Read every review reply Google's AI writes for you
A drafted reply takes seconds to approve and sits on your listing for years. Read it like a new hire's first customer email.
Co-Founder, Vyzz
The button that writes the reply for you
There is a new button turning up in some Google Business Profile accounts this year. It says "Reply to reviews with AI," and it writes an answer to a customer review for you. In the flow Search Engine Land described, you read that draft and submit it yourself, with whatever edits you want in between. Whether that step always holds is the part owners disagree about.
Search Engine Land reported the test in March. It has been seen in the U.S., Brazil and India, and it has barely shown up in Europe. Availability is uneven across accounts, and uneven from one review to the next inside the same account.
A local search specialist ran into that unevenness first-hand on his own listing. The tool steered him toward an old one-star review he had not answered yet, and the draft it wrote for that one came back decent. On his other unanswered reviews, the option never appeared at all. That report comes from Barry Schwartz at Search Engine Roundtable.
Owner accounts of this feature conflict, in that same Search Engine Land write-up, on whether a bulk-generated reply always waits for approval. Some owners say every draft still needs a manual submit. Others say replies went out with no edits. Treat that as unsettled and go look at your own account this week.
What the reply does after you post it
A review reply is public text with your business name attached to it. The next customer who opens your listing reads it, under the review that prompted it, for as long as both stay up. For years. It's also the kind of writing ChatGPT, Perplexity and Google's AI answers can pull from when somebody asks about a business like yours.
BrightLocal's 2026 Local Consumer Review Survey found 45% of consumers now use AI tools to find local services, up from 6% a year earlier.
Almost 9 in 10 of those AI users, 88% in the same 2026 BrightLocal research, check whether the reviews an AI puts in front of them are real, or where they came from. Somebody asks ChatGPT for a dentist, gets three practices back, then reads the reviews for all three. Your replies are sitting right there under those reviews, in your voice, with your name on them.
The customer in that moment has a name from a machine and no idea whether to trust it. So she opens the listing and reads how the business talks to people who were unhappy. That's the cheapest trust test there is, and it takes her under a minute.
Where a drafted reply goes wrong
Take a one-star review from a patient who waited 40 minutes past her appointment time while nobody at the desk told her why. The draft comes back polite. It thanks her for the feedback and invites her to call the office.
Now read that the way the next customer reads it. She learns that your office thanks people. The 40 minutes goes unmentioned, and the reply is the last word on that thread for years.
A reply that names what happened takes 90 seconds to write. She was kept waiting 40 minutes on a Tuesday because the surgery before hers ran long, the desk should have told her at minute 10, and the schedule now leaves a gap after long cases. Any customer who reads that learns something true about how the office runs. (I invented the Tuesday and the surgery. Your front desk knows the real reason, which is the whole point.)
The 5-star reviews are worth the same minute, for a different reason. Most owners have more of those than complaints, and a drafted thank-you burns the one spot on your listing where you can say what the visit was actually for. Name the same-day crown, the 6am emergency call, or the Saturday overnight cover in your answer, and that service goes into public text wrapped in a customer's own story.
Two other habits show up in drafted text. It can state something out of date or flatly wrong, like a service you stopped offering last year or a callback nobody has scheduled. A drafted reply promising a callback is a promise somebody at your desk now has to keep. And it repeats itself: 15 replies down one page carrying the same three sentences read like a machine wrote them, because one did.
What nobody has measured yet
I'd rather be straight about the limits here. Nobody has run a study showing that a better-worded review reply changes what ChatGPT says about your business. If anyone tells you they've measured that, ask to see the study.
What the research does cover is narrower and still worth your time. Close to half of consumers now use an AI tool when they look for a local service, and almost 9 in 10 AI users check whether the reviews they are shown are real, or where they came from. Your reply text sits under those reviews. 15 minutes a week is a fair price for owning the words on that thread, and that's the honest size of the claim.
The weekly habit that keeps this safe
Put it on the calendar for Monday morning, before the phones get going.
- Open your Google Business Profile and go to the reviews tab. Look for the AI reply option. If it's missing, check again next month, because the test is still moving.
- Read every draft out loud before you touch submit. A canned sentence looks fine on screen and sounds hollow in your own voice.
- Check the facts in it. The service it names, the hours, the price, the promise to call somebody back. Anything you cannot stand behind gets cut.
- Add one specific detail the tool could not know, like what actually happened or what has changed since.
- Post it, then read your last 10 replies in a row. If they blur together, rewrite two of them today.
Two is enough. You're breaking the pattern before it sets, and a year of old replies can wait for a slower week.
Treat that drafted reply the way you would treat the first customer email from a new front-desk hire. You would read it before it went out, and you would fix the line that sounded canned, because your name is on the bottom of it. Same job here, same 90 seconds, on text that stays public and gets read by both people and machines.
The button is a good thing. It gets the blank page out of your way on a Monday when you have a stack of unanswered reviews and a full schedule. The part that stays yours is the minute after the draft appears.