When ChatGPT or Google AI Mode recommends a business, it almost always explains why. And the explanation is nearly always a quote from that business’s own website. “They have been operating since 1998.” “They offer same day callouts.” “They are licensed under REC 12345.”
Which means the businesses that get recommended are the ones with something quotable on the page.
So we checked how many Australian businesses have that. In August 2026 we read the homepages of 500 Australian websites, all of them pulled from live customer searches across solar, legal, building, trades, dental, finance, health and removals. We looked for six specific facts.
The average homepage published 1.94 of the six. Only 3% published all six. 22% published none of them.
The six facts
These are the details that show up inside AI recommendations, in rough order of how often we see them quoted.
| Fact | Example | Australian homepages that publish it |
|---|---|---|
| Areas or suburbs served | “Servicing all Melbourne metro suburbs” | 42% |
| A price signal | “From $890 fully installed” | 39% |
| Licence, ABN or accreditation | “REC 12345” or “QBCC 987654” | 33% |
| Year founded or years in business | “Family owned since 2004” | 31% |
| Response time or availability | “Same day service, 7 days” | 28% |
| Review count or star rating | “4.9 stars from 312 Google reviews” | 21% |
Not one of the six is published by even half the market.
How thin it gets
| Facts published | Sites | Share |
|---|---|---|
| None | 108 | 22% |
| One | 115 | 23% |
| Two | 112 | 22% |
| Three | 79 | 16% |
| Four | 51 | 10% |
| Five | 22 | 4% |
| All six | 13 | 3% |
Two thirds of Australian business homepages publish two facts or fewer. Nearly a quarter publish nothing an AI could quote at all.
These are not empty websites. The median homepage we read had 9,681 characters of text, and only 4% had under 1,500. There are plenty of words. They are just words like “quality workmanship you can trust” and “your local experts” instead of “operating since 2004” and “licence 12345”.
Why this decides who gets recommended
An AI assistant building a shortlist has a problem. It needs a reason to name one plumber over another, and it will not invent one. It reads pages and looks for a claim it can repeat and attribute.
Give it “trusted local experts with quality service” and there is nothing to lift. Give it “REC licensed, servicing 42 Melbourne suburbs, same day callouts, 4.9 stars from 312 reviews” and it has four reasons to name you and a sentence for each.
This is why brand copy that would pass a marketing review can still lose you the recommendation. The AI is not judging whether your copy is good. It is looking for something checkable.
The structured data gap
We also looked at the code behind the pages.
| What we checked | Share of homepages |
|---|---|
| Has any JSON-LD structured data | 81% |
| Declares LocalBusiness or a trade subtype | 37% |
| Has FAQPage schema | 16% |
| Declares opening hours | 20% |
| Declares a postal address | 40% |
Four out of five sites have structured data. Most of it says WebSite, Organization and WebPage, which tells an AI almost nothing it could not work out. Only 37% actually declare what kind of business they are in a form a machine reads.
That gap is mostly a plugin default nobody changed. It is also one of the quicker things to fix. We covered the specifics in schema markup for AEO.
What good looks like
The 13 sites that published all six were not fancy. They tended to have a short block near the top of the homepage that read something like this:
“Melbourne electricians since 2004. REC 18422. Same day callouts across 42 metro suburbs, 7 days. Fixed price quotes from $180. 4.9 stars from 312 Google reviews.”
Five sentences. No design change. Every one of them checkable, and every one of them the kind of line we see quoted back inside AI answers.
You do not need a rebuild for this. You need a paragraph.
What to do this week
- Open your homepage and look for the six facts. Most people find one or two. That is normal, not a failure.
- Write the missing ones into plain page text. Not an image, not a PDF, not behind a tab. AI reads text.
- Be specific. “Over 20 years” is weaker than “since 2004”. “Great reviews” is weaker than “312 Google reviews”.
- Repeat them on your main service pages, not just the homepage. AI often lands deeper than your front door.
- Fix your business schema so it declares what you actually are.
- Check an AI crawler can fetch the page at all. One site in eight cannot be fetched by AI even though it loads fine for you. See our crawler blocking study.
Only publish claims that are true and current. An AI will repeat a stale licence number or an old review count as confidently as a fresh one, and you own what it repeats.
If you want the six-fact check run across your site and your competitors, that is where our answer engine optimisation work starts.
Common questions
What information does AI need to recommend my business?
Specific, checkable facts in plain page text. The six that come up most are years in business, licence or accreditation number, service area, a price signal, response time and review count. Only 3% of Australian homepages publish all six.
Why does AI recommend my competitor instead of me?
Usually because their page gives the AI something to quote and yours does not. Assistants explain their picks, and the explanation is almost always lifted from the business’s own website.
Does putting prices on my website help with AI search?
It gives the AI something concrete to repeat, and only 39% of Australian homepages have any price signal. A starting price or a range is enough. It does not have to be a full price list.
Is schema markup enough on its own?
No. 81% of the homepages we checked already have structured data and most of it is generic. Schema helps a machine understand the page. The facts in the visible text are what gets quoted.
Where should these facts go on the page?
In readable text on the homepage and on your main service pages. Not in an image, a PDF, a slider or a tab that only loads when clicked. If it needs a click to appear, assume it will be missed.
How long does this take to fix?
The writing is usually an afternoon. Getting it published depends on who controls your website. It is one of the few things in AI search that is genuinely quick.
Will this help my normal Google rankings too?
Indirectly. Specific service and location detail tends to help ordinary search as well, but this work is aimed at being quoted rather than being ranked. Treat any ranking improvement as a bonus.
Method: 500 Australian business homepages, drawn from 589 domains that appeared in live commercial searches, were fetched in August 2026 and their visible text checked for six categories of fact using pattern matching. Structured data was read from JSON-LD on the same pages. Pattern matching will miss facts written in unusual ways, so treat these as conservative floors rather than exact counts.
