Consulting and recruiting
Focus group recordings and transcripts: consent, retention and reuse
By SourceX Editorial · Reviewed by Noah Loul ·
Short answer
Focus group recording consent should cover who records, who watches, which tools transcribe, how long each file is kept and every planned reuse. Treat video, audio, identified transcripts and de-identified transcripts as separate records with separate retention clocks. Reuse beyond the client project, including AI training, generally needs consent that names that use.
Key takeaways
- Consent at recruitment and again on the record at the start of the session gives the clearest evidence.
- Video and audio identify people by face and voice, so they carry the tightest retention and reuse limits.
- A de-identified transcript is a different record from an identified one and can have a longer clock.
- The client contract sets ownership of recordings; participants' privacy interests apply whoever owns the file.
- New uses such as AI training generally need consent that names them, not a general research clause.
Who owns focus group recordings?#
Focus group recordings are usually governed by the client contract. Many master agreements treat recordings and transcripts as client deliverables; others give the client viewing rights and leave the files with the agency. Facilities and streaming platforms may hold copies under their own terms.
Ownership is only half the question. Whoever owns the file, the people on it have privacy interests that the consent form defines. A client that owns a recording still cannot use it for something participants were never told about, and an agency holding a copy is bound by the same limits.
Consent checklist for recorded sessions#
A consent form for recorded groups and interviews should state each point below in plain language. The moderator then confirms the key points on the record at the start of the session, which gives the agency evidence tied to the recording itself.
- That the session is recorded, and whether that means audio, video or both.
- Who will watch, live or later: client observers, a viewing room or a streaming link.
- Which services transcribe or analyze the recording, including AI notetakers and transcription vendors.
- How participants will appear in transcripts and reports, such as first name only or a code.
- How long recordings and transcripts are kept, and by whom.
- Whether clips may appear in client presentations or internal highlight reels.
- Any reuse beyond the project, such as methods training or AI training, stated separately with its own choice.
- How to withdraw, and what happens to recordings already made.
Recording laws, observers and remote sessions#
Recording laws differ by state, and some states require every party to a conversation to agree before it is recorded. Remote groups can put participants in several states at once, so a practical approach is to apply the strictest standard to every session instead of checking each participant's location.
Observers count too. Client staff watching a stream are part of the audience participants agreed to, so the consent form should name them and observers should sign confidentiality terms. If an AI notetaker joins a call, participants should know before it starts, and the tool's terms should be on the agency's approved list. Zoom's terms, for example, state that it does not use audio, video or chat content to train Zoom or third-party AI models; other tools differ, so check each one.
Where video is analyzed for faces or voices, biometric privacy laws may apply in some states. Counsel should assess those questions before any facial or voice analysis is run.
Separate retention clocks for each record#
Each recorded session produces several records with different privacy weight, and each needs its own clock. The table shows a typical structure; set the actual periods from your contracts and consent forms.
Platform settings can delete files before your policy does: Zoom offers an admin setting that removes cloud recordings a chosen number of days after they are created, and single recordings can be exempted. Check those settings against the schedule and exempt anything under legal hold.
| Record | Identifies participants? | Clock starts | Typical end action |
|---|---|---|---|
| Raw video | Yes, by face and voice | Session date | Delete after analysis or client handover |
| Raw audio | Yes, by voice | Session date | Delete once the transcript is checked |
| Identified transcript | Yes, names and personal details | Transcript delivered | Replace with a de-identified version, then delete |
| De-identified transcript | Not directly; check for indirect clues | De-identification sign-off | Keep per contract and consent |
| Highlight clips | Yes | Clip created | Keep only for the stated presentation use |
| Consent forms | Yes | End of the related records' life | Keep as long as the records they cover |
| Screener and contact data | Yes | Session date | Delete once incentives are paid |
Reuse: what each new use generally requires#
Reuse is where most disputes start, because a recording made for one client's project is easy to copy into another context. The table shows what usually needs to be in place for common reuses; counsel confirms the specifics project by project.
