Give the AI a voice sheet written from your own best replies, a short list of facts it may use, and three hard rules: no admission of fault, no finance figures, no customer details. Put every draft through a voice check before a manager posts it. The result is a library of reply templates, not one-off replies.
Responding to dealership reviews goes wrong when the AI knows only the star rating, so it writes the average dealership reply. Better inputs fix that faster than a cleverer prompt: one detail from the review, your own phrasing, and clear limits on what a public reply may say. A one-star finance review shows the stakes. An unguided draft explains the fees and turns a private dispute into a public argument, while a safe reply names no figure and moves the conversation to a named manager.
Why dealership replies sound canned, and the three causes
A canned reply fits any dealership, which is exactly why it reads as nobody's. There are three causes, and each has a fix that doesn't need a better AI model.
- No specifics. Given "write a reply to this 5-star review", an assistant produces the average of every dealership reply it has seen. The fix is a rule that every reply must contain one detail taken from this review, in your words.
- Form-letter apologies. Phrases such as "we apologise for any inconvenience caused" and "your satisfaction is our top priority" sound like a legal department. The fix is a never-say list in the voice sheet.
- Identical structure. If every reply opens with "Thank you for your review" and closes with "we look forward to serving you again", a reader scanning five of them sees a template. The fix is opener rotation and a sign-off from a named person.
The difference shows up on one made-up five-star review about buying a used hatchback with no pressure and a quick handover. The canned reply first:
Thank you for your 5-star review! We're thrilled you had a great experience at our dealership. Your satisfaction is our top priority and we look forward to serving you again!
And a reply written to the rules set out below:
Thank you for taking the time to write this. It's good to hear the handover was quick and nobody pushed you, because that's how we want every sale to go. I'll pass your comments on to the team. Enjoy the hatchback.
[First name], General Manager
A quick way to see the difference is the swap test. Replace your dealership's name with a competitor's and read it again. The first reply still works, so it is canned. The second still mentions a hatchback, a handover and a manager, so it doesn't fit anyone else. Google's own tips agree with the shorter style: keep replies short and simple, don't use them to offer deals or promotions, and don't share a reviewer's private information.
Three rules no public reply may break: fault, figures, files
A review reply is public, permanent in practice, and read by future buyers, by the reviewer, and sometimes by a solicitor. Three rules keep it safe, and they belong in the prompt as hard rules that no style instruction can override.
- Never admit fault, cause or liability. You rarely know every fact when you reply, and a public admission can be quoted back in a dispute. Apologise for the experience, not for a cause. Ask your insurer or solicitor for a short list of phrases they are comfortable with, and put it in the prompt.
- Never quote finance or money. That means no rates, monthly payments, deposits, fees, trade-in values, prices or discounts, even if the reviewer mentioned them. Figures depend on one customer's deal and credit, they may be wrong or unfair to quote, and in some places statements about credit terms can fall under advertising rules, so ask your compliance adviser. Move numbers to a private conversation with a named person.
- Never reveal or add customer details. Don't confirm a purchase, a vehicle, a service history or a date beyond what the reviewer wrote. Google's tips say never to share a reviewer's private information. The same applies to what you feed the AI: paste the public review text, not the deal file. The tutorial on what customer data never to paste into ChatGPT lists what to keep out.
| What breaks the rule | What to say instead |
|---|---|
| "That was our mistake" or "we should have called you" | "I'm sorry the visit didn't go the way you expected. I'd like to look into it properly." |
| "We can get your payment down to [figure] a month" | "Please call [name] on [number] and we'll go through the numbers with you in private." |
| "Your trade-in was valued at [figure] because of the condition" | "I'd like to talk through the appraisal with you directly. Please contact [name]." |
| "The repair on [date] was done to spec, according to our records" | "I'd like to hear your side and go through the work with you. Please contact me." |
| "We'll fix it free of charge" | "[A commitment you can keep, such as: someone will call you back the same working day.]" |
The last row matters because a public promise is a commitment the whole team must honour. Keep a short list of commitments a manager may offer, and let the prompt choose only from it.
