Blog · · 5 min read

How to reply to Google reviews with n8n and AI (draft → you approve)

Build an n8n agent that spots new Google reviews, drafts a reply in your brand voice and emails it for approval. You decide what gets published.

A 1-star Google review with no reply sits there for months. A 5-star one without a «thank you» does too. Replying builds trust and local ranking, but if you run a workshop, clinic or shop you can't always stop to write every response carefully. So you reply late, in a rush, or not at all.

With n8n, a bit of AI and Gmail (or Google Sheets) you can build an agent that drafts the reply as soon as a review arrives. You only read, tweak if needed and approve. The voice is yours; the decision to publish stays yours.

What you'll build

Every time a new review arrives, the workflow:

  1. Detects the review (Google Business Profile alert email, a new row in a sheet, or a webhook from a tool that syncs reviews).
  2. Normalises the data: star rating, text, author name, language, date and business link.
  3. The AI drafts a reply in your brand voice: thanks for positives, acknowledges the issue if it's negative, doesn't argue in public and suggests a concrete next step (call, email support…).
  4. Emails you the draft (and optionally saves it to Sheets with status «pending»).
  5. You approve (email reply, button or status change in the sheet). Only then does the workflow post — or leave you the text ready to paste into Google.

The critical bit: it doesn't publish on its own. A misread draft under a 1-star review does more harm than staying quiet for a day.

What you need

  • An n8n instance with HTTPS.
  • An OpenAI key (or another model compatible with the chat node).
  • Gmail OAuth2 credentials (and Google Sheets OAuth2 if you want a log).
  • A way to get reviews: simplest is the notification email from Google Business Profile; if you already sync reviews into Sheets, even better.
  • About 40 minutes and 4–5 sentences that sound like you (how you say thanks, how you apologise, what you'd never say in public).

Step 1: define your voice and rules in one place

Before the AI, fill a settings node with:

  • Business name and sector (salon, clinic, restaurant…).
  • Brand voice: 3–5 concrete traits. «Warm, short, no jargon, second person» works; «professional and friendly» is too vague.
  • Hard rules: don't argue facts in public; don't ask them to change the rating; don't invent compensation; on 1–2 stars, acknowledge the issue and offer a private channel; on 4–5, thank them and echo a detail they mentioned.
  • Max length (e.g. 400–600 characters) and default language.
  • Email for drafts and, if you use a sheet, the Google Sheet ID.

The more concrete the voice, the less you'll edit. If you're unsure, copy three real replies you like and describe what they share.

Step 2: capture the review

Three common paths (pick one):

  1. Gmail: a trigger watches for Google notification emails («New review» / «Nueva reseña»). A code node extracts stars, author and text from the body or HTML.
  2. Google Sheets: if you already sync reviews into a Reviews tab, a «new row» trigger starts the flow.
  3. Webhook: a third-party app (or the Business Profile API, if you have it set up) POSTs JSON like {"stars":2,"author":"Laura","text":"…","url":"…"}.

Always normalise to the same internal shape: stars (1–5), text, author, detected language, URL. If the text is empty (stars only), the AI should still reply: a short thank-you on 5 stars, or an invite to say what went wrong on 1–2.

Step 3: the AI drafts the reply

A chat node receives the review plus your settings (BRAND_VOICE, hard rules, business name). The prompt asks for:

  • a ready-to-post reply (no quotes or prefixes),
  • a tone that matches the star rating,
  • a one-line internal summary (for the approval email),
  • and a risk flag (ok / review): e.g. if the review mentions medical, legal or serious accusations.

Tell the AI not to invent apologies for facts that aren't in the review and not to promise refunds or discounts unless you authorised that in the config.

Step 4: you get the draft; you approve

A Gmail node emails you: stars, original text, proposed draft, risk flag and (if available) the review URL. Optional: a Sheets row with status pending.

Simple approval options:

  • Reply to the email with OK or with edited text; a second flow (or the same branch with a wait) publishes or archives.
  • Change the status in Sheets to approved / edited / discarded and let n8n read that cell.
  • Draft only: the flow stops at the email and you paste into Google Business Profile by hand (fine if you get few reviews a month).

Start with draft-only. When you trust the tone, add structured approval.

Test it with three reviews

  1. 5 stars with detail: should produce a short thank-you that cites something concrete from the text.
  2. 1 star with a clear complaint: should acknowledge the issue, not argue, and offer a private channel.
  3. Stars only, no text: should reply briefly and safely, without inventing a reason.

Check that the approval email arrives, that Sheets (if used) stores the row, and that nothing publishes without your manual step.

Honest limitations

  • Google doesn't always email instantly; there can be delay. If you need minutes, look at API or Sheets sync.
  • The AI doesn't know your back office. If the review mentions a specific order, the draft will stay generic until you add the detail when approving.
  • Reviews in another language: detect the language and reply in the same one, or force your business language in the config.
  • Serious accusations or legal/medical topics: the review flag should stop autopilot; you write those yourself.
  • It doesn't replace judgement. It's a fast drafter with a human in the loop, not an autonomous community manager.

Ideas to go further

  • Keep a Sheets history (date, stars, draft, published text, response time) so you can see if you reply faster.
  • Alert via WhatsApp or Slack as well as email on 1–2 star reviews.
  • Pair it with the lead qualifier if someone leaves a positive review and asks for a quote in the same message.
  • Use the content repurposer to turn 5-star reviews (with permission and no personal data) into social proof for LinkedIn or your newsletter.

Shortcut: the agent already built

This workflow is one of the eight agents in the AgentKiosk pack: an n8n template ready to import, brand voice and rules in one node, a star-aware reply prompt, approval email, status sheet and a PDF guide. It has been imported and run end-to-end in n8n with sample data and simulated services; run your own tests with real (or sample) reviews before relying on it for your Google listing.

To start for free, download the WhatsApp agent and also read 7 AI automations for small businesses.