Who Google’s AI Recommends: 446 Searches Across 262 Victorian Suburbs

What we sampled: 446 Google searches in the form “who are the best [trade] in [suburb]”, across 262 suburbs and towns in Victoria (metro Melbourne and regional) and six trades. Some searches added “Melbourne” or “Victoria” after the suburb, and bathroom and kitchen renovators were searched separately. On 23 September we re-ran the same searches through a search data service, on desktop and on mobile. Victoria only, and Google’s AI answer only.

We ran the first searches while researching local trade businesses for our own outreach, and saved every answer Google gave. This page reports what those answers show, and what changed when we looked again.

How often the AI answer shows up

In the first capture, an AI answer naming local businesses appeared on 375 of 446 searches.

For most of these trades, a homeowner asking this kind of question is now likely to read an AI answer that names a shortlist of businesses before anything else.

How many businesses it names

We counted the businesses named in the 369 answers where the names could be read.

Across every answer, several hundred different businesses were named. Everyone else trading in those suburbs was left out of the answer entirely.

The answer is not limited to businesses inside the suburb searched. Asked for the best pool builders in Docklands, Google’s AI answer said there are very few residential pool builders physically located there, then listed Melbourne pool specialists that service the area. One of them was based in Somerville on the Mornington Peninsula, roughly 50 km from Docklands.

Run it again and the list changes

On 23 September we re-ran 440 of the same searches, once on desktop and once on mobile, with the location set to the suburb searched wherever Google offered it.

Electricians and cabinet makers saw the biggest drop in answers. Heating and cooling, and bathroom and kitchen renovators, rose, plumbers rose from a small base, and pool builders slipped slightly. The by trade figures are in the appendix.

One city, nine points: checking a service area with Local Falcon

A single search only tells you what Google showed one person, in one spot, at one moment. Your customers search from all over your service area. To see the whole picture we use Local Falcon, which runs the same question from a grid of points across a suburb and records what Google’s AI answer says at each one.

Here is a real example from our own work: “who are the best concreters in Geelong”, scanned on a 3 by 3 grid with points one mile apart across central Geelong, on Google AI Overviews. We ran the identical scan twice on 14 September 2026, about four hours apart. The business we were checking for is left unnamed, as are the others.

Local Falcon’s data point export shows exactly who was named at each point. Across the eight points with an answer the AI named four to six concreters each time, 39 mentions in all. Two concreters were named at all eight points and a third at seven, but only one pair of points gave exactly the same line up. The rest of the list came and went from one side of town to the other.

In this grid the two most named businesses held their spots across both runs while the rest of the list reshuffled within hours. Across our 446 searches, even the leading names often changed over a few weeks. Both show why we never judge a client’s AI visibility from one search. Local Falcon also lists the pages Google’s AI drew on at each point, which in this scan included Google’s own pages, the businesses’ own websites and a local Reddit thread asking for concreter recommendations.

Where the AI gets its names

In the 23 September recheck we also saved the source links Google attached to each desktop AI answer, and the map pack (the three businesses shown on the map) from the same results page.

What the named listings look like

For three trades in the first capture (pool builders, bathroom and kitchen renovators, and cabinet makers in September) we saved the full answer text, including the star rating and review count Google showed beside each name. That gives 406 rated listings. The same business can appear in more than one suburb’s answer, so this counts listings, not different businesses. The recheck added 600 rated listings across all six trades.

What it means for a tradie

If the AI answer names about four businesses per suburb, being fifth is close to being invisible on that search. Three things follow from the data:

• The spots are winnable. Plenty of small businesses with a couple of dozen reviews are being named.
• The list moves. Being left out today is not permanent, and being named today is not guaranteed next month.
• The basics carry over. The businesses named overlap heavily with the map pack, and the websites Google cites often name the suburb in the page title or web address.

Methodology

First capture. Each search was run in a signed out browser in the form “who are the best [trade] in [suburb]”, and the AI answer was saved, either as the list of named businesses or as the full answer text. One search per trade per suburb (bathroom and kitchen renovators separately). Where the same trade and suburb was searched twice, only the later result is counted. A search counts as showing an AI answer if an answer naming local businesses appeared. Answers that failed to load, or appeared without naming any business, are counted as no answer. In 6 of the 375 answers the names could not be read, so name counts use 369 answers.

Recheck, 23 September 2026. The same searches were re-run through the DataForSEO search results service, once as a desktop search and once as a mobile search, with the location set to the suburb where Google offers that suburb as a location (382 of 440 searches) and to Victoria otherwise. Wording was standardised to “who are the best [trade] in [suburb] Victoria”, and heating and cooling searches used “heating and cooling companies”. Business names were compared after removing punctuation and words like “Pty Ltd”. Names, sources and the map pack come from the desktop results; the mobile results were used for whether an answer appeared.

Grid scans. Local Falcon, Google AI Overviews, 3 by 3 grid, points one mile apart, centred on central Geelong, run twice on 14 September 2026 for the question “who are the best concreters in geelong”. Counts are taken from the brands mentioned table and the data point export in each Local Falcon report, and the map image is Local Falcon’s own export.

Questions about the data, or want to use it? Email thomas@5startradie.com.au.

Limitations

Read the figures with these in mind.

• Not a random sample. The suburbs and trades are ones we were researching for our own outreach, so they lean toward places with active trade businesses.
• Different tools. The first capture used a browser and the recheck used a data service, with slightly different wording for some trades. Part of the change between them may come from the method, not from Google. The device disagreement within the recheck does not have that problem.
• One moment in time, twice. AI answers change from search to search and day to day, as the recheck and the grid both show.
• Small groups. Plumbers (25 searches) and electricians (57) are small samples. Treat those trade figures as indicative.
• Google’s AI answer only. ChatGPT, Gemini and Google’s AI Mode tab were not measured.
• Review figures cover only listings with a rating shown, and name counts can slightly undercount where a business was named without a rating card.
• The grid example is one question in one city, run twice. It illustrates the pattern; it is not a second sample.
• Anonymised. We report counts, not which businesses were or were not named.