A voice sheet built from your own best replies
The AI can only copy a voice it has seen. Collect six to ten of your best real replies, the ones a customer thanked you for or that you'd be happy to see on screen. If you have fewer, write three yourself. Paste them into a chat and ask the AI to describe the voice as eight short rules, quoting phrases from your examples. Edit its answer, add the never-say list and the facts list, and you have a sheet that every draft starts from. The tutorial on building a brand voice guide that AI can follow covers the general method. A filled-in sheet for a made-up family-run used-car dealership with a workshop looks like this:
VOICE SHEET: replies to Google reviews
WHO SIGNS: The department manager, first name and role, on a new line. Replies are written as "I" for the manager and "we" for the team.
HOW WE SOUND: Plain and warm, like a manager talking across the desk. Short sentences. Contractions are fine. One exclamation mark at most, and only on a positive review. No emojis.
ALWAYS: Include one specific detail from the review in our words. For neutral and angry reviews, give a named contact route. For angry reviews, add one commitment from the list below.
NEVER SAY: valued customer; your satisfaction is our top priority; we strive; we pride ourselves; we apologise for any inconvenience; please do not hesitate to contact us; going forward; as per our policy; we look forward to serving you again.
NEVER WRITE IN PUBLIC: prices, rates, payments, fees, trade values, discounts; vehicle or deal details the reviewer did not write; anything that accepts fault; staff discipline; refunds or repairs promised.
FACTS WE MAY USE: Service advisers show customers worn parts before replacing them. Service contact route: the service desk phone or email. Sales and finance contact route: the general manager's phone or email.
COMMITMENTS A MANAGER MAY OFFER: A manager will call back the same working day. A manager will review the job with the workshop team and come back to you. The general manager will arrange a time to go through the paperwork in private.
SAMPLE REPLIES WE LIKE: [paste three or four real ones here]Test the sheet before you trust it. Run five of your old reviews through it, including one angry one, and compare the drafts with what you actually posted. If the drafts could be signed by a different dealership, the sheet needs more of your own phrasing.
The drafting prompt that carries the voice and the rules
The standing prompt goes in a Project (or the equivalent saved-instructions feature in your AI tool) alongside the voice sheet, so every manager starts from the same one. Note the input rule: the AI receives only the public review, the stars and the department, never a deal file.
You draft public replies to Google reviews for a car dealership. A manager edits and posts every reply. The voice sheet is attached.
INPUT: the review text, the star rating, the department (sales, service or finance) and the signer. Nothing else. You will not be given customer files, vehicle details or prices. If the review contains them, do not repeat them.
HARD RULES
1. Never admit fault, cause or liability. Apologise for the experience, never for a cause. Never use: our fault, our mistake, we should have, we failed, that's on us.
2. Never write a price, rate, payment, deposit, fee, trade value or discount, even if the reviewer did.
3. Never confirm or add any detail about the customer, vehicle, purchase, repair or dates beyond what the review itself says.
4. Never argue, never say what our records show, never promise a refund, repair or outcome.
5. No offers, promotions or links.
6. If the review alleges fraud, deception, illegal conduct or a safety defect, mentions an injury, or threatens legal action or a report to an authority, write only: ESCALATE TO OWNER, and stop.
STRUCTURE
- Positive: thank, one specific detail from the review, name a staff member only if the reviewer did, short close.
- Neutral: thank, what went well in their words, a sorry for the experience (not for a cause), a named contact route.
- Angry: acknowledge the feeling, no defence, invite a private conversation, one commitment from the list in the voice sheet.
- 40 to 70 words. End with the signer on a new line. Do not start with the same words as the last three replies, which I will paste if I have them.
Review: [paste]. Stars: [n]. Department: [x]. Signer: [first name, role].Samples: positive, neutral and angry reviews, with the reply
Every review below is made up, and each reply follows the prompt and the voice sheet above. The notes under each one point out what the rules did.
Positive: a sale and a service visit
Sales review (five stars): "Bought a used hatchback here. The salesperson didn't push us, explained everything and the handover took under an hour. Would recommend."
Thank you for taking the time to write this. It's good to hear the handover was quick and nobody pushed you, because that's how we want every sale to go. I'll pass your comments on to the team. Enjoy the hatchback.
[First name], General Manager
The detail is the handover and the lack of pressure. The reviewer didn't name anyone, so the reply doesn't either. "Hatchback" is the reviewer's own word, so it isn't a revealed detail.
Service review (five stars): "Booked a service online, car ready when they said, and they showed me the worn brake pads before replacing them. Felt fair."
Thanks for the kind words, and for booking online. Showing you the worn parts before fitting new ones is something our advisers do on purpose, and I'm glad it made the visit feel fair. See you at the next service.
