Which Australian sources AI assistants reach for, counted across 1,010 hosts on one day

They know a few international directories, they know the Australian government far better, and the gap between those two facts is where a listing budget should go. Across 64 blind readings taken for Australia on 22 October 2026, Australian government hosts appeared 139 times and the international directories appeared about 25 times between them, while several of the Australian directories a business would expect appeared not at all. If you are choosing between a paid listing and getting your registrations and official record in order, the second one is what the engines are reading.

The method, stated so it can be re run

We put 33 batches of buying questions to ChatGPT and Gemini with Australia written into the text of every question. That is not a style choice. The search interface we use accepts a country parameter, a language parameter and a location parameter, returns HTTP 200 for all of them, and ignores every one, so the only reliable way to take an Australian reading is to put Australia in the prompt. The snapshot taken that evening holds 64 blind readings, 32 on ChatGPT and 32 on Gemini. The brand batch, which names us in the question, is excluded because a question containing the answer is a ceiling measurement rather than a blind one.

Those 64 readings returned 1,010 distinct hosts, of which 435 carry an Australian domain.

What the engines actually reached for

Source type	What appeared, and in how many of the 64 readings
Australian government	66 distinct hosts, 139 appearances, present in 25 readings. business.gov.au 15, accc.gov.au 13, acma.gov.au 8, oaic.gov.au 6, legislation.gov.au 5, business.qld.gov.au 5, nsw.gov.au 4, consumer.gov.au 4, buy.nsw.gov.au 4, service.nsw.gov.au 3, ato.gov.au 3, ai.gov.au 3
Australian specialist sites	localdigital.com.au 9, digitalnomadshq.com.au 8, 3pdigital.com.au 7, soupagency.com.au 7, firstpage.com.au 7, odindigital.com.au 6, iabaustralia.com.au 6, profitgeeks.com.au 5, rocketagency.com.au 5
International directories	clutch.co 11, designrush.com 6, agencies.semrush.com 5, goodfirms.co 4, sortlist.com 2, staging.clutch.co 2, capterra.com.au 1, g2.com 1
Australian directories and lead platforms	hipages, Gumtree, Airtasker, Oneflare, Yellow Pages, True Local and Yelp appeared in none of the 1,010 hosts
Australian media and industry bodies	iabaustralia.com.au 6, adnews.com.au 4, theaustralian.com.au 3, auda.org.au 4

How wide the Australian source pool is, for scale

1,010 distinct hosts across 64 readings is a wide pool and the width matters for budgeting. Our Singapore reading taken the same day with the same method returned 757 hosts across its own 64 readings, so Australia drew roughly a third more sources for the same number of questions. More states, more regional variation and more government layers produce more candidate pages, which means a single page competes against more alternatives here than in a smaller market. That is an argument for depth on fewer subjects rather than thin coverage of many.

Three findings a listing budget should turn on

The state outranks the directories many times over. 66 distinct government hosts appearing 139 times, against roughly 32 directory appearances in total. If your category touches consumer law, communications, privacy, tax or procurement, the engines are reading a regulator's page about the rules and then looking for a supplier who visibly meets them. That is a filings and compliance job rather than a listings job.

The Australian lead platforms were absent entirely. Zero appearances for hipages, Gumtree, Airtasker, Oneflare, Yellow Pages and True Local across 1,010 hosts. Those platforms still produce jobs inside their own apps, and they produce nothing in this channel, so a business whose entire presence is a platform profile has no presence here at all.

A directory's staging site got cited. staging.clutch.co appeared in 2 of the 64 readings. A staging environment is a directory's unfinished copy, and a buyer who followed it would be reading a page nobody meant to publish. What a citation proves is that the engine reached a url on that date, and it says nothing about whether the url was the one the publisher intended to serve.

The domain registry appearing is worth a line of its own

auda.org.au, the administrator of the .au domain space, appeared in 4 readings. That is an unusual host to see in a marketing answer and it points at a real Australian question: which domain shape a business should hold. The practical reading is that the engines treat the registry as the authority on that question rather than any vendor blog, so a page arguing about domain choice is competing with the registry's own documentation.

What is not in this dataset

Two things, and both matter for how far these numbers can be pushed. There is no per industry rate here, because the questions were marketing and supplier buying questions across a spread of sectors rather than a sector by sector sample. And this is ChatGPT and Gemini. Perplexity, Claude and Google AI Overviews were not in this run. Our product watches five surfaces; this measurement covers two, and the two stay separate in every sentence. Our British equivalent, the same question measured for the United Kingdom across 64 readings, was taken the same day with the same method and gives a different answer for a different market, which is the reason they are run separately.

Questions about Australian listings and AI answers

Should I cancel my directory listings

No. Keep a free tier anywhere a buyer genuinely looks and stop treating a paid listing as a visibility channel. Across our earlier three market audit of 597 readings, review directories were 0.8 per cent of all citations recorded.

Which government page should I care about

Whichever one governs the claims you make. The engines reached the consumer regulator in 13 readings and the privacy regulator in 6. If your sector has published guidance, your page should visibly answer to it.

Does a com.au domain help

435 of the 1,010 hosts in this set carry an Australian domain, so the engines are finding Australian sites in quantity. The page's subject is doing the work rather than the suffix, and the registry's own guidance is the authority on what the suffix requires.

How do I find which of my questions have no supplier in the answer

Fix a list of the questions your buyers ask, put Australia in each one, and run them on a schedule, recording the date and the sources. Our guide to measuring a mention rate without inventing an industry benchmark covers reading the result.

How often does this change

Enough that a reading carries its date. This one is 22 October 2026 and the scorecard behind it was still being written that evening, which is why the denominator here is the snapshot we froze.

The six platforms that reached none of 1,010 hosts

The clearest single fact in this dataset is an absence. Across 1,010 distinct hosts returned by 64 blind Australian readings on 22 October 2026, there was not one appearance from hipages, Gumtree, Airtasker, Oneflare, Yellow Pages or True Local. Seek did appear, across four of its hosts, and it appeared on the employment side of the question set rather than the supplier side. Those six platforms still produce work inside their own apps and they produce nothing in this channel, which means an Australian business whose entire presence is a platform profile has no presence here at all. The counterweight is that the domain registry itself, auda.org.au, appeared in 4 readings, and 66 distinct Australian government hosts appeared 139 times across 25 of the 64, so the sources that do win are institutions and small specialist sites rather than marketplaces. AI Knows Us runs a fixed question list across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews and shows the sources behind each answer, which is how a platform dependent business finds out what its own category looks like. Our own domain was among the 1,010 exactly zero times, which is the baseline we publish rather than the one we would choose.