As of 13 August 2026, AI can analyse comments from your CSAT surveys.

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. Open your survey or customer-feedback system and export the CSAT comments with the score, response ID, date, product, channel and any non-sensitive segment fields that are relevant.
   2. Remove names, email addresses, order numbers, free-text personal details and any other information that could identify a customer, while keeping a response ID so the analysis can be traced back to the source.
   3. Add a short note containing the survey question, the meaning of the CSAT scale, the date range and the definitions of your products, channels and segments.
   4. Paste the cleaned data and the prompt into a chatbot, then ask it to produce the themes, supporting response IDs, score comparisons and limitations in the requested format.
   5. Open the original survey export and compare every proposed theme and quoted comment with its response ID, correcting misclassified, duplicated or invented evidence.
   6. Ask a colleague who understands your service to challenge the interpretations and check whether important operational context is missing.
   7. Turn only the validated themes into follow-up actions, then record which source comments and scores support each action.

   Prompt

   Analyse the anonymised CSAT survey data below. Treat each row as one response and do not invent comments, causes, customer details or statistics.
   
   Data:
   [PASTE CSV OR TABLE HERE]
   
   Context:
   Company or service: [NAME]
   Survey question: [QUESTION]
   CSAT scale and meaning: [SCALE]
   Relevant date range: [DATE RANGE]
   Known products, channels or customer segments: [LIST]
   
   Produce:
   1. A short executive summary of the main findings.
   2. The most important recurring themes, with a plain-language description, the number of comments supporting each theme if the data allows it, and representative verbatim comments labelled by row or response ID.
   3. Positive themes and negative themes separately.
   4. The main apparent drivers of high and low CSAT, clearly distinguishing what the comments say from your interpretation.
   5. Differences by product, channel, segment and time period only where the data contains those fields and the comparison is meaningful.
   6. Comments that are ambiguous, contradictory, duplicates or too sparse to support a conclusion.
   7. A list of practical follow-up questions and actions, ranked by likely customer impact and ease of investigation.
   
   Use consistent theme labels. A comment may have more than one theme, so say when totals overlap. Do not present correlation as causation, do not infer demographic or sensitive traits, and do not identify individual customers. Include a short audit table showing each theme and the response IDs assigned to it. Finish with the limitations of this analysis and the checks a colleague should perform before acting on it.

   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.