How to write copy using Amazon review mining

Executive overview

Most copy fails because it starts inside someone's head or a boardroom. Amazon review mining replaces guesswork with language real customers have already used.

The method is built around the Problem Agitation Solution (PAS) framework. Know what you're looking for before you start reading reviews, and the whole exercise takes 20 minutes.

Works for product and service businesses alike. If your service isn't on Amazon, find the book people buy instead of hiring you — the language is just as useful.

The core insight: you don't write the copy, you find it — your job is to stitch real customer language into a persuasive structure.

Before you start: set up your PAS target

• Decide what you're writing and who it's for before opening any reviews.
• Map the four slots you need to fill under solution: how people describe it, the specific value they get, how it makes life instantly better, and how it makes life better long term.
• Problems and agitation are easy to find; the solution section is where most copy falls apart — knowing its sub-parts prevents that.
• Write the framework out as a blank template so you're scanning for specific things, not just "interesting stuff."

Finding the right books and reviews to mine

• Search Amazon for books that do the job your product or service does (e.g., project management books for a PM consultancy).
• Open at least six books in new tabs; prioritise titles with 35+ reviews — low-review books skew toward fake or thin feedback.
• Focus on two-, three-, and four-star reviews. Ones and fives exist on extremes; the middle is where honest, specific experience lives.
• Fake reviews are obvious at speed — short, glowing, sounds like the author's friend wrote it. If it looks fake, skip it and move on.
• The goal is not a perfect sample; it's more signal than you'd get from a blank page or an internal meeting.

How to scan: speed and instinct over analysis

• Move fast. You're not reading — you're scanning for language that makes you stop.
• The filter is simple: does this phrasing feel specific, real, and sticky? If yes, clip it. If not, keep moving.
• Don't summarise or interpret while scanning. Your brain will want to; don't let it.
• 20 minutes should yield around 20 usable snippets. If you have fewer after five minutes, you're going too slowly.
• You may go through 30 reviews and find nothing. That's normal. Keep the pace.

Tagging and organising what you find

• After scanning, go back through your clips and tag each one: problem, agitation, solution description, specific value, instant benefit, long-term benefit.
• Tags don't have to be final — a clip tagged as "agitation" might move to "solution" when you start writing.
• Archive anything that doesn't fit any category. Don't force it onto the page.
• If the same phrase or theme appears repeatedly across reviews, surface it to the top of its section — frequency signals hierarchy.
• Once tagged, drag everything into order. Gaps become visible; go back and fill them with another pass of reviews.

Building the page from mined language

• Merge all clips under their PAS sections. The structure of the page emerges from the language, not from your imagination.
• Any strong line can become a headline. "Whether a project sinks or sails falls on the shoulders of the leader" — that came straight from a review.
• Your job as a writer becomes stitching, not inventing. Connecting the clips with transitions is far faster than writing from scratch.
• After assembling, spend 20 minutes cleaning. Then layer in proven formulas for individual elements (headlines, CTAs, subheads) to optimise the details.
• Total time from blank page to draft: roughly one hour.

Applying PAS across contexts

• PAS works at every stage of customer awareness, not just problem-aware prospects.
• Even in unaware stages, you open around a problem — it just doesn't have to be dramatic.
• For emails, the solution can be as short as a single CTA. For long-form sales pages, agitation can run much longer.
• Homepages are harder because you don't know who's there or what stage they're at — PAS still works, but picking the right opening problem is the challenge.
• PAS is a reliable default. Other frameworks (4Ps, AIDA) are valid alternatives when the content demands it.

Where else to mine when Amazon isn't enough

• Other review platforms work: G2, Capterra, Trustpilot, Yelp depending on the category.
• Blog comment sections surface problems and emotional language, especially for B2C topics.
• Reddit threads contain raw, unfiltered voice-of-customer — filter for dramatic or off-topic noise.
• For nascent or niche categories with few reviews, go broader: adjacent categories, substitute products, upstream problems.
• The principle stays constant: find where customers talk openly about problems your solution solves, then listen.

Copyright and practical concerns

• You're lifting a short snippet of language, not reproducing substantial text. This falls comfortably within fair use in most jurisdictions.
• If concerned, look up local fair use rules. In practice, this has not been a real-world issue.
• Think of it as listening in on a public conversation — the same way a copywriter would sit in a shop and listen to customers talk.