[First name], Service Manager
The line about showing worn parts is safe only because it appears in the voice sheet's facts list. If it weren't a real practice, the AI would be inventing a policy for you, which is why the facts list exists.
Neutral: the honest three-star
Service review (three stars): "The repair was done well, but I waited three hours for a job that was meant to take one and nobody updated me."
Thank you for the honest feedback. I'm glad the repair itself went well, and I'm sorry the visit took longer than you were expecting. I'd like to hear more about it, so please contact me at the service desk and we'll look at it together.
[First name], Service Manager
Compare it with what a quick draft would say: "we're sorry nobody kept you updated". That concedes a failure in public. The reply above apologises for the experience, uses the reviewer's good news first, and ends with a named contact route.
Sales review (three stars): "Car was fine and the salesperson was pleasant, but the price was higher than the online listing and I only found out at the desk."
Thank you for your review, and I'm glad the sales team looked after you. I'd like to understand the difference you saw between the listing and what you were told, so please contact me directly and we'll go through it with the details in front of us.
[First name], General Manager
This one is a trap for an AI, which will happily explain the price difference or promise to honour the listing. The reply names no figure, doesn't say which side was wrong, and moves the conversation to a private channel.
Angry: repeat repairs and finance fees
Service review (one star): "Third visit for the same fault and it's still not fixed. Nobody takes responsibility. Avoid."
I'm sorry you've had to bring the car back more than once; that's frustrating, and I understand why you're angry. I'd like to review what's been done so far with you and the workshop team. Please contact me directly, and a manager will call you back the same working day.
[First name], Service Manager
It says sorry for the experience of coming back, acknowledges the feeling, and makes no statement about the cause or who is to blame. The commitment comes from the approved list: if your team can't keep a same-day call-back, take that line out of the list.
Finance review (one star): "They added fees at the finance desk that were never mentioned and pressured me to sign. Felt scammed."
I'm sorry you left feeling this way after your visit to our finance desk. I'd like to understand what happened and go through your paperwork with you in private. Please contact me directly and I'll arrange a time that suits you.
[First name], General Manager
The reply doesn't defend the fees, doesn't say they were disclosed, and doesn't repeat the word "scammed". It also doesn't mention a fee amount. A defensive draft such as "per our records, everything was fully disclosed" would turn a private dispute into a public argument.
Edge cases: a named salesperson and a star-only rating
Review that names a member of staff (five stars): "[First name] in sales was brilliant, patient with all our questions and never rushed us."
Thank you for such kind words. It's great to hear that [first name] was patient with all your questions and never rushed you, and I'll make sure [first name] sees this today. We hope the car is serving you well.
[First name], General Manager
The name comes from the reviewer, not from a roster, and the pass-on is a promise someone must keep. Add a rule that a manager forwards every named compliment the same day, or take the line out.
A star rating with no text is the case where the "one detail from this review" rule can't be met, so the reply should be short, below the usual 40-word floor, and must not invent a detail. An AI left alone will write "thank you for praising our friendly service team" about a review that said nothing. For five stars: "Thank you for the five stars. If there's anything you'd like to tell us about your visit, we'd love to hear it." For one star: "I'm sorry to see a one-star rating, and I'd like to understand what happened. Please contact me directly, and a manager will call you back the same working day." Both end with the signer's name, and neither mentions the car, the department or anything else the reviewer didn't write.
The review that gets no draft
Some reviews should never reach the drafting step. If the review says "they sold me a car with a fault they knew about", mentions an injury, threatens legal action, or says the writer will report the dealership to an authority, the prompt returns ESCALATE TO OWNER, and nothing else. The owner decides, with a solicitor if needed, whether and how to respond. A quick public reply to a serious allegation is the single most expensive mistake available in this process. For the general approach to negative reviews, see the tutorial on replying to negative reviews with AI.
