Write a one-page reply brief (your tone, what you can offer, what you must never mention), paste each review into ChatGPT, Claude or Gemini with that brief, and ask for a two-to-four-sentence reply that picks up one specific detail. Edit the draft, then post it yourself from your Business Profile. Once the brief exists, each reply takes about two minutes.
The brief is where the quality comes from. Without it, every assistant produces the same warm, vague paragraph ("Thank you so much for your kind words! We're thrilled…"), and after twenty of those your profile reads like a vending machine. Two practical points before you start: the reviewer is notified when you reply, and Google checks each reply against its content policies before it appears. Google's help page says that usually takes up to 10 minutes but can take up to 30 days.
Google has also been testing its own AI reply suggestion for a limited set of Business Profile accounts since March 2026. If it shows up on yours, the same discipline applies: it drafts, you decide. The routine below works with any assistant and any number of locations.
Five stages from a new review to a posted reply
The whole setup takes about an hour and a half, most of it spent on the brief. After that, the daily work is a few minutes. Here are the stages in order, with realistic times for a single-location business.
- Get reply access sorted (10 minutes). Only owners and managers of a Business Profile can reply. If a staff member will handle drafts, add them as a manager rather than sharing your Google password. Switch on review notifications in the Business Profile settings so new reviews reach an inbox someone actually reads.
- Write the reply brief (45 minutes, once). One page covering who signs replies, the tone, what you always do, what you never mention, and what you can offer an unhappy customer. There is a filled-in example below.
- Sort each review into a lane (seconds). Five-star with detail, star-only, mixed, negative, or policy-breaking. The lane decides how much editing the draft needs and whether AI should draft at all.
- Draft with a saved prompt (30 seconds). Keep the prompt in a Project (ChatGPT and Claude both have them) or a Gemini Gem (becoming a skill from November 2026) so the brief is loaded every time and you only paste the review.
- Check and post (1-2 minutes). Read it aloud once, run the five checks further down, and post from Business Profile via Read reviews, then Reply.
A filled-in reply brief for a car repair garage
This is the brief an illustrative four-bay car repair garage uses. The owner, two mechanics and a service adviser share the work, and the profile gets 10 to 15 reviews a month. Copy the structure and replace every line with your own rules.
REPLY BRIEF: Google reviews
Who signs: "[owner's first name], owner" on every reply
Tone: Plain, friendly, slightly dry. No exclamation marks,
except at most one in a five-star reply.
Length: 2-4 sentences. Never more than 60 words.
Always: Pick up one specific detail from the review (the job,
the staff member named, the timing).
Use staff FIRST names only, and only if the reviewer
named them.
Never mention: Registration numbers, prices, the customer's surname,
anything about the car's history they didn't write,
discounts, offers, "best garage in town".
Can offer: A call from the owner on the main number.
A free recheck within 30 days on work we did.
Negative: Acknowledge the problem, state at most one fact,
invite them to call. Never argue. The owner rewrites these.
Our words: "car" not "vehicle"; "booked in"; "collected".
Banned openers: "Thank you so much for", "We're thrilled",
"We appreciate you taking the time".The "never mention" list does the most work. AI assistants love to add helpful context, and a reply that mentions the customer's registration or what they paid is the kind of slip that makes a reviewer regret posting. The banned openers matter nearly as much: they are the three phrases every assistant defaults to, and banning them forces a reply that starts with the customer's own detail.
