TL;DR
A negative review is read by future buyers, not just the angry one. Feed the AI the verbatim review, your verified facts, and the resolution you can actually deliver, then ship a short, specific four-sentence reply. The big change as of June 2026: Google now holds every review reply for moderation, and generic AI boilerplate (“thrilled to hear”, “your kind words”) is the single most common reason replies get rejected. So the goal is no longer just “sound human” — it is to write something Google’s filter will actually publish.
The task
A 1-2 star public review just landed on your Google Business Profile, App Store listing, or a marketplace like Trustpilot. You need a reply that calms the unhappy reviewer, signals to every future shopper that you handle problems well, stays inside the platform’s character limit, and — on Google — clears moderation instead of getting silently blocked.
What changed in 2026 (read this first)
Google ended instant review replies. Through the Business Profile API (v4.9), every reply you send is now assigned a ReviewReplyState:
- PENDING — being screened by Google.
- APPROVED — live under the customer’s review.
- REJECTED — blocked, never shown publicly, often with no notification.
Moderation usually takes minutes but can take up to 30 days in edge cases. The biggest rejection trigger, per practitioners analyzing thousands of blocked replies, is templated AI filler: phrases like “thrilled to hear”, “your kind words”, and “look forward to welcoming you back” appear in the majority of rejected replies. Google’s own Search Central guidance is that appropriate use of AI is fine — what gets penalized is generic, spammy, keyword-stuffed output. That is exactly the failure mode an unguided chatbot produces, which is why the prompt below forces specificity.
When AI is the right tool
- You handle more than 5 negative reviews per week and consistency slips.
- Multiple agents reply and you want one on-brand voice.
- The reviewer is upset and you want a calm second pass before posting.
- You need replies in 2+ languages for global listings.
When not to rely on AI alone
- Legal threats, safety claims, or fraud accusations — escalate to a human and possibly legal counsel.
- Reviews naming a specific employee or transaction the AI cannot verify.
- High-profile reviewers (press, influencers) whose reply will be screenshotted.
Where you can (and can’t) reply: platform limits as of June 2026
Know the rules before you draft. Amazon is the trap: it removed public seller comments on product reviews back in 2020-2021 and never brought them back, so there is no public reply to write there.
| Platform | Public reply to reviews? | Character limit | Notes |
|---|---|---|---|
| Google Business Profile | Yes | 4,000 | Replies are moderated (PENDING/APPROVED/REJECTED) before going live |
| Apple App Store | Yes, one editable reply per review | ~5,970 | Appears within 24h; needs Account Holder/Admin/Customer Support role |
| Google Play | Yes | 350 | Tight limit — one or two sentences max |
| Trustpilot | Yes | Generous | Reply in 1-2 days; move sensitive details to private contact |
| Amazon product reviews | No (removed) | n/a | Brand Registry can contact 1-3 star reviewers privately via Customer Review Contact |
What to feed the AI
- The full review text, verbatim.
- Your verified version of events (order number, what actually happened).
- The concrete resolution you can offer (refund, replacement, store credit, escalation path).
- Brand voice notes — formal vs. warm, “we” vs. a singular shop voice.
- The platform’s character limit from the table above.
Copy-ready prompt
You are a customer experience manager replying publicly to a negative review.
Review text:
"""
[review_text]
"""
What actually happened: [our_side]
Resolution we are offering: [resolution]
Brand voice: [voice_notes]
Max length: [char_limit] characters.
Write a single public reply that:
1. Addresses the reviewer by first name if given.
2. Names the specific problem in their own words — no "sorry you feel that way".
3. Takes responsibility where appropriate without admitting unverified claims.
4. States the concrete resolution and exactly how to claim it.
5. Closes with one forward-looking line for future readers.
Hard bans (these get Google replies rejected):
"thrilled to hear", "your kind words", "we value your feedback",
"look forward to welcoming you back", any marketing slogan, any emoji.
Tone: warm, calm, adult-to-adult. Reference at least one concrete
detail from the review so it cannot read as a template.Any current model handles this well — GPT-5.5, Claude Sonnet 4.6, or Gemini 3.1 Pro. For a consistent house voice across a team, paste 2-3 of your own approved past replies into the prompt as style examples; that anchors tone better than any “brand voice” adjective.
Recommended output structure
Four sentences usually wins: address + specific acknowledgement + resolution + forward-looking close. Anything longer reads as defensive, and on Google Play you simply will not fit. Aim well under 100 words on a product page.
How to check the output before posting
- Read it as a shopper, not the merchant. Does it make you trust the brand more?
- Cut any line that could be pasted onto a different review unchanged — that is the line Google will reject.
- Confirm the resolution is something you can actually deliver this week.
- On heated cases, wait at least two hours before posting.
Common mistakes
- Generic openers (“We’re sorry you had a bad experience”) with zero specifics — now an active liability under Google moderation.
- Defending the company before acknowledging the customer.
- Posting verifiable claims you cannot back up.
- Walls of text; over 100 words on a product page reads as defensive.
- Trying to reply publicly on Amazon (you can’t) instead of using private Customer Review Contact.
Next steps to keep improving
Build a small library of your best approved replies, tagged by complaint type (shipping, defect, sizing, support delay), and feed them as examples in future prompts. Then track the one metric that matters: whether reviewers update their star rating after your reply.
FAQ
- Will an AI-drafted reply get me penalized on Google? No — Google explicitly allows AI assistance. What gets rejected is generic output. A reply that quotes the customer’s specific issue and offers a real fix passes; “thrilled to hear about your kind words” does not.
- Why did my Google reply never appear? Since 2026, replies are moderated and assigned PENDING, APPROVED, or REJECTED. Rejected replies are blocked silently. Rewrite with concrete detail and re-submit.
- Can I reply to Amazon product reviews? Not publicly — Amazon removed that in 2020-2021. Brand Registry members can contact 1-3 star reviewers privately via Customer Review Contact.
- How fast should I respond? Within 24-48 hours on Google and marketplaces; 1-2 days on Trustpilot. Faster looks attentive, slower looks dismissive.
- Can I ask the reviewer to update their rating? Only after the issue is resolved, and only privately. Never in the public reply.
- What if the review is fake? Reply professionally as if real, then flag it to the platform separately.
Related
See review reply prompts, negative review response prompts, and the negative review response AI workflow for deeper patterns.
Tags: #E-commerce #Workflow