☕🤖Tutorial: Turn Reviews Into High-Converting Messaging (With AI)
PLUS: check all prompts and resources to build this system...
Hey AI Breakers 👋
Most marketing doesn’t fail because your copy is “bad”. It fails because you’re guessing.
Guessing:
what customers actually care about
why they buy
what words they use
what they hate about alternatives
what finally pushed them to purchase
Meanwhile, the answers are sitting in plain sight:
✅ reviews
✅ Reddit threads
✅ competitors’ testimonials
✅ app store comments
✅ G2/Capterra
✅ YouTube comments
✅ support tickets / chat logs
Today, you’ll build an AI Customer Research Engine that turns raw customer language into:
✅ clear pain points
✅ emotional triggers
✅ objections + rebuttals
✅ messaging pillars
✅ positioning statement
✅ headline and angle library
Let’s build it 👇
🧠 How the Engine Works
This is the flow:
Reviews → Find Patterns → Why They Buy → Messaging → Positioning → Copy angles
You’re basically doing what elite marketers do… but in 45 minutes instead of 2 weeks.
All you need is input data.
🧾 Step 0 → Collect the Raw Inputs (10–30 minutes)
Aim for 30–100 snippets total.
Mix sources if you can:
For your product
reviews (Shopify, Trustpilot, Amazon, App Store)
testimonials
sales call notes
support tickets
live chat logs
For competitors
G2 / Capterra / TrustRadius
Product Hunt comments
Reddit threads
YouTube comments
“alternative to X” blog comments
Format tip: paste as a simple list with:
Source
Star rating (if relevant)
Review text
Example:
(G2, 2-star) “The UI is clunky and onboarding took forever…”
(Amazon, 5-star) “Saved me 2 hours a day because…”
Once you have your raw dump, we run prompts 👇
🔎 Prompt #1 → The Review Cleaner (turn messy text into usable data)
If you paste reviews directly, you’ll often get noise:
off-topic comments
vague praise
no clear “why”
This prompt cleans and structures everything into a dataset you can actually use.
✅ Run this first every time.
Prompt:
You are a customer research analyst.
I will paste raw reviews/comments.
Your task:
- Remove irrelevant or unusable lines (say why)
- Convert each remaining review into a structured row with:
1.sentiment (positive/neutral/negative)
2.customer type (guess if not explicit)
3.situation/context (what was happening in their life/business)
4.pain/problem they mention
5.desired outcome
6.feature or benefit referenced
7.emotional tone (frustrated, relieved, excited, etc.)
8.exact customer phrases worth saving (quotes)
Output as a clean table.
Here are the raw reviews:
[paste]💡 Tip: If you have multiple sources, label them so you can compare (your product vs competitor).
🧠 Prompt #2 → The Pain Pattern Finder (find themes that actually matter)
Now we look for patterns. Not “people like it”.
We want:
repeated complaints
repeated outcomes
repeated switching triggers
hidden anxieties
This is where your messaging comes from.
✅ Use this to create your “research summary”.
Prompt:
You are a senior insights strategist.
Using this structured review table:
[paste output from Prompt #1]
Deliver:
- Top 10 pain points (ranked by frequency and intensity)
- Top 10 desired outcomes (ranked)
- Top 5 “switch triggers” (what made them change tools or finally act)
- Top 10 objections and fears (what almost stopped them)
- Top 10 moments of delight (what surprised them positively)
- The 10 most valuable customer phrases (verbatim) that should be used in marketing
Make it specific and written like a research debrief.🎯 Prompt #3 → Jobs To Be Done Map (the real “why they buy”)
Most brands sell features. Customers “buy” products for a job.
This prompt turns your patterns/pains into a JTBD map you can build positioning around.
✅ This is where your offer becomes sharp.
Prompt: