Interviews & Transcripts

The list

Two counters head the page — Interviews and Completed — followed by the table:

Filters: Completed only, Simulated only, plus one filter per characteristic question you collect — so you can look at just the 35-44s. Page size is 10/25/50/100.

Studies whose interviews carry it also get filters for where people came from: Source (including No source, i.e. the interviews your campaign links missed), Your participant ID and Referrer, and — for embedded studies — an Embed metadata key with an optional value to match. Each control only appears when the study has data it can filter on, so a study whose links carry no tags shows none of them.

Download all transcripts exports the whole study, honouring the completed and simulated filters.

Reading a transcript

Each interview opens on its own page, showing every turn with:

• turn numbers (1:4) and who spoke
• the phase label for interviewer turns — welcome, consent_selection, ask_qualifying_question, ask_main_question
• the step or question id for flow studies
• elapsed time per turn
• Search this transcript
• With media and Reasoning toggles

Participant characteristics and screening answers are shown alongside the conversation, as is How they got here — the source tag, your own participant ID and the referring page, where the interview carries them.

Treat all three as untrusted: they come off the link the participant followed, and anyone holding that link can edit them. A participant ID here is a convenience for joining back to the sample you drew, not proof of who answered. If you need an identifier you can rely on, sign it into an embed bootstrap token — see signed embed metadata.

Recordings

If you enabled Session Recording on the Details page, audio and video for voice and video interviews appear on the transcript page, per turn. They are fetched separately from the transcript because listing them has a cost, so they appear a moment later. Nothing is recorded unless you turned it on, and video implies audio.

Known issue (September 2026). No recordings are being stored — see the note on the Details page. The media tab stays empty however the study is configured.

Searching across interviews

The Search transcripts panel covers every turn of every interview in the study. Three modes:

• Exact words — matches the text as typed, across both sides of the conversation
• Meaning — searches participants' answers for the same idea in different words; the useful one when twenty people describe one problem twenty ways
• Find a quote — locates a specific passage

Results show matching turns with a link into the transcript, and how many turns matched per interview.

Meaning search relies on embeddings, which are generated automatically when an interview finishes. Interviews that completed before automatic indexing existed are not searchable by meaning until you click Index older interviews.

With MCP

get_study_stats {"study_id": "…"}
list_interviews {"study_id": "…", "completed": true, "simulated": false, "limit": 50}
find_interviews_by_external_id {"study_id": "…", "external_participant_id": "panel-7781"}
get_interview   {"interview_id": "…"}
get_transcript  {"interview_id": "…"}
bulk_transcripts{"study_id": "…", "completed": true, "limit": 20, "offset": 0}
search_transcripts {"study_id": "…", "q": "\"too expensive\" -trial", "mode": "hybrid"}
get_interview_fetches_and_signals {"interview_id": "…"}

• list_interviews defaults to real (non-simulated) interviews; pass simulated: null for both, and filter by started_after / started_before.
• get_interview returns screening answers, characteristics and embed-widget metadata (with provenance — page-supplied or verified).
• get_transcript returns messages in order, with type separating interviewer from participant, plus captured variables and decision outcomes for flow studies.
• find_interviews_by_external_id looks someone up by your id for them — the one the entry link's ?participant_id= or the embed metadata carried — and returns every attempt they made, newest first. list_interviews, bulk_transcripts and search_transcripts take the same external_participant_id filter, and every row carries the id back so an export joins to your own records. See Recruiting participants.
• search_transcripts modes are string (websearch syntax: quoted phrases, -term), semantic and hybrid (default). Pass strict: true to fail loudly instead of silently falling back to string search when semantic is unavailable.
• get_interview_fetches_and_signals shows what an Interview Flow's side steps did outside the conversation.

Exports and privacy

Transcripts are research data about identifiable people. Export what you need, store it under your own retention policy, and strip identifiers before sharing outside your team. What you promised in the welcome and consent text governs.

Last updated: September 2026