Export your reviews into a spreadsheet, ask AI to group them by the reasons customers give for choosing and recommending you, then write headlines, service-page copy and ads from those reasons in customers' own words. Quote a review word for word only with the reviewer's permission, and never let AI tidy up a quote.
There are two separate uses here, and mixing them up causes most of the trouble. Mining reviews for language and themes, then writing your own copy from them, needs no one's permission and is where most of the value sits. Quoting a named reviewer is a different act: Google's own guidance for businesses says reviews belong to the people who wrote them, even on your listing, and that you need the reviewer's consent to use their review in your marketing.
Mining the language versus quoting the reviewer
Keep these apart from the start, in the spreadsheet and in your head.
| Mining the language | Quoting the reviewer | |
|---|---|---|
| What you publish | Your own copy, shaped by the words and reasons that recur | A review, word for word, with a name or initial |
| Permission | Not needed; you aren't reproducing anyone's review | Needed from the reviewer |
| Where it goes | Headlines, service pages, ads, emails, leaflets | Testimonial sections, case studies, social posts |
| AI's role | Finding themes, counting them, drafting copy | Finding candidate quotes; never editing them |
| Main risk | Claims that go further than the reviews support | Using a quote without consent, or altering it |
Getting every review into one sheet
AI can only find patterns in reviews it can see, so the first job is dull: gather them. Include every review, not only the five-star ones. The negative reviews tell you which promises you can't make yet, and the lukewarm ones often contain the most specific language.
Where the reviews come from:
- Google Business Profile. There's no one-click export in the profile itself, so for a few dozen reviews, copying them into a sheet takes about 20 minutes. Review platforms connected to your profile usually offer a CSV export.
- Review sites and marketplaces. Most business dashboards have a download option; check the export or reports menu.
- Your own inbox. Thank-you emails and cards are reviews too, often warmer and more detailed than anything posted publicly. They are also the ones most in need of permission before you quote them.
Use one row per review and these columns: date, source, stars, review text, reviewer's display name, service or product mentioned, and a "permission to quote" column that starts empty. Two illustrative filled-in rows from a care agency's sheet:
date | source | stars | text | reviewer | service | permission
2026-03-14 | Google | 5 | "Mum has had the same two carers since | J. Evans | home visits | not asked
| | | January, which means everything to her. | | |
| | | The office always picks up." | | |
2026-04-02 | Email | 5 | "When Dad came out of hospital at two days' | (family) | discharge | not asked
| | | notice you had someone there that evening." | | |
If you'll paste the sheet into a general chat assistant, a business plan is the better choice, because business plans don't train on your content by default. Reviews are already public, but the emails and cards in your sheet aren't.
The theme prompt and what it gives back
Paste the review text column (not the names) and use a prompt that forces the AI to show its evidence.
Below are [n] customer reviews of a [business type], one per line,
numbered.
1. Group them into 5-8 themes: the reasons customers give for
choosing us, staying with us or recommending us.
2. For each theme give: a short name, the number of reviews that
mention it, the review numbers, and 3 short phrases copied
EXACTLY from the reviews.
3. List separately any complaints or reservations that appear more
than once.
4. Don't invent phrases. If a phrase isn't in the reviews word for
word, don't include it.
Reviews:
[paste]
Illustrative output, first three themes, from 86 reviews of a home-visiting care agency:
Illustrative AI output: "1. Same carers, week after week (31 reviews: 2, 5, 9…). Phrases: 'the same two carers', 'she knows them by name now', 'not a different face every day'. 2. The office answers (18 reviews). Phrases: 'the office always picks up', 'someone rings back the same day', 'I never feel fobbed off'. 3. Carers who notice things (22 reviews). Phrases: 'spotted Dad's ankle was swollen', 'they chat, not just do the tasks', 'she noticed Mum hadn't eaten'."
