Do Google reviews affect whether AI recommends me?

What do reviews plausibly do, and through which engine?

They act as corroboration, and the clearest path runs through Gemini, because Gemini answers with Google Search grounding.

Follow the chain and every link is documented: reviews sit on your Google Business Profile, the profile appears on Google Search and Maps, and Gemini is queried with Google Search grounding. Our own scanner asks it that way, through the Generative Language API, and we publish that on The Receipts Engine.

For ChatGPT the path is a page rather than a platform. A review site listing, a best in Sydney roundup that cites ratings, a forum thread comparing three suppliers: if those are what its crawler returns for the question, the review content inside them is in scope. If they are not, your rating is invisible to that answer no matter how good it is.

What reviews supply that your own website structurally cannot is an independent second voice saying the same thing about you. That is the same property that makes a directory entry, a licence register or a news article useful, and it is the variable our own data keeps pointing at.

What did our study not measure?

Review volume and review rating. Neither is a field in the dataset.

We publish every column we recorded: industry, business name, domain, city, an anchor-brand flag, scan date, overall AI visibility score, site readiness score, per-engine questions asked, per-engine times named, a recognised flag, per-engine visibility scores, and the four site checks. There is no review column, so there is no correlation for us to report.

That is a deliberate line, not an oversight in the writing. Where we do have a measurement we state it with its denominator: the correlation between site readiness and mean per-engine visibility is Pearson r = 0.116 across the 137 businesses with complete records from both engines. That is what a measured claim looks like on this site. We do not have one for reviews, and a vendor who does should be able to hand you their sample, their period and their method the way we hand you ours.

Source: Answerable, The State of AI Visibility in Australia 2026. 300 Australian businesses, six industries, scanned 30 July 2026. Raw data published as CSV and JSON under CC BY 4.0.

What does a review need to contain to be useful to a machine?

The service, the place and the specifics, in the customer's own words.

A five-star review that says Great job, highly recommend corroborates nothing. There is no service in it, no location, no outcome, nothing for a retrieval system to match against a buyer question. A hundred of them still corroborate nothing.

A review that says which job was done, in which suburb, and how it went gives a system actual text to work with. You cannot write that, and you must not ask anyone to include wording that is not their own. What you can do is ask at the right moment and ask an open question: What did we do for you, and how did it go? produces far more usable text than Please leave us five stars.

Two other properties do work without any wording at all. Recency, because a business with nothing recent reads as one that may have stopped trading. And spread across platforms, because corroboration from one source is a single point of failure.

Which review platforms show up in Australian answers?

Start with Google, then find out empirically instead of taking a list from a vendor.

Google reviews come first because of the grounding path above, and because Google reviews feed the profile that other listings copy from. After that, the platforms Australians actually use vary sharply by category: ProductReview.com.au for consumer products and services, Trustpilot for e-commerce, hipages and Oneflare for trades, Whitecoat and HealthShare for health services, plus whatever the dominant directory is in your industry.

How to find out which of them matters for you, rather than for someone else:

1. Ask five buyer questions in your category and city, with no brand names in them.
2. Open every citation the answer carries and write down the domain.
3. Repeat with different phrasings, on different days, in both ChatGPT and Gemini.
4. The domains that keep reappearing are your list. It will not match your neighbour's.

We are giving you a method rather than a ranking because we do not have measured data on per-platform influence in Australian AI answers, and neither does anyone quoting you a percentage without a published sample.

Where is the line between encouraging reviews and buying them?

Encouraging is asking every customer. Buying is paying for the opinion, or filtering which opinions get through.

The ACCC is direct about it: It's against the law for a business to create fake or misleading reviews or to arrange for others to do so. Its own example of misleading includes reviews written by family, employees, or people paid in some way by the business, without stating the personal connection or commercial relationship.

Incentives are not banned outright, but they come with conditions. The ACCC's guidance is that incentives must be applied regardless of whether the reviewer leaves a positive or negative review and clearly disclosed so consumers know the review was incentivised. The practice that draws enforcement is the other one: suppressing or editing negative reviews, or only ever asking the customers you already know are happy.

This is not legal advice and we are not lawyers. The ACCC's own page is short and worth reading before you set up any review request process.

Source: ACCC, online reviews for products and services.

Find out what the answers actually cite.

Instead of guessing which review platform matters, run a free scan: we ask the buyer questions and show you the answers and the sources that came back.