How to Write an AI Review-Response Prompt That Sounds Like You
The difference between a form-letter reply and one that sounds like you comes down to four things in the prompt.
You can always tell an AI-drafted review reply that was rushed: "Thank you for your feedback! We strive to provide excellent service to all our valued customers." Generic, safe, and it could have come from any business on earth. Here's the fix — four specific things to put in the prompt, every time.
1. Name the business type and personality, not just the name
"Draft a reply for Ridgeline Auto Repair" gets you generic output. "Draft a reply for Ridgeline Auto Repair, a family-owned shop that's warm but not overly formal" gets you something with an actual voice. Say how you'd talk, not just who you are.
2. Paste the actual review, not a summary
The specific details in a customer's review — a name, a specific thing they mentioned, a complaint about one particular thing — are exactly what makes a reply sound personal instead of templated. Always paste the full review text into the prompt.
3. Set a word limit and ask it to reference something specific
"Under 60 words, and reference the specific thing they mentioned" forces the output to engage with the actual review instead of defaulting to boilerplate. Long AI-drafted replies read as more obviously AI-drafted; short, specific ones don't.
4. Give explicit instructions for negative reviews
Never let a review-response workflow auto-post a reply to criticism unsupervised. Instead: "If this review is critical, draft a reply that acknowledges the specific issue without being defensive, and flag it for me to personally review before it's posted." That one line is the difference between a helpful tool and a PR incident.
A full example prompt
"Draft a reply to this Google review for [business name], a [one-line description of the business and its tone]. Under 60 words, thank them by first name if given, and reference something specific from their review. If the review is critical, flag it as needing my personal review instead of drafting a standard reply. Here's the review: [paste full review]"
See this exact prompt run against a real review, with the output shown in full: before & after examples.