| Reuse | What usually needs to be in place |
|---|---|
| Clips in the client's own presentations | Consent naming clip use, and a client contract that allows it |
| AI-assisted analysis within the project | Disclosure of the tool and approved vendor terms |
| Internal moderator training | Consent naming training use, or de-identified transcripts only |
| Insights for a different client | Client contract permission; usually de-identified and aggregated |
| Licensing transcripts for AI training | Consent naming that use or robust de-identification, plus confirmed client and agency rights |
| Licensing video or audio | Explicit consent naming the use; otherwise usually out of scope |
Can transcripts and recordings be de-identified?#
Transcripts can often be de-identified; recordings rarely can be without destroying their value. An editor can replace names, employers, places and distinctive stories with placeholders. A face or voice cannot be removed so easily, and blurring or voice alteration changes what the recording shows.
The legal test is demanding. GDPR Recital 26 says that to decide whether a person is identifiable, account should be taken of all the means reasonably likely to be used, such as singling out, by the controller or by another person. A participant who describes a rare job in a small town may still be identifiable after the name is gone.
Keep the de-identified transcript as a separate record with its own sign-off. Store any key that maps participant codes to people apart from it, and delete the key when it is no longer needed.
Illustrative: a qualitative agency sorts its archive#
Illustrative: a fictional qualitative agency has run in-person groups at rented facilities and remote groups on Zoom for many years. Recordings sit on a file server, transcripts came from more than one transcription vendor, and the consent form changed several times.
The principal sorts projects by consent version. Older forms mention recording and client viewing only. Recent forms add AI-assisted analysis and a separate choice about use in methods training. Video from every period stays within its original project, while de-identified transcripts from projects with the broader consent are documented as candidates for internal reuse.
Raw video past its retention period is deleted, consent forms are filed with the transcripts they support, and the consent template gains a separate, optional question about licensing for future studies.
How SourceX approaches qualitative archives#
SourceX treats recordings and transcripts as carrying a high privacy burden, which under the SourceX Enterprise Data Value Framework reduces net value. Video and audio are usually out of scope unless consent explicitly covers licensing. In the SourceX five-step transaction, the Rights step checks consent versions and client contracts project by project, and Preparation covers de-identification of any transcripts that proceed. The initial assessment uses metadata only, and the agency signs off each later stage.
Frequently asked questions
Can we use an AI transcription tool without telling participants?
It is safer not to. Participants should know which services process their words, and some clients require disclosure of any AI tool. Name the transcription method in the consent form, confirm the tool's terms on training and retention before use, and record which tool handled each session in the project file.
What if consent forms from older projects are missing?
Treat those recordings as limited to the original project and your standard retention period. Without evidence of what participants agreed to, reuse beyond the project is hard to defend. Record the gap in your inventory rather than reconstructing consent from memory.
Can a participant withdraw after the session?
Most consent forms allow it, and the agency should be able to act on it. Remove the person's contributions from recordings or transcripts where practical, delete their contact data and record what was done. Findings already delivered in aggregate usually cannot be recalled, which the form should say.
Are B2B interviews treated differently?
The same principles apply with extra care. Professionals often discuss their employer, customers and deals, so transcripts can hold third-party confidential information as well as personal data. De-identification needs to cover company names and deal details, not only the participant's name.
Do client observers need to sign anything?
Usually yes. Observers see faces and hear personal details, so a short confidentiality agreement and a rule against recording, screenshotting or photographing the session protect participants and the agency alike. Keep the signed observer list with the session file, because it shows who saw the recording if a participant later asks.
Sources
- Zoom lets account owners, admins and licensed users enable deletion of cloud recordings after a specified number of days counted from each recording's creation, with an option to exempt individual recordings. Source
- Zoom's Terms of Service (Section 10.2) state that Zoom does not use audio, video, chat or other communications-like Customer Content to train Zoom or third-party AI models. Source
- GDPR Recital 26 says that to decide whether a person is identifiable, account should be taken of all the means reasonably likely to be used, such as singling out, either by the controller or by another person. Source
Related resources
See if your company qualifies
A short company assessment. No data uploads are needed.