AI prompt: reply to a negative Google review
A copy-and-paste AI prompt that turns a 1- or 2-star Google review into a calm, specific public reply: apologise once, name the problem, say what changed and take it offline.
When to use it
An unhappy customer has left a 1- or 2-star Google review and you want to reply in public without sounding defensive, copy-pasted or careless with their privacy.
- Copy the prompt below.
- Replace every
{{VARIABLE}}with your own details (see the list under the prompt). - Paste it into the AI chat tool you already use.
- Read the reply against the checklist before you post it on Google.
The prompt
You are helping a local business owner reply to a negative Google review. Write the public reply the owner will post under the review.
BUSINESS
- Business name: {{BUSINESS_NAME}}
- Type of business: {{BUSINESS_TYPE}}
- Who signs the reply (first name and role): {{SIGNER}}
- Best way for the reviewer to reach us directly: {{CONTACT}}
- Tone: {{TONE}}
THE REVIEW
- Reviewer name as shown on Google: {{REVIEWER_NAME}}
- Star rating: {{STARS}}
- Review text: {{REVIEW_TEXT}}
PRIVATE FACTS (for your understanding only; never quote them or reveal private details in the reply)
- What actually happened, from our side: {{WHAT_HAPPENED}}
- What we have changed or are doing about it (write "nothing yet" if nothing): {{WHAT_WE_CHANGED}}
- Anything the reply must not mention (industry rules, privacy, legal): {{DO_NOT_MENTION}}
RULES FOR THE REPLY
1. Open by using the reviewer's first name and thanking them for the feedback. Do not start with "We're sorry you feel".
2. Say sorry for the specific problem they described, in one sentence that starts with "We're sorry" and names the problem in plain words. Apologise once. No excuses, no blaming the customer, staff or "a busy day".
3. If we have made a change, describe it in one sentence. If the answer is "nothing yet", say only that we have shared the feedback with the team. Never promise anything we did not list.
4. Invite them to contact the signer directly using the contact above so we can talk it through.
5. Do not argue with the facts in the review, even if our private facts differ. Do not reveal job details, prices, addresses, health details or anything the reviewer did not write themselves.
6. Do not ask them to change, update or remove the review, and do not offer a discount, refund, gift or anything else in return for it.
7. Do not invent facts, numbers, guarantees, policies or awards.
8. Keep it between 50 and 100 words, in short sentences, with no exclamation marks, no emojis and no hashtags.
9. Finish with a sign-off line in this exact format: {{SIGNER}}, {{BUSINESS_NAME}}
Return only the finished reply, including the sign-off line, and nothing else.What to fill in
- {{BUSINESS_NAME}}
- Your business name exactly as it appears on Google.
- {{BUSINESS_TYPE}}
- What you do and where, in a few words.
- {{SIGNER}}
- First name and role of the person signing the reply.
- {{CONTACT}}
- The phone number or email the reviewer should use to reach that person.
- {{TONE}}
- Two or three words for how you sound, e.g. friendly and plain-spoken.
- {{REVIEWER_NAME}}
- The reviewer's name as Google shows it.
- {{STARS}}
- The star rating they left.
- {{REVIEW_TEXT}}
- The review, pasted word for word.
- {{WHAT_HAPPENED}}
- Your side of the story. The AI uses it for context only and is told not to repeat it.
- {{WHAT_WE_CHANGED}}
- A real change you have made, or "nothing yet". The prompt tells the AI not to invent one.
- {{DO_NOT_MENTION}}
- Anything that must stay out of a public reply: staff names, prices, addresses, health details.
Real example output
Fictional sample business. Barton Creek Plumbing Co., Rachel P., Dana and the (512) 555-0142 phone number are fictional. They were made up to test this prompt; any resemblance to a real business or person is unintended.
Show the inputs used for Barton Creek Plumbing Co. (fictional)
- BUSINESS_NAME
- Barton Creek Plumbing Co.
- BUSINESS_TYPE
- Residential plumber in Austin, TX
- SIGNER
- Dana, owner
- CONTACT
- (512) 555-0142
- TONE
- friendly, plain-spoken, no jargon
- REVIEWER_NAME
- Rachel P.
- STARS
- 1 star
- REVIEW_TEXT
- Booked a water heater repair for the morning window. Nobody showed up until almost noon and nobody called to tell us they were running late. The repair itself was fine, but I took the morning off work for nothing. Won't be using them again.
- WHAT_HAPPENED
- The job before hers ran long and the office never made the call to warn her that the technician was late.
- WHAT_WE_CHANGED
- The office now calls the next customer as soon as a technician is running late.
- DO_NOT_MENTION
- the technician's name, the customer's address, the price of the repair
What the model returned
Rachel, thank you for your feedback. We're sorry you had to wait until noon for the repair. The office now calls the next customer as soon as a technician is running late. Please feel free to contact us directly at (512) 555-0142 to discuss further. Dana, owner, Barton Creek Plumbing Co.
Model: Qwen2.5-3B-Instruct (Q4_K_M GGUF), llama.cpp server b11853, run locally, temperature 0.4. Tested on . Edits: None. The output is shown exactly as the model returned it. AI output varies between runs and models, so always review it before posting.
Check before you post
- Every fact in the reply is true and appears in the review or your inputs.
- Nothing private is revealed: no job details, prices, addresses or health information the reviewer did not post themselves.
- Any change you mention has actually been made.
- The reply does not ask the reviewer to edit or remove the review, and offers nothing in exchange for it.
- The contact details are correct and someone will answer them.
Local Business AI Prompt Pack
Get the full set of local business prompts.
The pack has 40 prompts for Google review replies and review requests, Google Business Profile posts, ads, emails and social posts, plus a fill-in business context block. One-time purchase, $24.
Related tools
Want to run prompts like this from your own code? These AI gateways in the Digidle directory put many models behind one API.