Thematic analysis of open-ended survey answers: a documented first pass
Thematic analysis is slow because it is careful: generate codes, apply them to every answer, review. This tool speeds up the first two steps and leaves the review to you. It proposes themes with descriptions from a spread sample of your answers, or uses your codebook, codes every answer with up to three themes, and gives you counts and example quotes per theme to check. The first answers are free; up to 20,000 answers are $5.
- Themes with one-sentence descriptions, at most 14 plus "Other"
- Every answer coded from the same list, recorded next to the answer in your file
- Example quotes per theme, taken from coded answers, for checking and reporting
- Your own codebook can be used instead, for deductive coding
- No inter-rater agreement is computed; the review step stays with you
Where it fits in the method
In inductive work, use the proposed themes as initial codes: read them, read the example quotes, merge or rename, then run the job again with your revised list as your own codes. In deductive work, start with your framework as the code list; answers that fit none of it are coded "Other", which is a useful signal that the framework misses something.
Every coded answer sits next to its codes in your file, so a second coder can review a random sample, or the whole file, against the same codebook.
How to describe it honestly
- Say that an AI model produced a first coding pass from a codebook, and name whether the codebook was proposed or your own
- Report that each answer could receive up to three codes, so shares can exceed 100%
- Describe how you reviewed and corrected the coding
- Keep the codebook CSV with the descriptions and counts as an appendix
Limits
One question per job. The proposed themes come from a sample of up to 150 answers across the file, so a theme mentioned by very few respondents may not get its own code; add it to your own list if it matters. Very long answers are shortened to their first 1,000 characters for the model; your file keeps the full text.
How the tool works, what it costs and how your files are handled is on the Code Survey Responses page; prices for every tool are on the pricing page.
Last updated
Questions and answers
Can ChatGPT do thematic analysis of my survey?
It can suggest themes from the answers it reads, which is helpful for exploring. It does not code every answer against a fixed list in a way you can audit row by row. This tool records a code for every answer so the analysis can be checked.
Is this inductive or deductive coding?
Either. Leave the codes empty and themes are proposed from your answers; give your own codes and only those are applied.
Can I revise the themes and run it again?
Yes. Upload the file again with your revised list as your own codes. Small files are free in full; bigger ones are $5 per job.
Does it compute inter-rater reliability?
No. It gives you one consistent coding pass and the evidence to review it.
How many answers fit in one payment?
Up to 20,000 answers for $5, or up to 100,000 with Unlimited at $19 a week.
Are the answers private?
Only you can open your job. Answers are sent to an AI model provider only to code them, through routes that do not use your data for training. Jobs uploaded without an account are deleted after 7 days, and you can delete any job yourself.