As of 13 August 2026, AI can analyse open-ended customer survey answers.

What to actually do

1. Use a tool built for this

   The route this page recommends

2. Do it yourself

   Second choice

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. 1. Export the open-ended answers from your survey tool, retaining a response ID and the survey question but removing names, email addresses, phone numbers and other unnecessary personal data.
   2. 2. Open a chatbot and paste the survey question, the relevant customer or product context and the business decision you want the analysis to support.
   3. 3. Paste the anonymised responses in a format with one response per row, or split the data into clearly labelled batches if the chatbot cannot accept it all at once.
   4. 4. Paste the prompt above and ask the chatbot to analyse the complete set rather than drawing conclusions from an incomplete batch.
   5. 5. Copy the themes, counts, recommendations and quoted responses into a working document, keeping each quote beside its response ID.
   6. 6. Compare every count and quote against the original survey export, and read a sample of responses assigned to each theme to check that the labels match what customers wrote.
   7. 7. Mark uncertain themes and individual complaints for a colleague or customer-service owner to assess before sharing the report or changing service.

   Prompt

   Analyse the customer survey responses below. The survey question was: [PASTE QUESTION]. The business context is: [PASTE BRIEF CONTEXT]. The decision this analysis will support is: [PASTE DECISION].
   
   First, clean only obvious formatting problems and do not rewrite customer meaning. Then produce:
   1. The main themes, with a short definition of each.
   2. The number of responses that clearly support each theme, separating clear mentions from uncertain or overlapping mentions.
   3. The strongest positive and negative patterns.
   4. Recurring complaints, requests and causes of dissatisfaction.
   5. Any urgent, sensitive or potentially individual customer issue that should be passed to a person, without diagnosing or guessing.
   6. Representative quotes for each theme, copied exactly and labelled with the response ID if one is supplied.
   7. A short list of practical actions, each linked to the evidence.
   8. A limitations section covering ambiguity, possible sampling bias, missing context and responses that do not fit a theme.
   
   Do not invent quotes, counts, customer details or reasons. Do not infer age, gender, ethnicity, health, income or other sensitive characteristics. Do not present sentiment as fact when the wording is ambiguous. If the data is too large or incomplete for a reliable count, say so and give a qualitative analysis instead. Separate observations from recommendations. Use plain UK English.
   
   Before finalising, check that every reported count and quote can be traced to the supplied responses. Here are the responses:
   [PASTE ANONYMISED SURVEY EXPORT HERE]

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

3. Hand it to a person

   The distant third

   A person who owns the outcome does this end to end, worth it when the failure is dear.