Skip to content

What an AI-drafted review reply is and how it works

Checked October 27, 2026

An empty table and dark chairs in front of large windows
Photo: Pexels on Pixabay

An AI-drafted review reply is text that a computer program writes in answer to a customer's review. You hand the program the review. It hands back a few sentences. Then you decide what, if anything, to post.

That last step matters most. The draft isn't a reply yet. It's a suggestion.

What the tool is doing

The kind of tool behind these drafts is called generative AI. That means software that makes new text, pictures or sound instead of just sorting or searching existing ones. Its models learn patterns from the data they were trained on, then produce new content in response to a prompt. A prompt is simply the instruction or question you type in, usually in plain language.1

So when you paste in a review that says the plumber arrived late but fixed the leak, the tool doesn't know your plumber. It has no memory of the job. It predicts text that fits a reply to that sort of review.

That explains two things you'll notice straight away.

  • The draft reads smoothly, with a polite tone and tidy sentences.
  • The draft knows only what you told it. Anything else is filled in by pattern.
From review to posted replyCustomer posts a reviewYou give the review to an AI toolThe tool writes a draftA person reads and edits itYou post the final reply
The draft is only a starting point. A person decides what goes public.

What you'll see on screen

You paste a review. A few seconds later, a reply appears. It usually opens with thanks, mentions something from the review, and closes with an invitation to come back or get in touch.

Read that draft slowly. Smooth text can hide a wrong detail. A draft might name a service you don't offer, promise a refund you never planned to give, or apologize for something that didn't happen. The tool wasn't there, so it guessed.

Australia's privacy regulator warns that AI systems can give wrong or false answers. It also advises people to treat the outputs as statistically informed guesses.2 That's a useful way to think about a review reply draft. It's a guess at what you'd say.

What a person still does

A reply goes out under your business's name. Customers will read it as you speaking. So the person posting it has to do the checking.

Before posting, a person should:

  • Confirm every fact in the draft against what actually happened.
  • Remove anything the tool added that the customer never said.
  • Cut promises you can't keep.
  • Make sure no private customer details slipped in.
  • Change the wording until it sounds like someone who works there.

The next pages cover each of these. What to hand the tool in the first place is in the prompt page. The privacy side has its own page, on customer details.

What happens after you post

On Google, you can't reply until your business is verified.3 Once you post, the person who wrote the review is notified.3 They'll read it. A reply that sounds copied or off-target is easy to spot, and the reviewer is the one most likely to notice.

The notes on Google's reply rules list what Google's help pages say about replies.

Where AI helps and where it doesn't

It helps with a blank page. A first sentence is often the slow part, and a draft gives you something to push against.

It doesn't help with judgment. Whether to apologize, whether to offer to fix something, whether the customer is right: those are decisions about your business. A tool can't make them for you, because it doesn't know the facts. Use the draft to save typing, not to skip thinking.

For angry reviews, take extra care. See drafting a reply to a negative review.

Reference list

  1. Generative artificial intelligence, Wikipedia
  2. Guidance on privacy and the use of commercially available AI products, Office of the Australian Information Commissioner
  3. Manage customer reviews, Google Business Profile Help