandon a subscription is a workable purpose. “Business improvement” leaves too much room for interpretation.

Decide whether the material is for research only or whether you want any additional use. If you later want a customer testimonial, treat that as a new permission question rather than assuming research participation covered it.

4. Recipients and access

Fill in: “Responses will be available to [organization and team/roles]. [Named external recipients or relevant categories] will receive [specific material].”

Separate the organization sponsoring the study from the vendor processing it. Within your organization, decide who needs raw recordings and who only needs findings.

Confirm relevant service-provider and subprocessor arrangements through the appropriate documentation. Also check account access and sharing settings. “Only our research team can view this” needs to be true in the product and in your internal process.

5. Clips, quotes and reuse

Fill in: “We may use [clips/quotes] in [specific settings], viewed by [audience]. [Describe any separate permission process].”

An internal research meeting and a public advertisement have different audiences. Make the intended use clear, including whether the participant's face, voice, name or other identifying details will appear.

My default would be to keep public promotional use outside research participation and request separate permission if a real need arises. Whatever boundary you choose, check download and link-sharing controls so the workflow supports it.

6. Storage and exports

Fill in: “Material will be stored in [platform and approved internal locations]. Exports are handled by [owner/process].”

Confirm hosting locations and relevant contractual safeguards if your study requires them. Do not guess from the vendor's headquarters.

Include the places your team controls: downloaded files, presentation folders and messaging channels. Decide whether exports are necessary at all. If they are, assign someone responsibility for tracking them rather than leaving deletion to whoever remembers making a copy.

7. Retention and deletion

Fill in: “We will retain [material type] until [date or defined event], then [action]. [State confirmed limits or exceptions].”

Consider raw recordings, transcripts, clips and summaries separately. You may need different periods, but each should have a reason and an owner.

Confirm how the vendor handles deletion requests, backups and exported material. Ask what your team must do itself. Avoid promising “everything is deleted immediately” unless that is a promise you can actually keep across the whole workflow.

8. Questions, stopping and requests

Fill in: “Contact [monitored address] about participation or your information. You can stop by [method]. After stopping, [explain what happens to material already collected].”

Specify how participants can ask questions, request withdrawal or seek deletion, including any applicable limits and process. Confirm what stopping the session does technically. Closing a browser tab may end an interview without deleting the answers already submitted.

Assign the inbox to a person or team who can act on a request. An address nobody monitors does not help the participant.

A concrete example

Suppose you sell refillable household cleaners and want to understand why customers stop ordering refills. You choose AI-moderated audio interviews because hearing the explanation may help you ask better follow-up questions. You do not need video to examine that decision.

In this hypothetical plan, the insights lead and customer experience lead can access recordings. The wider team receives a summary with identifying details removed. Audio clips stay out of sales presentations and public marketing. Your worksheet names a retention end date and an owner for deleting permitted exports.

Before launch, you still need to confirm that the platform settings, access controls and deletion process support that plan. If they do not, change the plan or the collection method. I would rather collect less than write a notice the study cannot honor.

A draft participant notice

Use this as drafting material, not a legally sufficient universal consent template. Replace every placeholder, remove anything that does not apply and arrange any legal review your study needs. Any required agreement should be captured through an appropriate process before recording starts.

About this interview

[Organization] is conducting this study to understand [specific experience] and help decide [decision]. You will speak with an AI moderator provided by [vendor]. [Explain any human observation and later review].

If you choose to participate, we will collect [audio/video/text] and create [transcripts and other outputs]. Recording begins [when]. [Explain available alternatives, or state that there are none]. Please avoid sharing information about other people or details unrelated to the study.

Your responses will be shared with [research organization/team] and handled by [vendor and relevant service providers, with a link to further information]. [Explain who can access raw material].

We will use [quotes/clips] in [settings] for [audience]. [Explain identification and any separate permission required for other uses].

We will keep [material] for [period or until a defined date/event]. [Explain deletion arrangements and confirmed exceptions].

You can stop the interview by [method]. [Explain what happens to responses already collected and how requests are handled]. For questions about participation or your information, contact [monitored address]. Further privacy information is available at [link].

How Probe fits

At Probe, we let you design a survey or interview study and send it to your own audience, or have us recruit participants from our recruitment partner's vetted panel. Own-audience studies are free up to 10 questions and 25 responses. For panel recruitment, you see one all-in price per study before launch. Results come back as charts, themes and quotes.

I would start by asking whether you need a recording at all. A focused survey may answer a question about why someone canceled without collecting their voice or face. Interviews make more sense when you need room to explore an explanation. Our guide to Qualtrics, AI Interviews or Surveys: Which Tool Fits? can help with that choice.

Choosing Probe does not remove the preparation work. Confirm the data-handling details for your chosen study format, and explain your team's intended uses before collecting responses.

Before inviting participants, run through the study as one. Read the notice, follow the recording steps and check where the resulting material goes. If the explanation and the actual flow disagree, fix that gap before asking a customer to turn on their microphone.