Use AI to Draft On-Brand Replies to Customer Reviews (Good and Bad)

Paste the review and a short tone guide into ChatGPT or Claude and ask for a calm, on-brand reply you approve before posting. Give it firm rules: thank the customer, never argue, never share private order details in public, and offer to move specifics to email or DMs. It drafts in seconds for five-star and one-star reviews alike, and you stay the final editor.

Replying to reviews is a small job you keep putting off, and the bad ones you put off longest. The happy reviews deserve a thank-you. The angry ones need a careful, calm answer that does not make things worse. ChatGPT or Claude is good at exactly this once you give it your tone and your rules. You paste the review, it drafts a reply that sounds like your brand, and you approve before it posts. The point is not to hand your reputation to a bot. It is to remove the blank-page stall so a stack of reviews becomes a few minutes of editing.

Before you open any AI, write five lines about how your brand sounds. This is the single thing that stops every reply reading like a generic corporate apology. Keep it short and specific:

• Warm and plain, like a real person, never stiff or formal.
• Short. Two or three sentences for a good review, three or four for a complaint.
• We thank by name, own mistakes directly, and never make excuses.
• No exclamation-mark spam, no salesy lines, no "we value your feedback."

Save this in a note. You will paste it above every review from now on, so the voice stays consistent whether it is one reply or twenty.

A reply can fix a bad impression or make it permanent. Hand the AI a short list of hard rules so it never crosses a line you would regret in public. Paste these every time:

• Never argue, never get defensive, never blame the customer.
• Never share private order details in public: no order numbers, names, addresses, or refund amounts.
• For anything that needs their order looked up, invite them to email or DM us.
• Thank the person, acknowledge the specific issue, offer one clear next step.

These four lines are the difference between a reply that builds trust and one that gets screenshotted. Treat them as required, not optional.

Open ChatGPT, Claude, or Gemini and give it everything in one message: your tone guide, your rules, and the exact review text. Pasting the real review matters, so the draft references the actual product and the actual complaint instead of vague filler.

Prompt: Here is my tone guide and rules, then a customer review. Draft a reply that follows both. Keep it short and specific to what they said. Review: [paste it].

That is the whole input. The AI now has your voice, your guardrails, and the real words it is responding to, which is everything it needs to write something you can actually use.

Do not settle for the first draft. Ask for a small set so you can pick the tone that fits the moment, since a mildly annoyed three-star and a furious one-star need different warmth.

Prompt: Give me three versions: one warm and brief, one more apologetic, one neutral and professional. All follow the rules above.

Now you choose instead of compose, which is far faster. Pick the version closest to right, and notice that the angry reviews are where having options helps most. One will land calmer than whatever you would have typed while irritated.

This is the step you never skip. Read the draft you picked and do three quick checks: no private order details slipped in, no defensive or sarcastic line, and it actually sounds like you. Then edit one sentence by hand so it carries a human fingerprint, not a template. This takes under a minute per reply and it is the whole safety net. Reviews are public and emotional, and one tone-deaf auto-reply does more damage than a slow one ever could. Fast drafting, human approval, then post. Keep that order and AI stays a help, not a liability.

Angry reviews are where this earns its keep, because the calm reply is the hard one to write when you are annoyed. Tell the AI the move: acknowledge the frustration, own anything that was genuinely your fault, do not relitigate the details in public, and offer to make it right by email. Future shoppers read this exchange closely, so the audience is them, not the reviewer.

Prompt: This is a one-star. Draft a calm reply that takes responsibility, does not argue or share order details, and invites them to email us to fix it.

A short, fair, non-defensive reply to a one-star builds more trust than the five-star reviews above it.

Open your reviews and pick one good and one bad. Write your five-line tone guide and the four rules, paste them with the first review, and ask for three versions. Choose one, read it against the rules, edit a sentence so it sounds like you, and post it. Then do the angry one. Two real replies is the fastest way to feel how much this removes the dread, and to tune your tone guide before the next batch of reviews lands.

Take the last review you actually received — the awkward one you put off — paste it in with your tone rules from Step 1, and post the reply once you have edited it. Save the prompt; it is now your reply machine.