Theme free-text survey responses
Turn hundreds of open-ended survey comments into a small set of evidenced themes, with counts and representative quotes.
Yours to copy, change, and make your own.
Replace every [BRACKETED PLACEHOLDER] with your own material before you send it.
Below are free-text responses from a survey. Question asked: "[SURVEY QUESTION]". Respondents are: [WHO RESPONDED]. Analyse the responses: 1. Identify the main themes (aim for 5-8; do not force responses into themes they do not fit). 2. For each theme: a short name, a one-sentence description, an approximate count of responses mentioning it, and two verbatim quotes as evidence. 3. Note anything said by only one or two people that is nevertheless serious (safety, discrimination, legal risk) under a separate "low-frequency, high-importance" heading. 4. State what proportion of responses fitted no theme. Rules: - Quotes must be verbatim from the data. Never edit a quote or merge two quotes. - A response can belong to more than one theme. - Do not treat frequency as importance; that is what section 3 is for. Responses: [PASTE THE RESPONSES]
- [SURVEY QUESTION]
- The exact wording of the question respondents were answering.
- [WHO RESPONDED]
- Who the respondents are and roughly how many, for context.
- [PASTE THE RESPONSES]
- The free-text answers, one per line, anonymised.
Where it shines, and where it falls over.
- Staff and member surveys with open comment boxes
- Event feedback and consultation responses
- A first-pass codebook before formal qualitative analysis
- Remove names and identifying details before pasting, especially for staff surveys.
- Run the same prompt twice; themes that appear both times are more reliable.
Counts are approximate at best. Models are poor at counting across long lists, so treat "roughly 30 responses" as "a lot" rather than a statistic, and never publish the numbers without checking them yourself. If you need exact counts, use the themes from this prompt as a codebook and tally in a spreadsheet.
The other failure is invisible: responses the model quietly ignores. The "no theme" percentage in section 4 is there to surface this, but with several hundred responses you should still spot-check a random sample against the themes to make sure nothing systematic was missed.
AI output is a first draft, not a finished product. You are responsible for whatever you send, publish, or decide with it.
More prompts worth exploring.
Specify a spreadsheet clean-up before you do it
Describe a messy spreadsheet and get a written clean-up plan: issues found, fixes proposed, and rules you can apply consistently.
Respond to a customer complaint
Draft a complaint response that acknowledges the specific problem, says what happens next, and avoids empty apology language.
Turn meeting notes into minutes and actions
Convert rough meeting notes into clean minutes with a decisions log and an owner-and-deadline action list.