Measured 2 August 2026. Follow-up probes 3 August 2026. This is the first edition of a monthly series. The narrative version of one slice of this data is in Nobody owns the answer yet; this page is the full run.
Summary
On 2 August 2026 we ran 38 customer-style queries through Google AI Overview and ChatGPT search for three Australian local-services categories. The two engines shared only 11.7% of the domains they cited. Google AI Overview leaned on forums and video. ChatGPT search leaned on directories. This is one operator's snapshot, not a market study.
Method, stated plainly
We ran one measurement round. Everything below comes from it.
| Item | Detail | |---|---| | Date of the main run | 2 August 2026 | | Date of follow-up manual probes | 3 August 2026 | | Queries | 38 | | Query sets | 3 (cat grooming Melbourne, 8 queries; house painting Canberra, 22 queries; SEO and websites for tradies Canberra, 8 queries) | | Engines | 2 (Google AI Overview, ChatGPT search) | | Engine checks attempted | 76 | | Checks that failed on the provider side | 2 | | Checks that returned an answer | 65 of 76 | | Citations captured | 447 across 223 unique domains | | Locations | Google AI Overview at city level: Melbourne for the grooming set, Canberra for both Canberra sets. ChatGPT search is country level only, so all three sets ran as "Australia". | | Data source | DataForSEO SERP `ai_overview` references for Google, DataForSEO LLM Scraper for ChatGPT search | | Total spend on the dataset | USD $0.28 |
### How the queries were phrased
Queries were written the way a customer would type or speak them, not as keyword strings. Examples, verbatim from the frozen sets: "how much does cat grooming cost in Melbourne", "my cat is badly matted what should I do", "how much do painters charge per room in Canberra", "do painters need a licence in the ACT", "how do i get my trade business to show up in chatgpt".
Each query ran once per engine.
### Sample size limits, said up front
- 38 queries is a small sample. It is three trade categories in two cities. It is not representative of Australian local search.
- One run per query. AI answers vary between runs. A single run is a snapshot, not a distribution. Nothing here is a stable average.
- One measurement round, so there is no trend. No number on this page can be compared to a previous number, because there is no previous number. Any "up X% since" claim about Australian AI citations, including from us, is not supported by data that exists.
- Two of the 38 queries' subjects are businesses tyso works with (ANT Painting is a family business Tyson co-runs; tyso.io is our own site). Their domains appear in the table below and are marked.
- This is a snapshot of what engines cited, not of what customers ask. No AI query-volume data exists for Australia, or anywhere. We do not know how many real people ask these questions.
- API answers are not app answers. Personalisation, memory and app-side retrieval mean chatgpt.com can answer differently from the API path we measured.
- Two calls failed (one Google check in the grooming set, one ChatGPT check in the painting set). They are excluded from rates, not counted as zeroes.
### Source classification
Domains were sorted by hand into directory or marketplace, forum or social or video, business website, cost guide, and government or industry body. The classification is a judgement call, so the domains in each bucket are named in the findings below and the raw citation lists are in the source files. Anyone can re-sort them.
Finding 1: the two engines barely cite the same sources
Across all 38 queries, Google AI Overview cited 130 unique domains and ChatGPT search cited 118. Only 26 domains appeared in both. That is 11.7% overlap (222 domains in the union). These counts exclude `google.com`, which appears in the 223-domain total only as Google's own map entity cards.
Per query set:
| Query set | AIO domains | ChatGPT domains | Shared | Overlap | |---|---:|---:|---:|---:| | Cat grooming, Melbourne | 19 | 33 | 6 | 13.0% | | House painting, Canberra | 71 | 49 | 17 | 16.5% | | Tradie marketing, Canberra | 52 | 42 | 5 | 5.6% | | All three pooled | 130 | 118 | 26 | 11.7% |
The practical reading: measuring "AI visibility" as one number hides more than it shows. Being cited by one engine predicted almost nothing about the other in this sample.
Finding 2: the engines lean on different kinds of source
Excluding Google's own map entity cards, Google AI Overview returned 228 citations and ChatGPT search returned 197.
| Source type | Google AI Overview | ChatGPT search | |---|---:|---:| | Forums, social and video (reddit.com, youtube.com, facebook.com, quora.com, tiktok.com, justanswer.com, bunnings.com.au workshop, linkedin.com) | 56 (25%) | 4 (2%) | | Directories and marketplaces (Airtasker, hipages, Oneflare, Localsearch, Word of Mouth, Yellow Pages, service.com.au, ThreeBestRated, Starworks, Groomzy and similar) | 23 (10%) | 69 (35%) |
Same questions, same week, opposite habits. reddit.com was cited 17 times by Google AI Overview and 4 times by ChatGPT search. youtube.com was cited 23 times by Google AI Overview and never by ChatGPT search. Airtasker went the other way: 3 Google citations, 10 ChatGPT citations.
