Code Interview Transcripts into Themes with Participant Counts and Verbatim Quotes

Turns pasted interview transcripts into a theme table with participant counts, exact supporting quotes, counterpoints, and a separate list of one-off mentions.

You getTheme table with columns, strong themes and one-off mentions

Prompt

Prompt
Act as a qualitative researcher who codes interview transcripts for a living.

I will paste one or more interview transcripts right after this message. Each participant is labeled with an ID (such as P1, P2) or a name. Code the transcripts into themes and show how much support each theme has.

Rules:
1. Use only verbatim quotes copied exactly from the transcripts. Never paraphrase inside quotation marks, and never create a quote that is not in the text.
2. Count participants, not mentions. One person raising a point five times counts as 1.
3. Mark a theme Strong if at least 3 participants or at least half the sample back it, whichever is smaller. Everything else goes under One-off mentions.
4. Tag every quote with its participant ID.
5. If any participant contradicts a theme, show that quote as a counterpoint.

Output format:
1. Theme table with columns: Theme | One-line definition | Participants (count and IDs) | Strength
2. Strong themes: for each, 2-3 supporting quotes with IDs, plus one counterpoint quote if one exists
3. One-off mentions: theme, participant ID, one quote
4. Gaps: questions the transcripts leave unanswered

End with one line on limits: sample size, participants who gave little detail, and anything you could not code with confidence.

Example output

At a glanceWhy FreshCrate customers cancelled, by participantsParticipants mentioning each theme, out of 6 interviews
  • Price felt too high over time: 4 of 6
  • Food waste: 3 of 6
  • Skipping weeks is a hassle: 3 of 6
  • Packaging guilt: 1 of 6
  • Recipe repetition: 1 of 6

Example for: 6 interviews with people who cancelled a meal-kit subscription (fictional brand FreshCrate).

1. Theme table

ThemeOne-line definitionParticipantsStrength
Price felt too high over timeCost stopped feeling worth it after the intro discount ended4 (P1, P2, P4, P6)Strong
Food wasteIngredients spoiled before they were cooked3 (P2, P3, P5)Strong
Skipping weeks is a hassleThe skip process was easy to forget or confusing3 (P1, P3, P6)Strong
Packaging guiltToo much plastic and ice packs1 (P4)One-off
Recipe repetitionMenus felt the same after a few months1 (P5)One-off

2. Strong themes

Price felt too high over time (4 of 6)

  • P1: "Once the 50 percent off ended, I was basically paying restaurant prices to do the cooking myself."
  • P4: "Eleven dollars a serving for chicken and rice. I did the math one night and that was it."
  • P6: "It was fine at first but the full price just didn't sit right."
  • Counterpoint, P3: "Honestly the price was okay, it was more that I wasn't using it."

Food waste (3 of 6)

  • P2: "I threw out a whole bag of herbs twice in one month."
  • P5: "By Friday the fish was already questionable, so we'd order pizza."
  • No counterpoint found.

Skipping weeks is a hassle (3 of 6)

  • P1: "You have to skip by Tuesday and I never remember Tuesday."
  • P3: "I thought I skipped it, then a box showed up."
  • Counterpoint, P2: "Skipping was easy, that wasn't my issue."

3. One-off mentions

  • Packaging guilt, P4: "The ice packs pile up under the sink and I don't know what to do with them."
  • Recipe repetition, P5: "Every week was some kind of grain bowl."

4. Gaps

  • No one was asked what price would have kept them subscribed.
  • Unclear whether waste came from portion size or schedule changes.
  • Only P3 and P6 mentioned household size; it may affect waste and price.

Limits: 6 participants only; P6 gave short answers, so their theme coding rests on two quotes; P5's comment about "the app" was too vague to code.

Written by AI from this prompt. Your results will vary.

Tips and common edits

Why this works

  • Requiring verbatim quotes tagged by participant stops the model from paraphrasing or inventing evidence.
  • Counting participants instead of mentions keeps one talkative interviewee from inflating a theme.
  • A set Strong threshold plus counterpoints separates real patterns from one-offs and shows disagreement.

Use after user interviews, customer calls, or research sessions. Have transcripts ready with participant labels. Long sets work best in Claude or Gemini; in ChatGPT, paste in batches if needed.

Common edits

  • Raise the Strong threshold to 4+ participants for larger studies of 12 or more interviews.
  • Add a column linking each theme to a product decision or next research question.

Make it repeatable

Keep using this prompt without pasting it in every time.

Prompt details

By
Prompts-GPT editors
Published
How prompts are written and checked

Did this prompt work?

Keep exploring

More in Research