Over the last week of August I ran the same set of checks across 100 service businesses in Sydney’s Eastern Suburbs.
Four businesses in each of 25 categories: physios, gyms, tattoo studios, tutoring centres, wedding venues, driving schools, dance schools, day spas, caterers, chiropractors, swim schools, music schools and more.
Between them they have earned 13,015 Google reviews. The median business has 59. The largest has 1,607.
Here is what the checks found.
62 of the 100 were not named when I asked Google for a recommendation
On Sunday 30 August I asked Google the question a customer would ask. Best physiotherapist in Bondi Junction. Best wedding venue in the Eastern Suburbs. Best driving school. Twenty-one searches covering all 100 businesses.
Google answered with an AI Overview, or a local map pack, or both.
- 23 were named in the AI Overview
- 14 appeared in the map pack
- 1 was referenced indirectly
- 62 did not appear at all
Nearly two thirds of established local businesses, several of them the biggest in their category by review count, were simply absent from the answer their own customers are being given.
Reviews help, but they are not the ticket in
The obvious theory is that Google names whoever has the most reviews. The data partly supports that and then breaks it.
Businesses that were named have a median of 111 reviews. Businesses that were not have a median of 40. So review volume clearly matters.
But 14 businesses with more than 100 reviews were not named at all, six of them with more than 250. Meanwhile three businesses with fewer than 25 reviews were named. One language school with 4 reviews was named first in its category, while a competitor with 38 was left out entirely.
So reviews are necessary and not sufficient. Something else decides, and whatever it is, it is not simply who has been around longest or who has the most stars.
I want to be straight about the limits here. This is 100 businesses, one afternoon, one location, one set of searches. AI answers vary by person and change constantly. Treat it as a snapshot, not a law.
The most common mistake takes an afternoon to fix
Of the 63 businesses whose websites I could check conclusively, 37 show no reviews on their own site at all. Not a rating, not a testimonial, not a count. A few link out to Google and leave it at that. One had placeholder testimonials still live, quoting a person called “John Doe Lee”.
Those 37 businesses have earned 2,249 Google reviews between them.
Think about what that means for a stranger. They search, they find you, they land on your website, and the single most persuasive thing about your business is not on the page. It is one tab away on Google, where they will not go back to look.
A tattoo studio in the group has 1,607 reviews and shows none of them. A physiotherapy practice has 148 and shows none, while its condition pages already bring in search traffic. The work of earning them is done. The work of using them has not started.
Almost a third of them are hard to contact
While checking, I tried to find a published email address for each business. Not a scraped one, an address they publish themselves.
- 71 publish one
- 16 offer a contact form and no address
- 13 offer neither
Three had no website at all, operating from an Instagram profile or a booking platform listing.
For businesses whose customers are increasingly likely to be an AI assistant acting on someone’s behalf, being uncontactable except through a form is a bigger problem than it was two years ago.
The AI file almost nobody chose to have
There is a convention called llms.txt, a plain text file at the root of a website that tells AI systems what the business does and which pages matter. It is early, unofficial, and adoption is thin.
Of the 64 sites where I could check conclusively, 17 had one. That is 27%, far higher than I expected.
Then I read them. Almost all were generated automatically by Wix, listing the homepage and nothing else. Two were written deliberately, and both of those businesses were named by Google in their category.
Two out of a hundred is not evidence of anything. It is worth noticing, and it is not worth acting on yet.
What I would do first
If you recognise your own business in any of this, the order matters more than the list.
Put your reviews on your website. Not a link to Google, the actual count, the rating, and three quotes with real names. This is the highest-return hour available to most of the businesses I looked at, and 37 of 63 have not spent it.
Publish an email address. A form is fine. A form instead of an address is a closed door to anyone who wants to reach you outside your website.
Then look at whether you are being recommended. Ask Google the question your customers ask. If you are not in the answer, you now know what the other 62 know.
How this was done, and what it does not prove
All checks were run on Sunday 30 and Monday 31 August 2026 against publicly available information: each business’s own website, its Google listing and its public social profiles. No client data, no private analytics, no paid tools.
The AI visibility check was 21 Google searches covering all 100 businesses, from one location, on one day, logged out. Results vary by searcher and change without notice.
The website checks were run by fetching each site’s pages directly. That method cannot see anything a page draws with JavaScript after loading, so nothing here claims a business has no tracking, and nothing claims a booking system is absent when it may simply load later. Where a site blocked automated requests, it is excluded rather than guessed at. Denominators differ between findings for that reason and are stated each time.
Businesses are described by category, not by name, except where a detail is already public on their own website.
Katie-Jane Michael
Founder, Human Theory · Connect on LinkedIn