Readable, not reputable

There is a great deal of confident advice circulating about how to get your business recommended by ChatGPT, Perplexity and Google's AI. Very little of it is supported by evidence, and a surprising amount of it is supported by figures that do not trace back to any real study at all.

We went looking for the evidence. Every statistic we intended to use was checked against its original source, and roughly a third were discarded, either because no primary source existed at all or because the figure had been superseded and nobody had noticed. What follows is what survived that check.

This is not a report that will tell you AI search is about to replace everything you do. The honest picture is more interesting than that, and considerably more useful. Referral volume from AI platforms is still small. Influence is not. The gap between those two facts is where the opportunity sits.

Seventeen point four million Australians aged sixteen and over now use AI. That is seventy seven per cent of the adult population, and daily use has grown by one hundred and sixty per cent in a single year. This is not an emerging behaviour. It is the current one.

The number that matters most for your business is smaller and sharper. Twelve per cent of Australians now name an AI tool as their primary way of finding information online. Not one option among several. The first place they go.

What happens next is the part most businesses have not absorbed. When an AI summary appears at the top of a search, people stop clicking. Pew Research Center tracked real browsing behaviour rather than asking people what they thought they did, and found that users clicked a traditional search result on eight per cent of visits where an AI summary was present, against fifteen per cent where it was not. Roughly half the clicks, gone. Only one per cent clicked a link inside the summary itself.

The bigger picture

Put those together and you get the defining feature of this shift. Your business can be read, summarised and described to a potential client without that person ever visiting your website. You may be recommended, or passed over, in a conversation you never see and cannot measure in your analytics.

That is a genuine change in how reputation travels. It is also why measuring AI visibility by looking at your referral traffic will tell you almost nothing.

This is the finding that reframes everything, and it comes from two independent datasets that agree.

Ahrefs analysed fifteen thousand queries and found that only twelve per cent of the URLs cited by ChatGPT, Gemini and Copilot appeared in Google's top ten for the same prompt. Around eighty per cent of citations came from pages ranking outside Google's top one hundred entirely. BrightEdge, working from a completely different keyword panel, found that roughly seventeen per cent of Google AI Overview citations came from pages ranking in the organic top ten. Two methods, two datasets, the same conclusion.

So what does correlate

Ahrefs measured seventy five thousand brands against their visibility in AI answers. The pattern is consistent and it is not the one most marketing plans are built around.

Being described elsewhere on the internet correlates far more strongly with AI visibility than any link based measure. Backlinks, the currency of twenty years of search optimisation, sit close to negligible.

For a professional services firm this is unusually good news, because being described elsewhere is something you can genuinely influence. Podcasts, industry publications, association directories, partner sites, conference listings, credible profiles with your name attached. That is the work.

This is the part of AI visibility that almost nobody discusses honestly, and it is the reason most AI ranking tools are selling you false precision.

Research published in June 2026 measured how much the citation sets overlap when the identical query is run repeatedly on the same platform. On Gemini, the median overlap sat between zero point two nine and zero point three one. In plain terms, two runs of the same question share under a third of their cited sources. On OpenAI's search product it ran between zero point three three and zero point four. Perplexity, the most stable of the three, managed zero point five.

The instability is not confined to individual queries. Semrush tracked citation sources across two hundred and thirty thousand prompts over thirteen weeks and watched Reddit's share of ChatGPT citations fall from close to sixty per cent to around ten per cent in roughly six weeks. Wikipedia fell from fifty five per cent to under twenty in the same window.

Two independent bodies of evidence, one academic in style and one commercial in scale, both showing severe instability. Any tool that hands you a single AI visibility score from a single test is reporting noise with a decimal point on it.

What we measure instead of a single number

Because citation presence is volatile, we do not lead with it. We build the picture from four layers, most stable first, so that a quiet quarter on the volatile layer never obscures the work that is genuinely moving.

Layers one and two are almost entirely within our control and carry no noise. Layer three is the one everyone wants and the one that moves unpredictably. Layer four is the only one that pays the bills, and it is the hardest to attribute, which is precisely why we ask every new client whether they used an AI tool while researching us.

Every approach to AI visibility involves a trade-off. Being clear about them is the difference between a strategy and a wish.

The second question has an answer you can actually work towards, and the work holds its value regardless of which platform is ascendant next year. The first question invites you to chase a system that changes weekly.

This is not a visibility test. It cannot tell you whether AI mentions you, and any tool claiming to do that from six questions is not being straight with you.

What it does measure is structural readiness: whether the signals AI systems rely on are present, consistent and attributable. Readiness is the part you control, and it is the part that has to be right before any measurement is worth taking.

Answer honestly. Nothing is stored or sent anywhere.

There is no single sequence that suits every business, and a generic checklist would be the same unmeasured advice this report set out to correct. Choose the description that fits you now.