Check two things before trusting it. First, the counts: assistants miscount often. Filter the review column in your sheet for a key word or two ("same", "carer") and see whether the numbers are in the right range. In this illustration, the "same carers" theme really had 26 reviews, not 31; the AI had counted five reviews that mentioned "the same day". Second, spot-check three phrases per theme against the original reviews. If one was paraphrased rather than copied, rerun that theme and remind the AI of rule 4.
A care agency turns 86 reviews into a new homepage
Follow the illustrative care agency from sheet to published copy. Before the review work, its homepage headline read: "Person-centred care that puts you first." Every care agency could write that sentence, and many do. The reviews said something much more specific, repeatedly, in words families actually use.
The owner took the top three themes (same carers, an office that answers, carers who notice things) and asked for headline options with this prompt: "Using only the themes and exact phrases below, write 8 homepage headlines under 10 words for families looking for home care for a parent. Plain words, no 'person-centred', no 'bespoke', no 'excellence'." From the eight, the one that went live was: "The same carers every week, and an office that picks up." Underneath, three short sections followed the themes, each ending with a line from a review, but only after the reviewer said yes (more on that below).
The before and after, side by side:
- Before: "Person-centred care that puts you first. Our dedicated team delivers high-quality, compassionate care tailored to your loved one's individual needs."
- After: "The same carers every week, and an office that picks up. We keep carer rotas small so your mum or dad sees familiar faces, not a new one each visit. If something changes, you ring us and a person answers, most days within a few rings."
Look at what the "after" does and doesn't claim. "Keep carer rotas small" is a practice the agency actually follows; it came from asking the manager how the "same carers" result is achieved. "A person answers" is backed by 18 reviews and by the office's phone policy. What the copy avoids is a promise the reviews don't support, such as "always the same carer", because the complaints list showed four reviews mentioning holiday cover by unfamiliar carers. Setting up the sheet took about two hours; the theme work and new homepage copy took an afternoon.
Six kinds of copy from a single theme
One strong theme can supply copy for most of your channels. Here is the "same carers" theme turned into six formats, all illustrative, each with the constraint that matters for that channel.
- Service page section heading: "Familiar faces, not a different carer each day." Echoes a customer phrase closely without quoting anyone.
- Search ad headlines (responsive search ads allow 30 characters per headline): "Same Carers Every Week" (22), "An Office That Picks Up" (23), "Carers Who Notice Things" (24).
- Search ad description (90 characters): "Small carer teams, so your parent sees familiar faces. Call and a real person answers." (86)
- Social post opener: "The thing families mention most in their reviews isn't what we expected. It's not the care plan. It's seeing the same two faces at the door every week." The claim about "most" has to match your counts.
- Leaflet line: "Ask any of our families what matters: it's knowing who's coming."
- Email subject line for enquiries: "Who will actually be visiting your mum?"
If the drafts come back sounding like every other agency, the prompt probably let the AI reach for its stock vocabulary. Feed it more exact phrases and ban the words you keep deleting; why AI marketing copy sounds generic goes through the usual culprits.
Asking permission to quote, and keeping a consent log
For Google reviews, Google's guidance suggests the simplest route: reply to the review and ask, because Google won't give you the reviewer's contact details. A reply that works, in public and short:
Thank you, this means a lot to the team. Would you be happy for us
to use your review (with your first name and initial only) on our
website and leaflets? A simple yes in a reply here is enough, and
no is completely fine.
For thank-you emails and cards, ask by email:
Subject: Could we share your kind words?
Hi [name], thank you again for your message about [detail]. We'd love
to include this sentence on our website:
"[exact quote]"
We'd show it as [first name, initial / "a family member"]. Please
reply yes or no; either is fine, and you can ask us to take it down
at any time.
Log every answer in the permission column: yes or no, the date, how they replied, and how they'd like to be named. Names and family details are personal data, so the log is also your record of what each person agreed to. An illustrative log entry: "J. Evans, yes, 2026-05-02, replied on Google, show as 'Jenny, daughter of a client', website and leaflet only." Honour the limits: agreement for the website isn't agreement for a paid ad.
