Turn 50 customer reviews into 14 ad angles with Claude

What they did

Mo wanted the words happy customers use when they explain why they bought. So Mo pulled a product's reviews into a spreadsheet with Apify, a web-scraping tool, and kept only the 4- and 5-star ones.

Claude then cleared out the empty praise and kept reviews that named a specific result, use or feeling. About half the pile went. The rest got grouped into four themes: problems solved, benefits noticed, objections removed and reasons people chose the product.

For each theme, Claude drafted three ad hooks in the customers' own language plus one landing-page headline. Your happiest customers have already written your marketing; Mo just let Claude do the sorting.

How it went

By Mo's count, 50 reviews became 14 ad angles in about 25 minutes, and Mo picked 3 ads to try out (source).

The first attempt was a miss. Claude lumped almost everything into one "Great product" bucket, and Mo traced it to noisy input rather than the tool. Cleaning the reviews first made the biggest difference, by Mo's account.

Mo hasn't shared how the ads performed. Mo does add a caution: "AI can find opportunities. It can't tell you which angle will win."

Worth knowing

• A review in an ad becomes a testimonial. The FTC says consumer reviews featured in your advertising or marketing count as testimonials, so its rules on testimonials apply (FTC Q&A).
• Never invent or reword in a way that changes the meaning. The FTC's rule, in force since October 21, 2024, bans fake reviews and testimonials, including AI-generated ones, and businesses that knowingly break it can face civil penalties (FTC, 16 CFR Part 465). If AI tidies a quote, put the customer's real words back.
• Ask before you quote. The FTC confirms you can ask customers to let you quote them in ads (Endorsement Guides Q&A). Asking first also saves an awkward message later.
• Watch for "best case" results. A review that promises a specific result ("lost 10 pounds", "doubled my bookings") reads as what everyone can expect. Either have proof it's typical or say what results people generally get (Endorsement Guides Q&A, 16 CFR Part 255).
• Cherry-picking has limits. Using only glowing reviews in marketing is allowed, but the FTC warns that non-representative reviews can still mislead (FTC Q&A).
• Strip personal details first. Remove surnames, emails, phone numbers and order numbers before pasting reviews into a chatbot. Claude and ChatGPT both let you choose whether your chats help train their models (Anthropic, OpenAI).

Paying for Claude? See cheaper Claude alternatives.

Try it yourself

Paste this into Claude or ChatGPT (both have free plans), then paste 10 to 50 real reviews underneath. Remove names and contact details first.

I run a small business. Below are real customer reviews.

1. Ignore reviews that only say something vague like "great service"
   or "fast delivery". Keep ones that mention a specific result,
   use, problem or feeling.
2. Group the reviews you kept into themes, such as problems solved,
   benefits noticed, worries that went away, and why people chose us.
3. For each theme, write 3 short social post hooks and 1 caption
   (under 40 words) using the customers' own wording.
4. For each hook, show the exact review line it came from,
   copied word for word.

Rules: never invent, merge or change a customer's words. Don't add
claims that aren't in the reviews. No customer names.

Reviews:
[paste reviews here]

More people who did this

• @Mike_Stelzner (Michael A. Stelzner, founder of Social Media Examiner): had Claude study the Social Media Examiner team's video ads, then cut a four-minute testimonial into a 9:16 vertical video ad, using ChatGPT for the end card and Gemini for the music. Stelzner puts it at about 90 minutes. Oct 15, 2026.
• @lkr (Laura Roeder, founder of Paperbell): set up a /testimonial command in Claude Code that files screenshots of nice ad comments into the right Google Drive folder, named after the person, ready to use as testimonials. Feb 13, 2026.

Original post

See @Mo_ali's original post on X → (Jul 15, 2026)

Filed under Marketing & Email hacks · Online Sellers · Restaurants & Cafés