How to Automate Customer Review Management with n8n, Gmail, and Gemini AI

Reviews decide whether a stranger trusts your business, yet asking for them is awkward and chasing them is worse. Send nothing and your happiest customers quietly move on; send a blast and you risk an angry one-star landing on your public Google listing where it sits forever. The naive fix, a manual "please review us" email, gets sent late, to the wrong people, with no way to catch a frustrated customer before they go public.

This guide builds a review engine that handles the whole lifecycle in n8n. It emails customers a review form after purchase, filters happy customers toward your public Google listing while routing unhappy ones into a private feedback channel, logs every response in Google Sheets, and uses a Gemini AI agent to draft replies to your public reviews. It covers the same ground as the popular WhatsApp-based review automation from GrowwStacks, but over email and without the click-tracking step.

Available resources#

This build uses three assets you will set up below:

1. n8n workflow "Customer Review Management" (email requests, feedback filtering, and AI review replies).
2. Google Sheet tab "Customers" (the contact list and send-status tracker).
3. Google Sheet tab "Reviews" (the log of every rating, comment, and reply status).

What you'll need#

Before you begin, make sure you have:

1. An n8n account (cloud or self-hosted) reachable at a public HTTPS webhook URL.
2. A Google account with one Google Sheet (two tabs) used as the system of record.
3. A Gmail account for sending the review request emails.
4. A Gemini API key for the AI agent that drafts review replies (any chat model n8n supports also works).
5. A hosted review form (a simple web page) that collects a star rating and posts low ratings to your n8n webhook.
6. A Google Business Profile if you want to pull and reply to public Google reviews.

Overview of the automation#

The automation runs in three phases that share one Google Sheet. A human decision sits between capturing feedback and replying to it.

1. Request reviews. On a schedule, n8n pulls customers who have not been asked yet and emails each one a link to your review form, then marks the row as sent.
2. Capture and filter feedback. The form sends 4 and 5 star customers to your public Google listing, while 1 to 3 star customers submit private feedback that posts to an n8n webhook and is logged for follow-up.
3. Reply to public reviews. A Google Business Profile trigger feeds each new public review to a Gemini agent, which drafts a short reply for your team to approve before it posts.

The important design decision is the rating split before anything becomes public. A delighted customer is one tap from a five-star Google review; a frustrated one is one tap from a private message that reaches your inbox instead of your listing. The filter happens on the form itself, so the unhappy path never touches Google, and you get a chance to fix the problem before it costs you a public star.

Step-by-step setup#

1. Set up the Google Sheets#

Create one Google Sheet with two tabs. The first tab drives who gets emailed; the second is the review log.

Tab one, "Customers" (the contact list and send tracker):

Tab two, "Reviews" (the log of every response):

2. Connect credentials in n8n#

Add three credentials so the Sheets, Gmail, and agent nodes can authenticate:

1. Google Sheets (OAuth2) with read and write on both tabs.
2. Gmail (OAuth2) for sending the review request emails.
3. Google Gemini for the chat model behind the review-reply agent.

3. Build phase one: email the review request#

Trigger the batch. Add a Schedule Trigger named Schedule Trigger and set the interval that fits your volume, for example once a day. Leave it disabled while you build and test, then enable it when you go live.

Pull customers who have not been asked. Add a Google Sheets node named Get row(s) in sheet pointed at the Customers tab. Add a filter so it only returns rows where Review Email equals Not Sent. That keeps the batch to people who still need the ask.

Guard against double sends. Add an If node that only continues when the send-status field is still empty:

={{ $json['Review msg Status'] }}   is empty

Loop and send. Add a Loop Over Items (Split in Batches) node so each customer is handled one at a time, then a Gmail node named Send a message on the loop output. Set the recipient to the customer email and the body to your review form link:

To:      ={{ $json.Email }}
Subject: Please help us with your valuable review
Body:    https://your-review-form-url

Mark the row as sent. After the Gmail node, add a Google Sheets node named Update row in sheet set to Update, matching on Email, and write Review Email = Sent. Wire it back into Loop Over Items so the batch advances to the next customer. This is what stops the same person being emailed twice.

4. Build phase two: capture and filter the feedback#

The filtering lives on the review form, not in n8n. Build a small page with a star rating that branches on the score: 4 and 5 stars open your public Google review link, while 1 to 3 stars reveal a short comment box and POST the feedback to your n8n webhook.

Receive the private feedback. Add a Webhook node named Receiving Customer Review (POST). Enable CORS for your form origin so the browser can call it. The form should send a JSON body like this:

{
  "rating": 3,
  "name": "Jane Doe",
  "email": "jane@example.com",
  "review": "The product was fine but delivery was slow.",
  "timestamp": "2026-06-08T06:12:53.691Z"
}

Normalize the fields. Add a Set node named Edit Fields (Edit Fields) that maps the incoming body into clean top-level fields:

rating = {{ $json.body.rating }}
name   = {{ $json.body.name }}
email  = {{ $json.body.email }}
review = {{ $json.body.review }}

Update the contact row. Add a Google Sheets node named Update row in sheet1 set to Update on the Customers tab, matching on Email, so you know this person responded.