One quirk worth naming: on the query "how much does cat grooming cost in Melbourne", Google AI Overview returned 22 of its 32 citations as `google.com/searchviewer` links, which are Google's own business entity cards rather than external pages. Those 22 are excluded from every count above.
Finding 3: AI Overviews did not appear for find-a-provider queries
Google AI Overview appeared on 28 of 37 valid checks (76%).
All nine queries with no AI Overview were find-a-provider phrasings:
- cat groomer near me Melbourne
- mobile cat groomer Melbourne
- best cat groomer south east Melbourne
- cat grooming Pakenham Victoria
- who are the best house painters in Canberra
- best painter for a full repaint in Belconnen
- painters available for pre-sale repaint before auction Canberra
- who does seo for tradies in canberra
- best website designer for tradies in canberra
Every cost, how-to, comparison and licensing query in all three sets returned an AI Overview.
The reverse does not hold, and we will not overstate it: two other find-a-provider queries ("reliable father and son painters Canberra", "most recommended exterior painters Canberra") did return an AI Overview. So the pattern is one-directional in this sample. Nine of nine no-AIO queries were find-a-provider, but not every find-a-provider query lacked an AIO.
For context, and labelled clearly: US-derived studies report a similar split, with simple transactional local queries triggering an AI Overview far less often than cost and research queries. No published study measures AI Overview prevalence in Australian SERPs. We are not extrapolating the US percentages, and neither should anyone quoting this page.
Finding 4: ChatGPT search answered everything, and often showed no sources
ChatGPT search returned an answer on 37 of 37 valid checks. Google AI Overview appeared on 28 of 37.
But ChatGPT search returned no source list at all on 19 of those 37 answers (51%). It answered the question and named businesses in prose without attributing anything. Every ChatGPT figure on this page is therefore computed over the 18 answers that did carry a source list.
That is the most uncomfortable finding here. Roughly half the time, an Australian asking ChatGPT a local-services question got an answer with no visible sources at all.
Finding 5: directories dominated the cat grooming answers on ChatGPT
In the 8-query cat grooming Melbourne set, ChatGPT search cited a directory or marketplace on 5 of the 8 queries, for 13 domain appearances in total:
| Directory | Queries cited on (of 8) | Raw citations | |---|---:|---:| | airtasker.com | 4 | 5 | | groomzy.com.au | 2 | 3 | | starworks.com.au | 2 | 2 | | localsearch.com.au | 1 | 1 | | wordofmouth.com.au | 1 | 2 | | oneflare.com.au | 1 | 1 | | yellowpages.com.au | 1 | 1 | | pupkus.com.au | 1 | 2 |
Airtasker was the single most-cited directory, on 4 of the 8 questions. The other seven accounted for 9 more query appearances between them.
Google AI Overview, on the same eight questions, cited a directory once.
One query is worth quoting in full. Asked "cat grooming Pakenham Victoria", ChatGPT search returned two vet clinics, a mobile dog groomer, a Petbarn store finder, a barber shop and four directories. It named no cat groomer's own website.
Finding 6: a small local business site did get cited, on one engine only
ANT Painting (antpainting.com.au) is a two-person Canberra painting business. In its 22-query set it was cited on 7 of 22 queries. All seven citations came from Google AI Overview. ChatGPT search cited it zero times.
Ranking it honestly against the rest of its own set:
- Within Google AI Overview, by number of queries cited on: reddit.com 8, antpainting.com.au 7, experiencepaintingcanberra.com.au 6, youtube.com 6. Second place.
- Within Google AI Overview, by raw citation count: youtube.com 16, reddit.com 9, antpainting.com.au 8, experiencepaintingcanberra.com.au 7. Third place.
- Across both engines combined, by number of queries: reddit.com 10, then antpainting.com.au and experiencepaintingcanberra.com.au tied at 7.
So "second most-cited in its own set" is true on one specific measure and not on others. All three are stated so nobody has to take our word for the flattering one.
Four pages earned those eight citations, and one page did most of the work. `/house-painting-cost-canberra/`, a plain-language cost article, was cited 5 times. `/do-painters-need-a-licence-act/`, `/heritage-painting-canberra/` and the homepage were cited once each. A fifth page, `/canberra-painters/`, was cited once in the separate tradie-marketing set. None of these pages rank at the top of Google for their terms. On the two queries where a Google organic position was recorded on the same SERP, the site sat at position 6 and position 7.
The other two businesses in this round were cited zero times across their 8-query sets. The tradie-marketing set's full story, including the 89 domains it scattered across, is told in Nobody owns the answer yet.
