Your Reviews Already Told You What Your Customers Want: How to Read Them With AI

Your customers have been writing to you for months. It's in the reviews, and in the DMs you meant to answer and didn't. Most of it is a tab you keep open.

I run a store, The Daily Paw Blog, live on Shopify, so I have that tab too.

Reviews are the one piece of customer research nobody has to schedule. People wrote them unprompted, in their own words, about a thing they paid real money for. That's the raw material, and you didn't have to run a survey to get it.

In this post I'll show you the two plays I run on raw customer language. Both work in ChatGPT or Claude on the free tier. Which one to reach for on a harder job is its own guide.

Play 1: Paste Your Reviews In and Ask What Keeps Coming Up

You get reviews. Some of them sting. Reading all of them takes an afternoon you don't have, so you fix whatever feels urgent and hope you guessed right.

Copy a batch of reviews, paste them into ChatGPT or Claude, and ask for the pattern. Here's my prompt:

"Based on these reviews, what do customers love most and what do they complain about most? Give me the top 5 of each, and quote the review each one came from."

That last clause is the one I'd fight for. It makes every line point back at a real person's sentence.

How to do it:

1. Open ChatGPT or Claude. The free tier on either is enough for this.
2. Paste in your reviews. A dozen gets you started. A hundred is better.
3. Run the prompt above.
4. Then push once more: "What's the single most common reason someone left less than 4 stars?"

Google reviews, Etsy reviews, DMs, email feedback. If someone wrote it about your business, it counts.

If you have a year of reviews, run the last three months as one batch and the nine before it as another. Same prompt, two answers. Anything in the recent batch that's missing from the older one is either something you changed or something a supplier changed without telling you.

Safety note: paste the review text only. Strip names, emails, and anything personal first.

Nothing about this is hard. It's the sitting-down that never happens. AWS found that more than half of small and medium-sized businesses lack the tools or experience to turn their own data into growth. Here the missing tool is a text box.

Play 2: Ask It to Describe the Person Who Keeps Buying

Try describing your buyer in one sentence, out loud. If it comes out as "people who love their dogs," your next caption is aimed at a stadium.

You have more raw material than you think. Reviews from repeat buyers. Emails from people who tell you what they did with the thing after it arrived. Four examples is enough to start.

How to do it:

1. Gather the good stuff: repeat-customer reviews, order notes, the best DMs.
2. Paste them in and ask: "Based on these, describe the type of person who keeps buying from me. Include their likely age range, what problem they're solving, how they talk about that problem, and what almost stopped them from buying."
3. Read it against who you thought you were selling to.
4. Lift the exact phrases. Those go in your next caption and your next subject line.

The "what almost stopped them from buying" line is the one I'd never skip. It hands you the objection to answer, in the words of someone who bought anyway.

One judgment call here: feed it nothing but five-star reviews and you'll get a flattering portrait of a person who doesn't exist. Include the customers who bought twice and still had something to say. They tell you what the product has to survive.

There's a second use for that top-5 complaint list: replying. 75% of businesses don't respond to their negative reviews, and companies that respond to at least 25% of their reviews earn 35% more than those that don't. Your complaint list is your reply queue, already ordered.

The Part I'd Warn You About

The AI will hand you a pattern whether one exists or not. It's the same lean toward a flattering answer that shows up when you ask AI to review your website instead of your reviews: it favors telling you what you already believe.

Ask for "the top 5 complaints" from a thin batch and you'll get five, even when three of them showed up once. The fix costs you one more prompt: ask it to put a count beside every theme, "how many separate reviews mentioned this?" Then open a few of the cited quotes and read them in place.

One mention is somebody's bad Tuesday. Fourteen means you have a real problem, and now you know its name. If it won't give you counts, treat the list as a reading order and verify it yourself.

The AI will also flatten your one furious review, the one with the actionable detail, into "some shipping complaints." Summaries reward the middle of the batch. The useful outlier is the customer who got specific: the box arrived crushed, the paint numbers didn't match the printed key.

Specificity is the first thing an average throws away. So read the one-star reviews yourself and let the AI take the ninety you'd never open.

Quick Recap

• Your reviews and DMs hold the customer research you were going to pay for.
• Paste review text into ChatGPT or Claude and ask for top complaints and top praise, with the quote behind each.
• Feed your best-customer feedback in and ask who keeps buying, including what almost stopped them.
• Write your captions and subject lines in the phrases your customers already used.
• Check any claimed pattern against the quotes, and read the one-star reviews yourself.

Start Here

If the reviews are piling up and a summary is as far as you get, that's the gap to close. Deciding what matters stays yours. I build the thing that runs it.

At daisyguti.ai/work-with-me, there's a short intake: about nine questions, a few minutes. I read every submission myself, assess whether a review workflow is a fit for your store, and reply with what to hand off first. I'm a 20+ year engineer and I build these systems for small business owners.

Sources

1. AWS SMB Blog - "Why Small and Medium Businesses Are Missing Out on the Full Benefits Data Can Provide": https://aws.amazon.com/blogs/smb/why-small-and-medium-businesses-are-missing-out-on-the-full-benefits-data-can-provide/
2. Exploding Topics - "81 Online Review Statistics": https://explodingtopics.com/blog/online-review-stats
3. OpenAI Help Center - ChatGPT Free Tier FAQ: https://help.openai.com/en/articles/9275245-chatgpt-free-tier-faq
4. Claude plans and pricing (free tier listed at $0): https://claude.com/pricing