Table of Contents
Hey creative strategists, AI enthusiasts, and my family + friends that I forced to subscribe,
Happy Friday! Hope your week was more productive than mine, which isn't saying much considering I spent Tuesday arguing with my dog, Linus, about who gets the good spot on the couch. (He won. Again.)
So remember last week when I said there were no new AI models dropping? Well, that was apparently the universe's cue for me to actually USE the tools we already have instead of just collecting them like Pokémon cards (damn, what they would be worth today!).
My Humbling (Beginning? Idk)
Picture this: It's 2 AM, I'm three coffees deep, James is asleep, the shusher is shushing, and I'm using our old systems to analyze customer reviews. Classic Thursday night, really.
Then I realize I’ve been stuck in my ways for too long, so I start reinventing the AI Review Analysis that I have staked my career on and have given multiple talks about. You know, the one where I acted like I knew what I was talking about while secretly having no idea if it actually worked.
Plot twist: The version works. Really well. Embarrassingly well.
Within 30 minutes, Claude Sonnet:
Identified customer personas I'd completely missed
Found emotional triggers buried in reviews that I'd skimmed past
Reverse-engineered our competitors' strategies better than our $5k/month agency
I felt like someone who had just discovered fire while everyone else was already grilling burgers
The Actual Step-by-Step Process (With Real Prompts)
Here's the embarrassingly simple process that's now questioning my entire career:
Step 1: Collect the Data (The Smart Way)
Stop manually copying and pasting reviews like a caveman (that’s how they did it, right?). Here's what actually works:
For Amazon: Use Apify's Amazon Reviews Scraper. It's stupidly simple: just paste the product URL, set your max reviews (I usually grab 1,000+ per product), and let it rip. Apify can return up to 500 reviews per product (100 per star rating) and handles all the technical stuff like avoiding CAPTCHAs.
For Everything Else: GPT-5 is actually incredible at scraping reviews from across the internet. Just tell it what you need:
"I need you to find and collect customer reviews for [PRODUCT/BRAND] from these sources: Reddit, Quora, Trustpilot, Google Reviews, and any other relevant review platforms. Focus on authentic customer experiences and organize the data by platform."
For Shopify: Well, we will get there.
The sources that matter:
Reddit: Real talk, no filter, brutally honest
Quora: Detailed experiences, often from people comparing options
Trustpilot: Structured reviews with ratings
Google Reviews: Local business insights
Facebook comments: Social proof and complaints
Industry forums: Niche-specific feedback
Step 2: Feed Everything to Claude with This Exact Prompt
Once you have your review dump (aim for 500-1000+ reviews), here's the prompt that broke my brain:
You are an expert customer research analyst. I'm going to give you customer reviews for [PRODUCT CATEGORY]. Your job is to identify emotional patterns and customer motivations that most marketers miss.
FIRST, analyze these reviews for:
1. POSITIVE EMOTIONS (what drives purchase):
- Motivations (desired outcomes that prompt purchase)
- Values (perceived benefits/advantages)
- Pain points being solved (industry problems this product addresses well)
2. NEGATIVE EMOTIONS (what creates friction):
- Frustrations (what annoys customers)
- Irritations (minor but recurring complaints)
- Deal-breakers (what makes them never return)
For each emotion category, provide:
- The specific emotion/motivation
- Example quotes from reviews that demonstrate this
- Frequency score (how often this appears, 1-10)
IMPORTANT: Base your analysis ONLY on what customers explicitly say in these reviews. Do not infer or assume anything not directly stated. Use the customers' actual language, not marketing jargon.
AVOID HALLUCINATIONS: If you're unsure about a pattern, mark it as "LOW CONFIDENCE" and explain why.
Here are the reviews: [PASTE YOUR REVIEW DATA]After getting the emotional breakdown, use this follow-up prompt:
Based on the emotional patterns you just identified, create 2 distinct customer personas:
1. EVERGREEN BUYERS: Loyal, repeat customers who consistently engage with this brand/category
2. DIAMONDS IN THE ROUGH: Potential customers using the product in niche ways or showing untapped potential
For each persona, provide:
- Persona name (something memorable, not generic)
- Core emotional drivers from your previous analysis
- Specific language they use (direct quotes)
- Their main use case/context
- What messaging would resonate vs. what would repel them
Format as: Persona → Emotional Drivers → Key Quotes → Messaging Angles
Remember: These should feel like real people, not marketing segments.Step 4: Reverse Engineer Your Competition
Here's where it gets spicy. Take screenshots of competitor ads (Facebook Ad Library, Google Ads, social posts) and upload them with this prompt:
Analyze these competitor ads and identify:
1. WHO are they targeting? (match to personas if possible)
2. WHAT emotional triggers are they using?
3. WHAT unique selling propositions are they highlighting?
4. WHAT format/style mix are they using? (UGC vs. professional, static vs. video, etc.)
For each ad, provide:
- Target persona (based on our previous analysis)
- Primary emotional hook
- Key USP being emphasized
- Production style and why it works for this audience
Then summarize:
- What % of their ads target each persona type?
