As of 13 August 2026, AI can only partly compare customer feedback across your products.

What to actually do

1. Hand it to a person

   The route this page recommends

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

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Export the feedback for each product from your survey, support or review system, and combine the rows into one spreadsheet with a product column.
   2. Remove customer names, email addresses, order numbers and other personal data that the comparison does not need, while keeping the feedback text and any useful product, date, segment and rating fields.
   3. Open a chatbot that accepts file uploads, paste the prompt, and upload the cleaned spreadsheet or CSV.
   4. Check the model's data-quality list against the spreadsheet, correcting product labels, duplicate rows and blank feedback before asking it to rerun the comparison.
   5. Compare each reported count, score, quotation and product assignment with the source spreadsheet, and delete any claim that cannot be traced to a row or clearly marked interpretation.
   6. Ask a product or customer-service colleague to challenge the three findings against known releases, service changes and customer context before sharing the report or changing priorities.

   Prompt

   Compare customer feedback across these products: [PRODUCT NAMES].
   
   I will provide a table or CSV with these columns where available: product, feedback text, date, customer segment, rating or NPS score, channel and issue type.
   
   Tasks:
   1. Check the data for missing product labels, duplicate comments, blank feedback, mixed languages and obvious personal data. List problems before analysing and do not invent replacements.
   2. Compare the products using only the supplied data. For each product, report the number of comments, the available rating or NPS information, the five most common themes, the main positive themes, the main negative themes and any notable changes by date or customer segment.
   3. Separate direct evidence from interpretation. For every important theme, give the product, theme, number of matching comments if you can calculate it reliably, and two short representative quotations. Do not present the quotations as statistically representative.
   4. Identify themes that are unique to one product, shared across products, and materially different in wording or frequency. Say when the evidence is too sparse or ambiguous to support a comparison.
   5. Do not infer customer identity, intent, causes, revenue impact or churn risk unless the data explicitly supports it. Do not invent percentages, scores or dates.
   6. Finish with a comparison table, three evidence-backed findings, three questions for a human product or customer-service team, and a list of every claim that needs checking against the original feedback.
   
   Use plain UK English. Keep customer names, email addresses, order numbers and other personal data out of the analysis where they are not needed.

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