The voice check: eight questions before anyone posts
The drafting step and the checking step should be separate. A draft can read well and still break a rule, and the writer of a draft is the worst person to check it. Put each draft through these eight questions. The first three are hard fails: a reply that fails any of them doesn't go out until it is fixed.
| # | Question | Type |
|---|---|---|
| 1 | Does any wording accept fault, cause or blame? | Hard fail |
| 2 | Is there any price, rate, payment, fee, trade value or discount? | Hard fail |
| 3 | Does it add or confirm customer, vehicle or deal details beyond the review? | Hard fail |
| 4 | Is there one specific detail from this review? | Fix |
| 5 | Does it pass the swap test (would it fit another dealership)? | Fix |
| 6 | Are any never-say phrases or the same opener as the last three replies present? | Fix |
| 7 | Neutral and angry reviews: is there a named contact route? Angry reviews: is there one commitment from the list? | Fix |
| 8 | Is it 40 to 70 words (shorter only for a star-only rating) and signed by a named person? | Fix |
You can use a second AI prompt as a first-pass checker, as long as a person makes the final call:
You are a strict reviewer, not a writer. I'll paste a voice sheet, a Google review and a draft reply. For each of these eight checks answer PASS or FIX and quote the exact words that fail: 1 accepts fault or blame; 2 contains a price, rate, payment, fee, trade value or discount; 3 adds or confirms customer, vehicle or deal details not in the review; 4 lacks one specific detail from this review; 5 would fit any dealership; 6 uses a never-say phrase; 7 lacks a named contact route (neutral and angry reviews) or, on an angry review, a commitment from the voice sheet's list; 8 is outside 40 to 70 words or unsigned. Do not rewrite the draft.See it work on a bad draft for the finance review above (made up, with a made-up figure):
We apologise for any inconvenience caused. Our finance team should have explained the $499 documentation fee more clearly. Your satisfaction is our top priority. Per our records, the paperwork you signed was fully disclosed. Please do not hesitate to contact us.The checker should fail it on 1 ("should have explained"), 2 ("$499"), 3 ("the paperwork you signed was fully disclosed" asserts something about the deal), 4 (no detail), 5 (it fits anyone), 6 (three stock phrases), 7 (no named contact) and 8 (it is unsigned), so it passes none of the eight. The three hard fails mean it can't go out. After a rewrite to the finance reply above it passes all eight. The point of the score is not the number but the habit of reading every reply against the same eight questions.
From good replies to a template library
Once a reply has passed the check, strip out the specifics and keep the skeleton. After a month you have a library that covers most reviews, with slots for the part that must change each time. The tutorial on writing reply templates that keep AI replies on-script goes further on keeping a library tidy.
| Scenario | Skeleton | Slots to fill |
|---|---|---|
| Positive sale | Opener, thanks for [DETAIL], pass-on, short close | DETAIL, STAFF (only if the reviewer named them) |
| Positive service | Opener, [DETAIL], one true service practice | DETAIL, PRACTICE from the facts list |
| Neutral wait or update | Opener, glad about [WENT WELL], sorry the visit took longer than expected, contact | WENT WELL, CONTACT |
| Neutral price or listing | Opener, glad about [WENT WELL], want to understand the difference, contact | WENT WELL, CONTACT |
| Angry repeat repair | Sorry for the experience, acknowledge the feeling, direct contact, one commitment | COMMITMENT |
| Angry finance or fees | Sorry you left feeling this way, private review of paperwork, contact | CONTACT |
| Allegation or legal threat | No template: escalate to the owner | None |
Rotate openers so no two consecutive replies start alike. A starter set is "Thank you for taking the time to write this", "Thanks for the honest review", "I'm glad you shared this", "This is helpful to read, thank you" and "Thank you for telling us how it went". Keep a short log of the last ten openers and paste it into the prompt, as the structure section above allows. Retire any template you see yourself using three times in a week, because a reader scanning your reviews will see the pattern.
Review the library monthly. Every time a manager heavily rewrites a draft, ask which sentence failed and whether it was a missing fact, a missing rule or a bad slot. Fix that, not the symptom. The before-and-after habit in the tutorial on tone of voice examples before and after editing shows how to keep your edits as teaching material.
Who approves what, and what a month costs in time
| Review type | Drafted by | Edited and approved by | Posted by |
|---|---|---|---|
| Positive | AI | Department manager | Department manager |
| Neutral | AI | Department manager, checked by the general manager | General manager |
| Angry | AI, as a starting point | General manager | General manager |
| Allegation or legal threat | No draft | Owner, with a solicitor if needed | Owner |
Take an illustrative month for a dealership with 60 new reviews: 42 positive at 2 minutes each (draft, one detail check, post), 12 neutral at 6 minutes, and 6 angry at 15 minutes of manager time. That is 84 plus 72 plus 90, or 246 minutes, a little over four hours spread across several managers, before any allegation needs the owner. If your own timings differ, time a week and recalculate; the shape matters more than the numbers.
