Last updated:
Quick answer
Paste 50 to 100 reviews into Claude or ChatGPT and ask for the top complaint themes with counts and example quotes. Fix the pattern, not the review.
At a glance
All outcome figures are estimates from cited sources, not independent verification by us.
- What it is allowed to do
- Read-only
Only reads, researches, or summarizes. It never drafts outbound content or changes anything.
- Time saved (estimate)
- ~3 hours per deep review
- Difficulty
- Intermediate
- Tools needed
- Claude, ChatGPT
- Typical cost
- Free tiers handle ~50 reviews. Larger batches are easier on paid tiers with bigger context windows.
Reviewed Oct 19, 2026
Basis (estimate): Estimate basis: reading and tallying 100 reviews by hand takes ~4 hours; an AI pass plus your spot-check takes under an hour.
What does this setup do?
Analyze a pile of reviews to surface the real recurring complaints and praises, ranked by frequency, with receipts.
Who is it for?
Owners who feel reviews blur together and want to know the actual top three problems instead of reacting to the loudest one.
How do you set it up?
Tools
- Claude or ChatGPT with a large context window (free tiers work for ~50 reviews; Plus helps past 100)
- Your reviews copied as plain text, with star ratings attached
- A spreadsheet or doc to log the themes you will act on
Steps
- Copy 50 to 100 recent reviews into a doc. Include the star rating with each one. Mix 5-star and 1-star.
- Paste them into chat in chunks if needed, and ask for the top 5 complaint themes and top 3 praise themes, each with a count and 2 example quotes.
- Ask which themes are getting worse in the last 3 months, if dates are included.
- Spot-check the counts. Read 5 reviews per theme yourself to confirm the AI grouped them sensibly.
- Pick the top 2 complaint themes you can actually fix. Write one action per theme.
- Share the summary with your team. One page. No blame, just the pattern.
Example output
Illustrative. Your results will differ.
- 5 new reviews found: 4 positive, 1 needs a reply
- Reply draft ready for the unhappy one
- You approve every reply before it posts.
Where to paste this prompt
Claude: Claude → Project knowledge & custom instructions
Copy-paste prompt
Adapt bracketed fields to your situation.
Analyze these customer reviews for my business and find the real patterns. Do not flatter me.
Business: [name, type]
Reviews (with star ratings):
[paste reviews]
Return:
1. Top 5 complaint themes, ranked by how often they appear. For each: theme name, count, and 2 short example quotes.
2. Top 3 praise themes, same format.
3. Any theme that appears more in recent reviews vs older ones.
4. One sentence on what the single most fixable problem looks like.
Rules:
- Count only what is actually in the text. Do not inflate.
- Keep quotes short and exact.
- If a review mentions two issues, count it under both.
- Tell me if the sample is too small to trust any theme.Newsletter
Get the free starter pack
Free starter pack with five beginner copy-paste prompts after you confirm, plus three sourced setups in your inbox each week. Unsubscribe anytime.
What usually goes wrong?
Where agents tend to stall on this job. More patterns in Where AI gets stuck.
- AI miscounts. Always spot-check the numbers against the raw reviews before acting.
- Small samples mislead. Twenty reviews can make one grumpy regular look like a movement.
- Do not confuse correlation with cause. "Slow service on Fridays" might be a staffing problem, not a speed problem.
- Ignoring the praise themes. Your strengths are marketing copy waiting to happen.
Which chat tools work with this prompt?
Paste the prompt into any of these assistants (free or paid tiers).
- Claude
- ChatGPT
Common questions
- How many reviews do I need for this to work?
- Fifty is a decent start; one hundred is better. Under thirty, just read them all yourself. The AI adds little at that size.
- Can I include reviews from Google, Yelp, and Facebook together?
- Yes, and you should. Note the source with each review in case one platform skews negative and drags the whole picture.
- What do I do with the themes once I have them?
- Pick the top two you can fix, assign each an owner and a deadline, and re-run the analysis in 90 days to see if the counts moved.
What tips make it smoother?
- Include the good reviews too. Knowing what people love tells you what to protect while you fix the bad.
- Recency matters. A complaint theme from 2022 that vanished is a win, not a problem.
- One loud reviewer is not a pattern. Demand at least 4 to 5 mentions before calling it a theme.
- Turn the top fix into an operations change, then watch the next 30 reviews to see if it worked.
Where do the claims come from?
Summaries use public sources linked below, not independent verification. How setups are picked.
- Read Clootrack’s write-up (www.clootrack.com)
Clootrack: Breaking feedback into units, detecting recurring phrases, and clustering into themes.
- Read Chisel Labs’s write-up (chisellabs.com)
Chisel Labs: Categorizing complaints, prioritizing by frequency vs severity, spotting patterns.
Related setups
- How to Use AI to Reply to Google Reviews (Even the Bad Ones)
Draft on-brand replies to Google reviews.
- Reply to 1-Star Reviews Without Sounding Like a Robot
Draft calm, human replies to angry 1-star and 2-star reviews that apologize once, offer one real next step, and take the fight offline.
- Ask Happy Customers for Reviews at the Right Moment
Write short, natural review requests for text, email, and in-person handoffs, timed to the moment the customer feels good.