Tool: Claude
Time: 5 minutes
Result: A one-page report of what your customers actually think, backed by their own words.
Your product has 1,000 reviews. You've read maybe 20…probably the angriest ones. The other 980 are telling you exactly what to fix, and nobody is listening.
AI Win #6
Turn your entire pile of customer reviews into a one-page insight report. Claude reads every review and returns ranked complaints, ranked praise, real quotes, and the single change most likely to move your ratings. Works for product reviews, app store feedback, support surveys, or course evaluations.
How To Do It
Collect your reviews in one file. Export a CSV from your platform (Shopify, Amazon Seller Central, App Store Connect, G2), or copy and paste review pages into a plain text file. Rough formatting is fine, Claude doesn't need clean columns.
Go to claude.ai, start a new chat, and attach the file with the plus button (CSV and TXT both work, up to 30MB per file). For a small batch, paste the reviews directly into the chat instead.
Paste the prompt below, replace the placeholder with your product name, and send.
Dig into anything surprising: ask "show me 5 more quotes about [theme]" or "re-run this on 1-star reviews only."
Why This Is Useful
Reading reviews one by one, you remember the loudest voice, not the most common one. Frequency-ranked themes turn anecdotes into priorities, and requiring verbatim quotes keeps the analysis honest. You can check every claim against a real customer's words.
Copy / Paste Prompt
You are a customer insights analyst. I'm giving you customer reviews for [PRODUCT OR SERVICE NAME]. Read all of them and produce a one-page insight report.
Output format:
Snapshot — approximate sentiment split (% positive / neutral / negative) and the 3 words customers use most to describe us.
Top 5 complaints — ranked by how often they appear. For each: a one-line summary, rough frequency, and one verbatim quote.
Top 5 things customers love — same format.
Surprises — 2 or 3 things I probably don't know: unexpected use cases, competitor comparisons, or quiet dealbreakers.
The one fix — if we could change only one thing to improve ratings, what should it be and why.
Rules: quote reviews word-for-word and never invent a quote. Frequency estimates can be approximate but must reflect actual counts. If reviews contradict each other, say so. Ignore generic praise with no specifics.
[ATTACH YOUR REVIEW FILE OR PASTE REVIEWS HERE]Run it on your last 90 days of reviews today!