Sorting reviews into lanes before you draft
Treating every review the same way wastes time on the easy ones and under-prepares you for the hard ones. The table shows the garage's lanes and an illustrative month of 14 reviews.
| Lane | Garage's month | What the reply does | AI's job |
|---|---|---|---|
| Five-star with detail | 9 | Thank, echo the detail, sign off | Draft; light edit |
| Star-only, no text | 2 | Short varied line, or skip | Write a bank of one-liners once |
| Mixed (three or four stars) | 2 | Thank for the good part, answer the gripe with one fact | Draft; careful edit |
| Negative (one or two stars) | 1 | Acknowledge, one fact, offer a call | Draft only; owner rewrites |
| Policy-breaking (spam, abuse, never a customer) | 0 | Report first, then a short neutral reply | Match the review to a policy |
Negative reviews and policy-breaking ones get their own tutorials, because the stakes and the steps are different. If a one-star review lands, read how to reply to negative reviews with AI before drafting anything, and if you suspect the reviewer was never a customer, start with handling a fake or unfair review instead.
The prompt, a sample reply and what to fix in it
Save this prompt once, with your brief pasted in, and reuse it for every review. The optional notes line is where you add something from your own records that the reviewer would recognise, such as who did the job.
You draft Google review replies for our business. Follow the brief
below exactly.
BRIEF:
[paste your reply brief]
REVIEW ([n] stars):
"[paste the review text]"
Notes from our records I'm happy to mention (optional):
[e.g. "Clutch job, Tuesday, [technician] did the work"]
Write ONE reply of 2-4 sentences that picks up one specific detail
from the review. Don't use any banned opener. Don't mention anything
that isn't in the review or my notes. Don't offer anything that isn't
in the brief.
Then, on a separate line starting "CHECK:", list anything in the
review I should confirm before posting (names, facts, claims).Here is an illustrative five-star review the garage received, followed by what a typical assistant returned with that prompt.
★★★★★ "Booked in for a clutch that was slipping. [Staff name] on the desk kept me updated by text all day and it was ready by 4. Not cheap but fair, and they showed me the old part."
Illustrative AI draft: "[Staff name] will be pleased the texts helped. A slipping clutch can make for a long, worrying day, so we try to keep everyone in the loop from drop-off to collection. Showing the old part is something we do on every job, and we're glad it helped make sense of the bill. [Owner], owner"
CHECK: Confirm the named colleague is happy to be named in a public reply.
It is close, and it followed the brief: no banned opener, staff first name only because the reviewer used it, no price. Two edits make it better. "A slipping clutch can make for a long, worrying day" is filler the reviewer didn't say, so cut it. "Make sense of the bill" nudges towards the price comment without dwelling on it, which is the right instinct; keep it. The posted version is three sentences and 38 words, and it took the owner about a minute.
The CHECK line is worth keeping in every prompt. It catches the small things that go wrong in public: a staff member who left last month, a reviewer who named the wrong person, or a claim ("they fixed it for free") you'd rather not confirm in writing.
A month of replies at the garage, before and after
Before the routine, the owner answered reviews in a batch on Sunday evenings. Each reply took five or six minutes because he started from a blank box, and some reviews waited three weeks for an answer. Several of his replies began "Thanks for the great review!", which is fine once and odd twelve times in a row.
After the brief and prompt were set up, the illustrative month looked like this:
- Nine detailed five-star reviews: about 1.5 minutes each including edits, so roughly 14 minutes.
- Two star-only ratings: a line from the pre-written bank, under a minute in total.
- Two mixed reviews: about 4 minutes each, because the gripe (a wait for a part, a missed call-back) needed checking against the job sheet first.
- One negative review: 15 minutes, mostly spent reading the job history and phoning the customer before the public reply went up.
That is about 45 minutes for 14 reviews, against roughly 80 minutes before, and every review had an answer within two days. The time saving is modest. The bigger gain is consistency: replies arrive while the customer still remembers the visit, and none of them read as if they came from a template.