Showing Google reviews on your own site without breaking the rules
When you do quote Google reviews or show your rating, Google's marketing guidance has a few specific rules worth following to the letter:
- Describe ratings as being "on Google", not "Google-rated" or "Google rating".
- Add an "as of" date when you show your overall rating or number of reviews: "4.8 stars from 86 reviews on Google, as of May 2026".
- Don't put stars next to the Google name or logo, and don't use unofficial Google badges.
There's also a search-results rule that surprises people. Google's structured data guidelines say that when a business controls the reviews about itself, its pages using LocalBusiness or Organization markup aren't eligible for star ratings in search results. They also say not to aggregate reviews or ratings from other websites in that markup. So adding review markup to a testimonial page won't produce stars under your listing, and copying your Google rating into markup is against the guidelines. Show testimonials for people, not for the markup.
The wider legal line on testimonials, including what counts as misleading when you select or edit them, is covered in using AI to write testimonials and where the legal line is. The short version: never alter a quote's meaning, and never present a composite as one person's words.
Writing from a funeral director's reviews without exploiting grief
Some sectors have reviews full of emotion, and the temptation is to turn raw grief into marketing. A funeral director's reviews might contain phrases like "they made the worst week of our lives bearable" and "nothing was too much trouble". The first is a quote that belongs to a family at a vulnerable moment; even with permission, many directors decide it shouldn't appear in an ad. The second is a theme you can write from without quoting anyone.
An illustrative theme result for a small funeral director, from 54 reviews: "Calm and unhurried (19 reviews): 'never rushed us', 'gave us time to decide', 'no pressure at all'. Explained costs clearly (12 reviews): 'no surprises on the bill', 'went through every item'." The copy that came from it: "We'll never rush you. Take the time you need, and we'll explain every cost before anything is agreed." That line does commercial work, since clarity on costs is a real worry for families, and it doesn't reproduce anyone's pain. For this kind of business, add a rule to your theme prompt: "Don't suggest quotes that describe the deceased or the family's grief." If you want a family's words for a testimonial, ask a few weeks later, and accept a no without follow-up.
A laboratory's reviews: checking claims before they become promises
Reviews praise what happened to one customer. Marketing copy promises it to everyone. An illustrative water and food testing laboratory found a strong theme in its reviews: "fast results" (15 of 40 reviews, with phrases like "results in two days" and "quicker than the lab we used before"). The AI's headline suggestion was "Results in 48 hours, guaranteed."
That headline would have been a problem. When the lab manager checked the turnaround log for the last three months, the middle value for standard tests was three working days, and one test in five took longer, usually because samples arrived after the daily cut-off. The reviews came from customers whose samples arrived early in the week. The copy that went live: "Most standard results within three working days of receiving your sample. Need it faster? Ask about our express option." It's less exciting, and it's true, which is the version customers will review well next time.
Any claim built from reviews that includes a number, a time, a guarantee or a comparison with competitors should be checked against your own records before it goes live. The routine in fact-checking AI marketing copy works for review-based claims too.
Where AI goes wrong with review-based copy
These are the failures to look for when you edit, each with the way it usually shows up.
- The invented quote. Asked for "testimonials we could use", an assistant will write convincing ones that no customer said. An illustrative example: "'The best care agency we've ever used' — Margaret, 72". There was no Margaret. Never ask AI to write testimonials; ask it to find quotes, then check each against the source.
- The composite. Two reviews merged into one smoother sentence. It reads better and it isn't anyone's review. Copy quotes from the sheet yourself, not from the AI's output.
- Upgraded language. Customers say "really helpful"; the AI writes "life-changing". Keep the customer's modesty. It's more believable.
- A review from the wrong business. When copying reviews by hand, it's easy to pick up one from a similarly named business in the search results. Check the source link for any review you plan to quote.