Log the review. Add a Google Sheets node named Append row in sheet set to Append on the Reviews tab. Map the rating, name, comment, and set the workflow defaults so new feedback lands in a clear queue:

Name                        = {{ $('Edit Fields').item.json.name }}
Email                       = {{ $json.Email }}
Rating                      = {{ $('Edit Fields').item.json.rating }}
Customer Review             = {{ $('Edit Fields').item.json.review }}
Handle With AI Or Manually  = Manually
Replied Or Not              = Not Replied

Defaulting private feedback to Manually and Not Replied is the human gate: a low rating is a service-recovery moment, so a person reads it and decides whether to reply by hand or hand it to the AI agent.

5. Build phase three: AI replies to public Google reviews#

Trigger on new reviews. The shipped workflow uses a Manual Trigger named "Change this with Google Business Profile Trigger" as a stand-in. Replace it with a Google Business Profile trigger (or a scheduled HTTP request to the Business Profile API) that emits each new public review.

Shape the review. Add a Set node named Edit Fields1 that pulls the fields the agent needs from the review payload:

reviewId    = {{ $json.reviewId }}
displayName = {{ $json.reviewer.displayName }}
comment     = {{ $json.comment }}
createTime  = {{ $json.createTime }}
reviewReply = {{ $json.reviewReply }}

Draft the reply. Add a LangChain AI Agent named AI Agent with a Google Gemini Chat Model attached. Give it a tight system prompt and feed it the review:

System: You are a helpful assistant. Write simple and short replies to the
        customer's feedback.
 
User:   This is the customer review: {{ $json.comment }}
        Customer name: {{ $json.displayName }}

Post the reply. Add the node that writes the reply back to Google Business Profile. In the shipped file this is a No Operation placeholder named Add Reply to Google Business Profile; swap it for an HTTP Request to the Business Profile API once your credentials are in place. Keep a human approval in front of the post step until you trust the tone.

6. Import the workflow JSON#

Grab the ready-made workflow with the download button above, or export your own by opening the workflow in n8n and using the menu to Download the JSON. On the target instance, import the file, then re-create or re-select the three credentials from Step 2. Finally, update the Google Sheet document id, the two tab names, your Gmail sender, and the review form URL to match your setup.

Testing the workflow#

Validate each phase before pointing it at real customers:

1. Run the request batch. Add one test row to Customers with Review Email set to Not Sent, run the schedule branch manually, and confirm the email arrives and the row flips to Sent.
2. Submit a happy rating. Open the form, pick five stars, and confirm it sends you to the Google review link with no row added to Reviews.
3. Submit an unhappy rating. Pick two stars, type a comment, submit, and confirm a new row appears in Reviews with Handle With AI Or Manually = Manually and Replied Or Not = Not Replied.
4. Draft a reply. Run the review-reply branch with a sample review and confirm the agent returns a short, on-topic reply before any post step.

If the webhook never fires, check CORS and the form's POST URL first; a blocked cross-origin request is the most common failure point.

Customization options#

• Delay the ask. Use Date of Purchase in a filter so the request only goes out a few days after the sale, when the experience is fresh but settled.
• Swap the channel. Replace the Gmail node with WhatsApp, SMS, or Telegram if your customers prefer it; the rest of the flow is identical.
• Swap the model. Replace Gemini with OpenAI, Anthropic, or any chat model n8n supports. The agent and prompt stay the same.
• Auto-reply to five-star reviews. Let the AI agent handle clearly positive reviews end to end, and keep critical ones in the manual queue.
• Add a follow-up. Email customers who were sent a request but never responded a gentle reminder after a few days.

Common mistakes that quietly break this#

• No double-send guard. If you skip the If check and the status update, a customer can get the same request on every run. Always flip Review Email to Sent after emailing.
• Filtering in n8n instead of the form. If unhappy customers are routed to Google and only filtered afterward, the bad review is already public. The split has to happen on the form, before the click.
• An open webhook. Anyone who finds the URL can inject fake reviews into your log. Restrict CORS to your form domain and validate the payload before writing.
• Auto-posting AI replies. A tone-deaf automated reply to a real complaint makes things worse. Keep a human approval in front of the post step until you trust it.
• Treating Sheets as disposable. The Reviews tab is your audit log and follow-up queue. Match on a stable key (Email) and keep the status columns accurate.

Conclusion#

You now have a review engine that asks at the right moment, protects your public rating, and keeps a human in control of the replies. Happy customers flow to your Google listing, unhappy ones land in a private queue you can actually act on, every response is logged in Google Sheets, and a Gemini agent drafts the replies so your team only has to approve. More five-star reviews, fewer public surprises, and far less manual chasing.