How to Automate Google Review Responses With AI (Without Sounding Like a Robot)

How to automate Google review responses with AI

A dental office gets 30 Google reviews a month. The office manager means to reply to every one, does it for two weeks, then falls behind, and now there are 47 unanswered reviews, including a two-star complaint about billing that has sat there for three weeks where every new patient can see it.

That is the job AI is good at, with one condition. Let it write and post the easy thank-you replies, and let it draft, but never post, the replies to anything negative. This guide shows how to set that up, in GoHighLevel if you already use it, or in n8n if you do not, and what each approach costs.

Why replies matter, and what they do not do

BrightLocal's 2026 Local Consumer Review Survey found that 89 percent of consumers expect business owners to respond to reviews, and 81 percent expect a reply within a week. Roughly four in five said they are likely to use a business that responds to all its reviews, and 42 percent said they are unlikely to use one that never replies. The same survey found that templated, generic replies put off about half of consumers.

That last number is the case against lazy automation. Fifty identical "Thank you for your kind words!" replies look worse than none. The point of using AI is that each reply can mention what the customer actually said, at a speed no office manager can keep up.

One thing replies are not: a proven ranking lever. Google's help page says replies show you are responsive to customers. It does not say they improve your position in Maps, and there is no official statement that they do. Reply because customers read them.

The rules Google enforces

Before you automate anything, know the lines. Google's contribution policy bans offering payment, discounts or free services in exchange for reviews, and bans discouraging negative reviews or selectively asking only happy customers, which is known as review gating. Reviews that break these rules can be removed. Nothing in your reply flow should touch those lines, and your reply text should never offer a discount for changing a review.

Replies themselves are checked against Google's content policies. Google says this usually takes up to 10 minutes and can take up to 30 days, and the reviewer is notified when you reply. Replies appear under the business name.

For health care, legal and similar businesses, there is one more rule that matters more than Google's: never confirm that the reviewer is a patient or client, and never mention anything about their visit. A reply that says "we're sorry your crown took three visits" can be a privacy violation. Your AI instructions need to say so in plain words.

The workflow in one picture

The design that works is a split by risk. Every hour, the system fetches reviews that have no reply. AI drafts a reply for each one. Five-star reviews with short or no text get posted automatically, because there is little that can go wrong in a warm, specific thank-you. Anything at four stars or below, or any review that mentions a problem, goes to the owner as a draft with an approve button. The owner can approve it, edit it, or pick up the phone first. Everything posted is logged.

Option one: GoHighLevel Reviews AI

If your business already runs on HighLevel, this is the shortest path. Reviews AI has three modes: Off, Suggestive, where the AI drafts up to four suggestions and a person picks one to send, and Auto Responses, where AI agents reply on their own. You connect your Google Business Profile in the sub-account, create a Reviews AI agent with your tone and rules, and choose the mode.

As of October 2026, HighLevel charges $0.01 per AI-assisted review reply sent on pay-per-use, and Reviews AI is unlimited, subject to fair use, on the AI Employee Growth add-on at $50 a month and the Unlimited add-on at $97 a month per sub-account. It is not included in the $20 AI Employee Starter tier. HighLevel's base plans start at $97 a month.

The sensible setup is Suggestive mode for everything at first, then Auto Responses only for five-star reviews once you trust the drafts.

Option two: build it in n8n

If you are not on HighLevel, n8n has a native Google Business Profile node with Review operations for Get, Get Many, Reply and Delete Reply. The catch is access: Google's Business Profile APIs are not open by default.

Step 1: Request Google Business Profile API access

Google's prerequisites say the profile must be verified and active for more than 60 days, and the business needs a website. Create a Google Cloud project, then submit the Business Profile API contact form, choose Application for Basic API Access, include your project number, and send it from an email that is an owner or manager on the profile. You can tell whether you are approved by the quota in the Cloud console: 0 requests per minute means pending, 300 means approved. Start this first, because approval is not instant.

Step 2: Connect n8n to your profile

In n8n, create a Google OAuth2 credential using your Cloud project's client ID and secret, with the business.manage scope, and sign in with the Google account that manages the profile. Add a Google Business Profile node, choose Review and Get Many, and pick your account and location. Run it once to confirm reviews come back.