Star-only ratings and one-line praise
A rating with no text gives the AI nothing to echo, which is where replies start to look automated. You have two honest options: skip them, or rotate a short bank written once. Ask your assistant for the bank with this prompt:
Write 10 different one-sentence replies to a five-star Google rating
that has no written review. Signed "[owner's first name], owner". Plain and friendly,
no exclamation marks, no "Thank you so much", no mention of prices,
offers or rankings. Each must start with a different word.An illustrative selection from what came back: "Good to see you back on the road, and thanks for the five stars." "Appreciated. We hope the car behaves itself until next time." "Five stars from a customer is the best part of a Monday for the team." Delete any that sound forced, keep six or seven, and never use the same line twice in a row on the profile. For one-line text reviews ("Great service, would recommend"), use the full prompt anyway; there's usually a detail in the date or the service type you can mention from your own notes.
Where to run it: chat window, Google's suggestion or an automation
The prompt works in three places. Which one suits you depends mostly on volume and on whether you want a human to press "Reply" every time.
| Route | Setup | Monthly cost | Suits |
|---|---|---|---|
| Chat assistant with a saved Project or Gem | 30 minutes | $0 on free plans; about $20 on a paid individual plan | Up to about 30 reviews a month |
| Google's built-in AI suggestion (limited test) | None, if your account has it | $0 | Quick drafts, but it can't see your brief |
| Zapier or Make: new review, AI draft, email to the owner | 2-3 hours | Zapier Professional from $19.99 a month billed annually; Make from about $9 | Several locations, or a team that approves drafts |
| Review management platform with AI replies | 1-2 hours | Roughly $75-$125 and up | Businesses that also want automated review requests |
The automation route deserves a closer look, because it is where people overspend or over-automate. Zapier's Google Business Profile integration has a "New Review" trigger and a "Create Reply" action, and Make has matching modules for watching reviews and creating a reply. The safe design keeps a human in the loop: the trigger fires, an AI step drafts the reply using your brief, and the draft lands in the owner's email or team chat for approval.
A quick sum for the garage on Zapier: the trigger costs nothing, an "AI by Zapier" step uses 1, 3 or 5 tasks per run depending on the model tier, and creating an email draft is one more task. That is 2 to 6 tasks per review, or 28 to 84 a month at 14 reviews. It fits comfortably inside the Professional plan's 750 tasks, but not the Free plan, which only allows two-step Zaps and doesn't include AI steps. If you are tempted to wire the "Create Reply" action straight to the AI step so replies post themselves, read whether you should let AI auto-post replies first. For most small businesses, the answer is only for five-star reviews with no complaints in them, if at all.
Replies for a funeral director, where the default tone is wrong
AI's default register is cheerful, and in some businesses cheerful is exactly wrong. Funeral directors are the clearest case. Here is an illustrative review and the unedited draft an assistant produced without a brief.
★★★★★ "They looked after Mum with such dignity. Special thanks to Claire, who guided us through everything when we couldn't think straight."
Illustrative draft, no brief: "Thank you so much for your wonderful review! We're thrilled you had such a positive experience with Claire and our team. We look forward to serving you again!"
Every phrase in that draft jars: "wonderful", "thrilled", "positive experience" and, worst of all, "serving you again". With a brief that bans exclamation marks, bans "experience", and says replies should be quiet, short and signed by the director, the draft became: "Thank you for trusting us to care for your mother. Claire will be very touched to read this. We are here if you need anything in the weeks ahead." That is the version to post. For a business like this, add a line to the brief saying the owner writes every reply by hand when the review mentions a complaint, a child, or anything about how the person died. AI can draft the rest, but those few deserve a human from the first word.
Business-to-business reviews: a packaging supplier's profile
Suppliers get far fewer reviews than garages or cafés, often a couple a month, so each reply is read by a larger share of prospective buyers. The reviewers are usually buyers or founders at other companies, and they sometimes name their own brand. An illustrative review for a packaging supplier: "Ordered 2,000 printed mailer boxes for our candle brand. The proof came back the same day and the colour matched our brand swatch exactly."
