AI Voice Recorder for Focus Groups: Consent, Speaker Attribution and Auditable Themes

Reviewed and improved: 4 August 2026.

Focus groups create insight through interaction. Participants agree, challenge, qualify and reshape answers in response to the group. That richness cannot be reduced safely to an automatic theme list without context, uncertainty and documented researcher judgement.

Start with the research question

Recording should serve a defined research purpose rather than becoming the default simply because the technology is available.

• What is the exact research question?
• Why is group interaction more useful than individual interviews?
• Who should participate, and who should not be grouped together?
• Is identifiable audio necessary?
• Which transcription or AI services are approved?
• Who can access audio, transcripts, coding files and reports?
• How will quotations be anonymised or attributed?
• What withdrawal arrangements apply?
• How long will each record type be retained?
• What happens if distress, safeguarding or sensitive disclosure occurs?

Explain the complete data journey

Participants should understand that the session is recorded, why audio is needed, whether AI processing is used, where processing occurs, who can access the files, how names and quotations may be used, the limits of group confidentiality, retention periods and how questions or withdrawal requests are handled.

Do not describe the process only as “recorded for notes.” Explain the material stages clearly enough for participants to make an informed decision.

Confidentiality in a group has limits

The researcher can control their own storage and reporting, but cannot guarantee that participants will never repeat what others said. Set ground rules, explain the limitation honestly and avoid grouping people where disclosure could create professional, social, financial or personal harm.

Design the room for attribution

Use a moderator and observer where possible

The moderator focuses on questions, participation, balance and safety. The observer can maintain a seating map, note attendance changes, capture analytically relevant non-verbal context, mark overlap, record references to visual materials and follow the approved process for distress or safeguarding concerns.

Make the session easier to transcribe

• Record introductions or participant codes in sequence.
• Ask one question at a time.
• Use participant codes when directing follow-ups.
• Allow silence rather than filling every pause.
• Stop persistent side conversations.
• Repeat material points after overlap.
• Spell specialist names and terms.
• Announce activities, stimulus changes and breaks.
• State when someone joins or leaves.
• Read back important factual claims where accuracy matters.

Separate the machine transcript from the research transcript

A machine transcript is a draft. A research-ready transcript may require corrected speaker labels, participant codes, verified terms, marked overlap, removal of identifiers, uncertainty markers, field-note links and a correction log.

Use honest uncertainty

A visible gap is safer than a plausible invented phrase.

An AI summary is not thematic analysis

Researchers should define the unit of analysis, coding approach, codebook development, treatment of contradictory evidence, reviewer roles, relationship between themes and the research question, and how AI suggestions are accepted, modified or rejected.

Frequency alone does not determine importance. A rare comment may reveal a serious risk, while repeated agreement may reflect group pressure, social desirability or dominance by a confident participant.

Use a documented coding workflow

1. Correct and anonymise the transcript according to the approved plan.
2. Read the full transcript before accepting automated themes.
3. Create initial codes linked to exact passages.
4. Record code definitions and exclusions.
5. Compare coding where more than one reviewer is involved.
6. Search actively for contradictory and negative cases.
7. Develop themes that answer the research question.
8. Check every reported claim against the source evidence.
9. Record material interpretation decisions.

Protect quotations from re-identification

1. Check the quotation against its full context.
2. Confirm that it represents the stated theme.
3. Remove unnecessary identifying detail.
4. Assess whether the remaining combination could identify the participant.
5. Apply the approved attribution rule.
6. Keep a controlled source link where required.

Changing a few words may not be enough where occupation, location, age, project or unusual circumstances identify the speaker indirectly.

Maintain an auditable evidence chain

Security and retention

Use approved accounts and controlled storage. Set separate retention periods for contact details, source audio, identifiable transcripts, anonymised analysis files, coding records and final reports. Delete temporary exports and working copies when they no longer serve the approved purpose.

When a compact recorder is not enough

Use specialist multi-microphone or professional research capture where the group is large, the room is difficult, remote and in-person participants are mixed, precise attribution is critical or the consequences of error are high. Test the exact arrangement before collecting real data.

How NERALVO Halo fits

Check whether NERALVO Halo fits focus groups work provides NOTE mode, 64GB local storage, up to 35 hours of recording and Bluetooth transfer to DOWAY for transcripts and structured outputs. One year of DOWAY Max is included from activation.

Halo may support approved small-group research, pilot sessions and researcher debriefs. It does not replace participant information, ethical review, room design, transcript correction or qualitative analysis.

Frequently asked questions

Can AI perform thematic analysis automatically?

It can suggest patterns, but researchers remain responsible for method, interpretation, contradictory evidence and the final claims.

Can one compact recorder handle every focus group?

No. Group size, room acoustics, overlap and attribution requirements may justify specialist equipment.

Can participants be promised complete confidentiality?

No. Explain that other participants may repeat what they hear, even when ground rules are agreed.

What if a speaker cannot be confirmed?

Mark the speaker as uncertain rather than guessing.

Should the raw AI transcript be shared with participants?

Only where the approved research process requires it. First consider identifiers, third-party information, access controls and whether sharing could expose other participants.

Can recordings be kept for future research?

Only where that future use is properly defined, permitted and communicated. Do not retain identifiable audio indefinitely merely because storage is available.

Useful resources

• ICO data-protection guidance
• Transcription Accuracy vs Audio Quality
• How to Compare AI Transcription Systems Fairly
• Check whether NERALVO Halo fits focus groups work

Final focus-group checklist

• Research question and processing plan approved
• Participant information accurate and complete
• Limits of group confidentiality explained
• Room and speaker mapping tested
• Overlap and uncertainty marked honestly
• Transcript corrected before analysis
• Coding method documented
• Contradictory evidence considered
• Claims linked to checked evidence
• Quotation identification risk assessed
• Retention and deletion controlled

Bottom line: AI can accelerate transcription and organisation, but accountable focus-group research depends on participant transparency, reliable capture, corrected evidence and a visible analytical trail.