- What emotional patterns appear most frequently?
- What USPs are they avoiding that we could own?
- What gaps exist in their messaging strategy?
[UPLOAD COMPETITOR AD SCREENSHOTS]The Results That Hurt My Feelings
After running this process on our client's reviews, here's what the AI found that I'd completely missed:
Persona 1:
What I saw with GPT: Overwhelmed Optimists
What Claude found: People buying hope in product form
Emotional driver: Relief from decision fatigue
Persona 2:
What I saw with GPT: Skeptical Settlers
What Claude found: People who've been burned before and want the "safe" choice
Emotional driver: Avoiding disappointment over achieving perfection
The wild part: Our competitor analysis revealed they were targeting these exact personas with 73% of their ads, while we were still chasing "premium quality seekers" who made up maybe 12% of actual buyers.
Advanced Prompts for Deeper Insights
Once you've got the basics, try these for next-level analysis:
For finding messaging gaps:
Compare our brand's current messaging to what customers actually say they value. Identify:
1. Messages we're using that customers don't care about
2. Customer values we're completely ignoring
3. Emotional triggers our competitors are missing
4. Specific language customers use that we should adopt
Provide specific recommendations for messaging pivots with examples.For seasonal/context analysis:
Analyze these reviews for contextual patterns:
- When/where/how is this product typically used?
- What situations trigger purchase vs. regret?
- What external factors (season, life events, etc.) influence satisfaction?
- How does usage context affect what customers value?
Use this to suggest context-specific marketing angles.For competitive advantage identification:
Based on all review data (ours + competitors), identify:
1. What we do better that customers actually notice and care about
2. What competitors do better that we're ignoring
3. What industry pain points NO ONE is addressing well
4. What "obvious" improvements customers keep requesting
Prioritize by: customer impact × competitive differentiation × ease of implementation.TLDR; The Tools That Actually Matter
Data Collection:
Apify + Amazon: Best combo for Amazon reviews, handles up to 10,000+ reviews per ASIN with advanced filtering
GPT-5 for everything else: Surprisingly good at finding and organizing reviews from Reddit, Quora, Trustpilot, etc.
Manual collection for niche sites: Sometimes you still gotta do it the hard way
Analysis:
Claude: Better for nuanced analysis, longer context windows. Period.
Competitor Research:
Facebook Ad Library: Free, comprehensive
TikTok Creative Center: Underutilized goldmine
Google Ads Transparency Center: For search ad insights
Screenshot everything: Still the most reliable method
Total monthly cost: ~$10-25, depending on volume
The Elephant in the Room: Shopify Problem (And Our Solution)
Here's the thing: while there are decent tools for scraping Shopify product data, there's no great automated script for collecting customer reviews from Shopify stores specifically. You've got browser extensions and manual tools, but nothing that scales.
That's actually part of what the SelfMade team is building for Q4. Think of it as "AI review analysis" but fully automated from data collection across ALL platforms (including those hard-to-scrape Shopify stores) to persona identification to competitor analysis to actual ad concepting.
The whole process I just walked you through? Imagine it happening automatically every day/week/month, with fresh data, updated personas, and new competitive insights delivered to your inbox.
👉 Can't say much more yet (lawyers made me promise), but if you want early access when we launch, just hit reply and say:
"I want it"
The Uncomfortable Truth About Your Current Strategy
If you're like me (and statistically, you probably are), you're making these mistakes:
You're targeting who you THINK should buy your product instead of who actually does
You're highlighting features that matter to you instead of benefits that matter to customers
You're using marketing speak instead of customer language
You're guessing at competitor strategies instead of analyzing what actually works
The fix isn't more creativity or better targeting options. It's better understanding of what's already working and why.
Bottom Line
Your customers have been telling you exactly what they want in their reviews for years. You just haven't been listening properly because you're human and humans are terrible at pattern recognition compared to AI.
The good news? Now you know how to fix it. The bad news? You'll realize how much money you've wasted on campaigns targeting imaginary people. The worse news? AI is legitimately better at understanding your customers than you are. The actually good news? Everyone else is making the same mistakes, so you can get ahead just by being less wrong.
Claude is currently king here.
Anyway, take all this for what you will, and have a great weekend!
Will
P.S. If you try this and discover your entire marketing strategy has been backwards, don't @ me. I'm still processing my own existential crisis.
P.P.S. Seriously, if you want early access to the SelfMade tool that automates all this (including those tricky Shopify reviews), just reply. We're looking for beta testers who aren't afraid of having their assumptions challenged by data.