Two brief rules matter here. First, echo only what the reviewer made public: if they named their brand, you can congratulate them on the launch; if they didn't, don't mention the order size or the client name from your records, even though the AI would happily use your notes. Second, resist the urge to turn the reply into a sales pitch. An illustrative before and after:
- Before (AI, no brief): "Thanks! We offer custom printed mailer boxes, shipping boxes and eco-friendly packaging for all businesses, with fast turnaround and colour matching. Contact us for a quote today!"
- After (with brief): "Colour matching is the part our print team worries about most, so this made their week. Good luck with the candle launch, and send us a photo of the boxes on the shelf if you get the chance."
The "before" reads like keyword stuffing, and Google's own reply tips say to be conversational, not promotional. The "after" gives a prospective buyer the same information (they colour-match, they care about print quality) in a way that sounds like a real company.
Signs your AI replies are drifting into a template
Replies drift slowly. Nobody notices the fifth reply that starts with "It's great to hear", but a customer scrolling your profile notices the pattern straight away. Once a month, paste your last 20 replies into your assistant with this audit prompt:
Here are my last 20 Google review replies. List:
1. Any opening words used more than twice.
2. Any sentence or phrase that appears in three or more replies.
3. Any reply that mentions something the customer might consider
private (prices, car details, health, family details, order size).
4. Any reply over 60 words.
Quote the reply number for each finding.An illustrative result from the garage's third month: "1. 'Great to hear' opens replies 3, 7, 11, 14 and 18. 2. 'We look forward to seeing you next time' appears in 6 replies. 3. Reply 9 mentions the safety inspection result the customer didn't write about. 4. None." The fix took ten minutes: "Great to hear" and the look-forward sign-off went on the banned list, and the brief gained a line saying never to mention inspection results or test outcomes unless the reviewer did. That third finding is the one that matters most. It came from the notes line in the prompt, where the service adviser had pasted the full job card instead of one detail.
Other warning signs to watch for:
- Keyword stuffing. Replies that repeat your service and town name in every answer. This reads as spam to people and adds nothing a search engine needs.
- Invented promises. "We'll be in touch to arrange a discount" when nobody plans to. The brief's "can offer" list prevents this, provided the prompt says not to offer anything else.
- Answering the wrong review. In automations, a mismatched review ID can post a cheerful reply under a complaint. Another reason to keep a human approval step.
- Replies to your own replies. Some tools treat an edited review as new and draft a second reply. Google's "Updated Review" trigger in Zapier fires separately from "New Review", so decide deliberately whether edited reviews get a fresh draft.
Five checks before you press Reply
These take under a minute once they are habit. Keep them next to the prompt or at the bottom of the brief.
- Is every fact in the reply in the review or your notes? Delete anything else.
- Would the customer be comfortable with everything it says being public? No prices, health details, addresses, registration numbers or order details they didn't share.
- Does it start differently from your last three replies? Glance at the profile.
- Is any offer in it one you will actually honour? If it says "call [name]", that person must be expecting calls.
- Does it sound like the person who signs it? Read it aloud. If it sounds like a hotel chain, cut a sentence.
If you already have a wider review process for AI output, fold these into it; setting up human review for AI work covers how to keep checks quick enough that people actually do them.
Measuring whether the routine is working after a month
Four numbers tell you whether the routine is doing its job. Keep them in a simple sheet: one row per month, filled in from the profile in about ten minutes.
| Measure | How to get it | Healthy sign | If it's off |
|---|---|---|---|
| Reply rate (text reviews) | Replied ÷ reviews with text | 100% | Notifications are going to the wrong inbox |
| Typical time to reply | Days between review and reply, middle value | Two days or fewer | Give drafting to a second person as manager |
| Heavy rewrites | Drafts where you changed more than half | Under 1 in 4 | The brief is missing a rule; add it |
| Repeated openers | From the monthly audit prompt | None used more than twice | Extend the banned openers list |
The heavy-rewrite figure is the most useful of the four. If you are rewriting most drafts, the AI isn't the problem; the brief hasn't told it something you know. Write down what you keep changing ("always mention the recheck", "never say 'team'") and add it as a rule. After two or three rounds of that, most owners find the drafts need only light edits, and the brief becomes something you could hand to a new manager on their first day.
