Answer every review like the owner wrote it — with help from Claude and GPT
Prospects don’t just read your reviews — they read your replies. A thoughtful response to a bad review is often more persuasive than ten five-star ratings, because it shows how you behave when something goes wrong. And replying consistently signals to Google that the profile is actively managed, which matters for local ranking.
The problem is that reply-writing is exactly the kind of task that dies on a to-do list. It’s emotionally loaded (nobody enjoys reading criticism), it feels optional, and it piles up. This is a perfect job to share with AI.
Teach the model your voice first
Open Claude or GPT and paste in your five best past replies — the ones that sound like you. Then paste the new review and ask for a draft “in the same voice.” With real samples to imitate, either model gets remarkably close to how you’d actually write; without them, you get generic corporate mush. Keep the samples saved so every session starts warm.
The negative-review protocol
- Acknowledge the specific complaint — never a form response. Name the job, the date, the thing that went wrong.
- Say what you did or will do about it. One sentence, concrete.
- Move the rest offline: invite them to call or email you directly.
- Never argue, never blame the customer publicly, and never let a model talk you into either.
The hard rule for AI drafts: the model never invents details. If the draft mentions anything you can’t verify — a date, a technician, an offer — cut it before it ships.
Make it automatic
The copy-paste workflow works, but the failure mode is the same as before: you stop doing it in a busy month — which is exactly when reviews spike. Coff closes that loop: it watches for new reviews, drafts a reply in your saved voice, and queues it for a one-tap approve. It also asks happy customers for reviews right after a finished job, which is when they actually say yes — so the flywheel runs on both ends.
A profile where every review has a considered, owner-voiced reply reads as a business that cares. That impression is earned one reply at a time — AI just makes sure the replies happen.