- Themes that ignore the complaints. If three reviews praise quick call-backs and four complain about slow ones, "we always call back" is not a theme; it's a problem to fix first.
Checking the review-based copy earns its place
New copy should be judged by what it does, not by how much better it reads. The fairest test for a small business is a straight comparison over the same length of time. For the care agency, that meant noting homepage enquiries (form fills plus calls from the website number) for the eight weeks before the change and the eight weeks after, and running the three review-based ad headlines alongside the old ones so the ad platform showed which drew more clicks.
A quick sum shows why eight weeks, not two: if the page brings in about three enquiries a week, two weeks gives you six enquiries either side of the change, and the difference between six and eight is chance. Over eight weeks, 24 against 34 starts to mean something. Note anything else that changed in the same period, such as a price rise, a new ad budget or a local news story, before crediting the words. If the new copy loses, go back to the sheet: the theme may be right and the phrasing wrong.
Once the sheet exists, keep adding to it monthly. The themes shift as the business changes, and fresh reviews keep the copy honest. The same sheet also makes a good starting point for longer pieces; turning one customer story into a month of marketing shows how to build out from a single review once the reviewer has agreed.
Further reads
- How to Write Case Studies From Customer Interviews Using AI — Go deeper with one customer when a review isn't enough.
- How to Build a Brand Voice Guide That AI Can Follow — Blend customers' phrases into a voice guide AI can follow.
- How to Write Your Website Copy With AI: Home, About, and Services — Put the new review-based copy into your main website pages.
- How to Get More Google Reviews With Automated Review Requests — More reviews coming in means more material to mine.
- How to Write Facebook and Instagram Ad Copy With AI — Turn the strongest review themes into social ad variations.
- How to Build a Landing Page With AI That Converts — Use review themes as proof on a page built for one action.
- How to Write Menu Descriptions With AI That Still Sound Like You — A five-stage method for AI-drafted menu descriptions that keep your kitchen's voice, with a voice card template, a drafting prompt and a before-and-after dish.
- What Are Diners Really Saying? Using AI to Analyse Your Reviews — A method for turning months of restaurant reviews into a few counted, checked themes, with a codebook, a tagging prompt and a worked nine-month example.
- How to Reply to Hotel Reviews With AI Without Sounding Canned — Why AI review replies read as templated, and the voice sheet, reply shapes and prompt that make each one sound written by the person who runs the place.
- ChatGPT Prompts for Pet Sitters and Dog Walkers That Win Clients — Twelve ready-to-paste prompts for pet sitters and dog walkers, from profile bio to daily updates, each with what to feed in and what to check.
- How to Write Shopify Product Descriptions With AI That Convert — A Shopify-specific method: fact sheet first, a prompt that answers buyers' real doubts, the right field for every line, and a fair before-and-after test.
- How to Get Your Products Recommended in ChatGPT Shopping — How ChatGPT picks products in 2026, how your data reaches it from Shopify, Etsy or your own site, and the fixes that make a recommendation more likely.
- How Small Amazon Sellers Use AI for Listings and Reviews — How a small Amazon seller can use AI for listings under the 2026 title rules, analyse reviews for fixes, and grow reviews without breaking policy.
- How Florists Can Use AI for Orders, Reviews and Social Posts — Worked examples for a flower shop: a rushed phone order turned into a docket, a substitution message, a year of reviews analysed, and accurate captions.
- Getting More Reviews After Every Job With AI Follow-Ups — When to ask for a review after each kind of trade job, the AI prompt that makes each request personal, and the Google rules that catch firms out.
- How to Use AI to Respond to Google Reviews — A reply brief, a reusable prompt and a two-minute routine for answering every Google review with AI while still sounding like the owner.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google Partner Marketing Hub guidance on using customer reviews (consent, attribution wording, rating dates, logos); Google Search Central review snippet guidelines (self-serving reviews, third-party ratings); Google Ads responsive search ad character limits.