For the detail of Google's own steps, including how to edit or delete a reply, Google's Manage customer reviews help page is short and current. And if your replies need to sound like your brand across other channels too, the same brief grows naturally into a brand voice guide that AI can follow.
Questions owners ask about AI review replies
Does Google penalise replies that were drafted with AI?
Google has not said it penalises AI-drafted replies, and it has been testing its own AI reply suggestion inside Business Profile. What Google checks is the content: every reply is reviewed against its policies before it appears. A reply that is accurate, specific and polite is fine however you drafted it. A reply stuffed with keywords or containing personal details is not.
Should I reply to every single review, including star-only ones?
Reply to every review that has text, positive or negative. Star-only ratings are optional: a short, varied line is fine, and skipping them is also fine. What looks bad is the same sentence pasted under forty ratings. If you reply to star-only reviews, keep a bank of ten different one-liners and rotate them.
Do review replies help my ranking on Google Maps?
Google's own guidance on local ranking says replying shows you value feedback and that positive reviews and helpful replies can help a business stand out. It separately says more reviews and positive ratings can help local ranking. It does not promise that replies themselves move rankings, so treat replies as a trust signal for readers first.
Can I change a reply after it has been posted?
Yes. Open your Business Profile, select Read reviews, and use the edit or delete option under your reply. The edited reply goes through Google's policy check again, so it may take a few minutes to show. Editing is useful when a customer updates their review after you have sorted out a problem.
Further reads
- How to Get More Google Reviews With Automated Review Requests — Set up review requests so there is a steady flow worth replying to.
- AI Review Management Tools for Small Businesses (2026) — Compare tools that draft or post replies across several review sites.
- How to Turn Customer Reviews Into Marketing Copy With AI — Turn the praise you are replying to into website and ad copy.
- How to Optimise Your Google Business Profile With AI — Tidy the rest of your Business Profile while you are in there.
- How to Reply to Hotel Reviews With AI Without Sounding Canned — See how the same approach works in a high-volume review sector.
- How to Write Google Business Profile Posts With AI Every Week — Add weekly posts to the profile alongside your review replies.
- How to Answer Restaurant Complaints With AI Without Escalating — A triage table, a reply prompt and three worked restaurant complaints showing how AI drafts calm replies and where a person must take over.
- Best AI Tools for Cafés and Coffee Shops, Grouped by Task — The AI tools a café can use, sorted by the job they do across the day, with what each costs and where each one lets you down.
- How Small Tour Operators Use AI for Bookings and Itineraries — How a small tour operator can use AI around its booking system, build private itineraries from a route library, and catch the errors before guests do.
- Where AI Actually Saves Time in a Small Restaurant — A task-by-task look at a small restaurant's admin week: where AI cuts real minutes, where it only moves them, and where it adds work.
- 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.
- How Much Time Can AI Save a Barber Each Day? — A minute-by-minute look at a barber's day: which admin AI shortens, which it can't touch, and a two-day tally to measure your own saving.
- 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 Garages Use AI to Send Repair Updates Customers Appreciate — How a small garage can turn technician shorthand into clear, timely repair updates with AI, and where those messages go wrong.
- What AI Can and Can't Do for Your Local SEO — Which local SEO jobs AI does well, which it can't touch, and an afternoon's worked example for a dry cleaner, mapped against Google's own ranking factors.
- What Is Review Gating and Can It Get Your Business Penalised? — Check whether your feedback forms, AI sentiment labels or review reminders quietly exclude unhappy customers.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google Business Profile Help (Manage customer reviews; Tips to get more reviews; Tips to improve your local ranking), Zapier Google Business Profile integration page, Make Google Business Profile app documentation, industry coverage of Google's AI reply test (March 2026).