A few practical details about Google itself. Replies appear under the review as your business, and your personal name isn't shown, which is why the sign-off sits inside the text. Google reviews replies against its content policies, and says replies usually take up to 10 minutes to be reviewed, though it can take up to 30 days. You can edit or delete a reply later. The reviewer is notified when you reply and can change their review afterwards, which updates its date. Never ask for that change, and never offer anything in return for one. For the private messages that go to customers rather than the public page, the tutorial on ChatGPT for car sales, follow-ups and what never to promise applies the same discipline.
Questions car dealers ask about AI review replies
Can AI auto-post replies to Google reviews?
Some tools can, but a dealership shouldn't switch it on. Angry and neutral reviews touch fault, finance and customer privacy, and an automatic reply can break all three rules in public before anyone reads it. Let AI draft, let a manager edit, and keep the posting step human. Positive replies can be approved in bulk, because they carry the least risk.
Should the reply name the salesperson?
Only when the reviewer named them first, and ideally when that person is happy to be named. Use a first name, never a surname, and never in a neutral or angry reply, where naming staff drags them into a public argument. For positive reviews, a private message passing the praise on is often worth more than the public mention.
What if the review looks fake or comes from a competitor?
Don't argue in public. Check your records quietly, flag the review through Google's own reporting route, and reply neutrally, if at all: say you can't match it to a customer and invite them to get in touch. An accusation of fakery can inflame a real customer, so keep the tone calm until you know.
How long should a reply be?
Short. Google's own tips say to keep replies short and simple, and a house limit of 40 to 70 words suits most reviews. Positive replies can sit near the lower end. Angry ones need enough room to acknowledge the experience and give a named contact, and no room to explain, defend or negotiate in public.
Further reads
- How to Reply to Hotel Reviews With AI Without Sounding Canned — The same not-canned discipline applied to hotel reviews.
- AI Review Management Tools for Small Businesses (2026) — What reputation tools add if you outgrow a prompt and a log.
- Should You Let AI Auto-Post Replies to Google Reviews? — Why automatic posting of replies is risky, in more detail.
- AI Marketing for Car Dealerships: A Four-Week Plan for One Lot — Fit review replies into a wider four-week plan for a lot.
- How Can AI Help Car Dealerships? A 30-Day Plan for a Small Lot — A 30-day plan for where AI helps a small independent lot.
- AI Chatbot for Car Dealerships: Rules to Set Before Launch — Rules to set before a chatbot talks to your customers.
- AI Automation for Car Dealerships: Answer a Web Lead in Minutes — A form-to-CRM-to-AI-draft-to-approval flow for dealership web leads, built in Zapier or Make, with the monthly cost worked out.
- AI Tools for Car Sales: Test-Drive Follow-Ups, Quotes and Call Notes — A salesperson's routine for AI follow-ups after a test drive and tidy call notes, with prompts, samples and the lines AI must never write.
- AI for Used Car Dealerships: Honest Listings and Photo Clean-Up — House rules for AI listings and photos: source-tagged descriptions, a photo-edit decision table and honest canned replies that never hide a flaw.
- AI-Powered Marketing Tools for a Motorcycle Dealer: Spring Campaigns — A ten-week spring plan for a motorcycle dealer: which AI tools to use for used-bike posts, service reminders and demo days, with list prices.
- ChatGPT Prompts for Car Dealers: 10 for Follow-Ups and Trade-Ins — Ten short ChatGPT prompts for follow-ups, trade-in replies and reviews, each with sample output and a check to run before sending.
- Franchise Review Management: Replies for Several Stores, On Brand — A store-level workflow for franchisees: a reply library, escalation rules and per-site fact cards so AI drafts stay on brand and a person approves each reply.
- How to Use AI in Car Sales: Trade-In Enquiries Before the Appraisal — An intake that gathers photos, mileage, service history and finance owing into a one-page pack, with guardrails so the AI never quotes a trade-in value.
- AI for Auto Dealerships: Where a Small Independent Lot Should Start — A decision rule for a lot of about 20 cars: why listings come first, lead follow-up second and phone answering last, and which finance topics stay human.
- 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 (reply to reviews; tips for responding to reviews).