Step 3: Trigger hourly and keep only unreplied reviews

Start the workflow with a Schedule Trigger set to every hour. After the Get Many node, add a Filter node that keeps reviews with no existing reply. Add a second check against a Google Sheet log of review IDs you have already handled, so nothing gets answered twice if a run fails halfway.

Step 4: Draft the reply with AI

Add an AI node, such as the Anthropic or OpenAI node, and pass it the reviewer's first name, star rating and review text. A prompt that works: "You reply to Google reviews for [business], a [type] in [city]. Write 2 to 4 sentences. Thank the reviewer by first name. Refer to one specific thing they mentioned, in your own words. Never offer discounts, refunds or anything of value. Never confirm they are a customer or patient or mention details of their visit. Never argue. If the review is negative, apologize once, do not make excuses, and invite them to contact [name] at [phone]. Do not use exclamation marks more than once. Do not start with 'Thank you for your review'." Also ask it to return a field saying whether the review contains a complaint, even when the rating is high, because some four and five-star reviews carry a real problem.

Step 5: Route by risk

Add an IF node. If the rating is five stars and the AI flagged no complaint, go straight to the posting step. Everything else goes to approval: n8n's Gmail and Slack nodes have a send-and-wait operation that pauses the workflow until someone clicks approve or replies with edits. Send the review, the draft and two buttons. When the owner approves, the workflow continues.

Step 6: Post and log

Add a Google Business Profile node with Review and Reply, pass the review ID and the approved text, and finish with a Google Sheets node that appends the review ID, rating, reply text, date and whether it was auto-posted. That log is how you audit the AI, and how you prove you are answering everything.

Step 7: Check a week of output

For the first week, set everything to approval, including the five-star reviews. Read every draft. If you are editing more than one in five, fix the prompt before turning on auto-posting.

A worked example: an HVAC company with two locations

Take an HVAC company with two Google profiles and about 40 new reviews a month between them. These numbers are illustrative. About 32 are five stars, five are four stars and three are one to three stars.

A five-star review says: "Carlos showed up on time, found the bad capacitor in ten minutes and explained everything. AC is cold again." The AI drafts: "Thanks, Jen. Glad Carlos tracked down the capacitor quickly and that the house is cool again. We'll pass your note on to him." The workflow posts it within the hour.

A two-star review says: "Tech was friendly but I was charged a diagnostic fee on top of the repair and nobody told me." The AI flags a complaint and drafts: "Hi Mark, thank you for the feedback, and we're sorry the diagnostic fee was a surprise. That should have been explained before any work started. Please call Dana, our office manager, at [number] so we can look at your invoice together." The owner gets the draft in Slack, calls Mark first, then approves an edited version. That call matters more than the reply.

The company's owner spends maybe ten minutes a week approving drafts instead of several hours writing, and every review gets an answer within a day.

What it costs

As of October 2026. GoHighLevel: $0.01 per AI review reply on pay-per-use, or unlimited within the $50 or $97 AI Employee add-ons, on top of a base plan from $97 a month. For 40 reviews a month on pay-per-use, that is 40 cents.

n8n: Cloud Starter is listed at 20 euros a month billed annually for 2,500 workflow executions, which covers an hourly schedule easily; self-hosting is free if you run your own server. AI model cost is close to nothing at this volume. Claude Haiku 5.5, for example, lists $0.10 per million input tokens and $0.50 per million output tokens, so a few hundred replies cost cents. The Google API itself has no fee.

Dedicated tools: Reviewflowz lists AI replies with auto-post or approval at $45 a location a month on Pro and $70 on Premium. It makes sense if you want no build at all and many locations to monitor; for one location, it costs more than the DIY route for the same result.

The real cost is time to set up. Expect an afternoon for the n8n build plus however long Google takes to approve API access, or about an hour in HighLevel.

When not to automate it

If you get five reviews a month, write them yourself. It takes ten minutes, and your own voice is better than any prompt. If you are in health care or law and cannot trust the prompt to stay clear of confidential details, keep everything on approval forever. And never let the AI auto-post replies to negative reviews. The best reply to an angry customer is a phone call, and no workflow makes that call for you.

Frequently Asked Questions