Two small jobs in one file. The first asks every finished customer for a review a day or so after the work, once. The second drafts a calm public reply whenever a new review arrives and emails it to you to approve, flagging anything sensitive. It asks everyone, offers nothing in exchange, and never posts anything for you.
Import it into n8n
The short lesson
- Why it matters
- People read reviews before they call. Asking every customer, every time, is the habit that builds them.
- Common mistake
- Asking only the customers you think were happy, or offering a reward. Both go against review-site rules.
- Next step
- Pull new reviews in automatically and track replies by location.
What's inside
- An n8n workflow that asks completed jobs for a review
- A second flow that drafts replies to new reviews
- A flag on reviews that mention injury, legal threats or safety
- The prompts, to use in any other tool
What it does
- Each morning it reads your completed-jobs sheet.
- It finds jobs finished a day or more ago that haven't been asked yet.
- The AI writes a short thank-you with your review link. Every customer is asked, and nothing is offered in return.
- The email is drafted or sent, and the sheet is marked.
- Separately, when a new review is sent to the workflow, the AI drafts a calm public reply and emails it to you, with a warning if the review is sensitive.
Set it up
Pick the path that matches the tools you already use.
- Download the workflow file from the box on this page. It's a .json file.
- In n8n, create a new workflow, open the ⋯ menu at the top right, and choose Import from file. Pick the file you downloaded.
- Nodes that need a login show a warning. Click each one and connect your own account. n8n walks you through it.
- Open the node called Settings and replace the sample values with yours.
- Put your Google review link in the first Settings node. The second flow's web address is on the second Webhook node, if you have a tool that can send it new reviews.
- Click Execute workflow to run it once with test data and check what comes out.
- When you're happy with it, switch the workflow to Active.
| You'll connect | What it's used for |
|---|---|
| Google Sheets | Holds your completed jobs |
| Gmail | Sends the request, and emails you the draft replies |
| An AI model | Writes the requests and the reply drafts |
| A reviews source (optional) | Anything that can send a new review to the workflow's web address |
| Google Sheet: Completed Jobs |
|---|
| Customer · Email · Phone · Job · Completed Date · Review Requested |
Using Claude, Gemini or another model instead of OpenAI? Delete the model node under each AI step, add the one you want, and drag its connector into the AI step.
New to n8n? Start here →The AI prompt
This is exactly what the n8n file sends to the AI. Square brackets are the details your tool fills in for each customer.
Review request: instructions
You write a short message asking a customer of [business] for a review. Rules: - Under 60 words. Thank them by first name and mention the job. - Ask once and include the link exactly as given. - Never offer a discount, gift or anything in exchange for a review. Never ask for a positive review, only for an honest one. Return only JSON with the keys "subject" and "body".
Review request: what you send it
Customer: [name] Job: [job] Review link: [reviewLink] Sign off as [owner], [business].
Reply drafter: instructions
You draft public replies to online reviews for [business]. Rules: - Under 80 words. Calm, specific, human. - 4 or 5 stars: thank them and mention one detail they wrote. - 1 to 3 stars: apologize for the experience, do not argue, do not make excuses, and invite them to contact [contact] so it can be fixed. - Never admit legal fault. Never reveal private details about the customer or the job. Return only JSON with the keys "reply" and "flag". Set "flag" to true if the review mentions an injury, a legal threat, discrimination or a safety problem. Otherwise false.
Reply drafter: what you send it
Platform: [platform] Reviewer: [reviewer] Rating: [rating] out of 5 Review: [text]
Test it before it touches a customer
- Add one finished job with your own email and check the request reads naturally.
- Send the reply flow a fake 5-star and a fake 2-star review and compare the tone.
- Send one that mentions an injury and check it's flagged for your attention.
Ask every customer, not just the ones you think were happy, and never offer a discount or gift for a review. Review platforms discourage both, and can remove reviews or penalize listings over them.
When you'll outgrow this
This is the right size for most small businesses. These are the signs it's time for something sturdier.
- You want reviews pulled in automatically from Google and other sites rather than sent to the workflow.
- You run several locations and want reviews and replies tracked by branch and by team member.
- You want to monitor and respond to reviews across many platforms in one place.