Replying to reviews used to be the worst part of running a small business. Modern AI changes that — but only when you avoid three traps that quietly hurt your Google ranking. This playbook is the version we wish someone had written for us when we started Yerly.
What works in 2025
- AI drafts in your tone. Trained on 10–20 of your past replies, it gets brand voice within 24 hours.
- Five-language detection. The customer wrote in German? Reply in German. Google rewards multilingual responsiveness.
- Specific acknowledgement. AI quotes one phrase from the original review back at the customer — proves you read it.
- Human approval at the start, auto-send for 4–5 stars later. Most owners flip to auto-send for positive replies after 2 weeks.
The three traps to avoid
Trap 1: identical openings
Google's spam filter flags profiles where every reply starts with "Thank you for your review!" Vary the opening — by mood, by name, by detail. Yerly does this automatically; ChatGPT-with-templates doesn't.
Trap 2: over-formal language
"We sincerely apologise for the unacceptable experience" reads as robotic in any language. Future customers can smell it. Aim for the tone you'd use texting a regular.
Trap 3: contradicting the reviewer
AI will sometimes argue back if you don't muzzle it. "Actually, our cake is fresh" is the worst possible response to a 1-star review. The playbook is: acknowledge → context → fix → invite. Argue privately, never in public.
Setting AI up the right way
- Load 10–20 of your hand-written replies as voice samples.
- Choose your reply length (short = 1 sentence, medium = 2–3, long = 3–4).
- Decide who gets auto-sent: usually 5★ and 4★ without text.
- Add a kill switch: under 4★, always human-approve.
Yerly default
Yerly drafts every reply and auto-sends only 4–5 star replies, with 6 safety locks (time delay, per-user cap, global cap, etc.). 1–3 stars always wait for you. This is the safest setup for new accounts.
See how Yerly's AI replies work →
Frequently asked questions
Will Google penalise AI-written replies?↓
No — Google's spam filter targets duplicate content and policy violations, not AI per se. Replies in your voice with specifics pass without issue.
Should I auto-send everything?↓
Not at first. Spend 2 weeks approving every reply, build trust with the AI's tone, then enable auto-send for 4–5★ only.
What languages should I support?↓
At minimum: your local language + English. Tourist-heavy areas should add the top 2 visitor languages. Yerly auto-detects 5.