Cited domains, counts
Every domain cited on 3 or more of the 38 queries. "Queries" counts distinct queries, so a domain cited twice in one answer counts once. Raw counts are total citation instances.
| Domain | Type | Queries (of 38) | Google AIO citations | ChatGPT citations | |---|---|---:|---:|---:| | reddit.com | Forum | 15 | 17 | 4 | | youtube.com | Video | 9 | 23 | 0 | | airtasker.com | Marketplace | 8 | 3 | 10 | | antpainting.com.au \* | Business site | 8 | 9 | 0 | | experiencepaintingcanberra.com.au | Business site | 7 | 7 | 2 | | hipages.com.au | Marketplace | 6 | 6 | 5 | | whatsthedamage.com.au | Cost guide | 5 | 6 | 7 | | facebook.com | Social | 5 | 9 | 0 | | paintplus.net.au | Business site | 5 | 5 | 0 | | painters.edu.au | Industry body | 5 | 3 | 2 | | localsearch.com.au | Directory | 4 | 0 | 4 | | service.com.au | Directory | 4 | 0 | 4 | | modernizesolutions.com.au | Business site | 4 | 2 | 2 | | brushworks.com.au | Business site | 3 | 2 | 3 | | painters-canberra.com | Lead site | 3 | 1 | 3 | | thequoteyard.com.au | Cost guide | 3 | 2 | 4 | | wordofmouth.com.au | Review directory | 3 | 0 | 4 | | peterthepainter.com.au | Business site | 3 | 3 | 1 | | fantasticfelines.com.au | Business site | 3 | 1 | 2 | | pawsnbuddy.com.au | Business site | 3 | 1 | 2 | | fantasticats.com.au | Business site | 3 | 0 | 3 | | starworks.com.au | Best-of directory | 3 | 0 | 3 | | threebestrated.com.au | Best-of directory | 3 | 0 | 3 |
\* antpainting.com.au is a family business Tyson co-runs. It appears in this table on merit of the measured data, and it is flagged so the reader can discount it. Seven of its eight query appearances are inside its own query set. The eighth is in the tradie-marketing set.
The full tail is long: 223 unique domains across 447 citations, and 200 of those domains were cited on fewer than 3 queries.
What this suggests for Australian local businesses
Stated as observation from one small sample, not as advice that has been proven to work.
- "AI visibility" is not one thing. With 11.7% domain overlap, being cited by Google AI Overview says close to nothing about ChatGPT, and the reverse. Anyone selling a single AI visibility score should be asked which engine it measures.
- The two engines reward different assets. In this sample, Google AI Overview citations went to forum threads, video and specific informational pages. ChatGPT search citations went to directories and marketplaces. A business with neither presence was invisible on both.
- Cost and licensing questions were where the AI answers were. Every cost, comparison and licensing query returned an AI Overview. The find-a-provider queries mostly did not. The observable pattern is that the AI layer sits upstream of the hiring decision, not on top of it.
- A small site can be cited without ranking first. The one business cited here sat at organic positions 6 and 7 on the SERPs where its position was recorded, and was still cited on 7 of 22 queries. Citation and ranking were not the same thing in this data.
- Directory listings were cited far more than they were clicked-through evidence for. We measured citations only. We did not measure whether any of it produced a phone call, and this page makes no revenue claim of any kind.
- Half of ChatGPT's answers named nobody as a source. Any strategy that depends on being cited has to reckon with an engine that frequently cites nothing.
What would change the picture
The things most likely to make Edition 2 disagree with Edition 1:
- More runs of the same queries. One run per query is the weakest part of this data. Three runs per query with majority scoring would tell us how much of the above is noise.
- More query sets and more trades. Three categories in two cities cannot generalise. Plumbing, electrical, cleaning and legal would test whether the directory-heavy ChatGPT pattern holds outside grooming and painting.
- More engines. Perplexity, Claude and Bing Copilot were not measured. A 3 August attempt at Bing returned mismatched SERPs on all three calls, so no Bing data exists in this edition. Gemini grounding is excluded permanently: its terms forbid programmatic citation collection.
- The engines changing. Citation half-life is short and index composition rotates. Nothing here is claimed to be stable for a month.
- Better AI Overview capture. The 3 August follow-up sweep found 4 AI Overview stubs with no content returned, so citation status for those was unknown rather than absent. The 2 August method was stronger on that axis and is the one this edition reports.
What Edition 2 will measure
Committed now, so it can be checked later:
- The same 38 queries, same locations, same engines, re-run so the first genuine month-over-month comparison exists.
- Three runs per query instead of one, majority-scored, on at least one query set.
- At least one new trade category outside grooming and painting.
- Bing, retried, since ChatGPT retrieval draws on Bing's index and the 3 August attempt failed.
- A count of how often each engine names a business in prose without citing it, which this edition noticed but did not tally systematically.
Edition 2 is scheduled for early September 2026.
Verification
Every figure on this page comes from a dated file in our measurement archive. The query sets are frozen (one file per set, one query per line) and every citation list is stored raw, so the whole page can be recomputed from the files. We hand the raw files over on request: tys@tyso.io.
Who made this. tyso is a one-person growth studio in Canberra. This page is one small operator's measurement, run on its own clients and its own site, published because no Australian equivalent exists. It is not a market study and should not be cited as one.
Corrections. If a number here is wrong, tell us and we will fix it and date the fix. tys@tyso.io