Two independent Australian studies from the same week show AI now helps three in four purchase decisions, with almost all of that influence landing upstream before any click your analytics can record. The panic about getting cited in AI misreads the problem. This is not a visibility issue, it is a measurement one, and the fix is to stop scoring your marketing on the last click.
The AI is not the last step before the sale. It is the first step before the shortlist, and your reporting starts counting one stage too late.
The Take: Three in four Australians now use AI to help decide what to buy, and almost all of that pull happens before anyone clicks a link you can track. This is not a visibility problem, whatever the panic about getting cited tells you. It is a measurement problem, and the businesses that win the next few years are the ones who stop scoring themselves on the last click.
Is being cited in AI the game, or is it something else?
No. 94% of AI use in a purchase decision happens right at the start, before any click or referrer your analytics can record. The choice forms in a room your measurement never enters.
Here is the week that should have changed how Australian marketers think about their reporting. Two independent studies landed within days of each other, and read together they draw a line straight through the middle of how this market spends its budget.
The first is WPP Media's Hidden Pathway study, 2,500 nationally representative Australian consumers across 12 categories. It found three in four Australians already use AI Search to help make purchase decisions. Not to buy. To decide.
Then the texture. 58% use the AI summaries sitting inside normal search results while 40% go straight to tools like ChatGPT, Gemini or Claude. About a third use both. Buyers run an average of 18 AI actions per purchase decision. Here is the part that matters: 94% of that AI use happens toward the beginning of the journey.
By the time a person hits your site, clicks your ad or opens a review page, the AI has already done its work. It framed the category. It named the options. It ruled some brands out before you knew you were in the running. 84% of Australians say AI surfaced a product or service they had not planned to buy, and 26% completely changed an intended purchase after an AI interaction.
Daniel Benton of WPP Media put the measurement gap plainly: "AI's influence begins much earlier in the journey than most measurement systems can currently see." Read that as an admission, not a headline. The people who ran the study are telling you the influence is real and unmeasured.
Australians who say AI surfaced a product or service they had not planned to buy
The decision moved upstream, your measurement did not
Australia's advertising market is forecast to reach A$31.1 billion in 2026, up 7.4%, led by retail media, search and social. Most of that money is optimised on last-click signals: the ad that got the tap, the keyword that got the session, the page that got the form. Across the Australian businesses we assess, a large share still run weak or broken analytics and credit almost everything to the final touch. So the national picture and what we see up close say the same thing. A market pouring billions into the last click, against a decision being formed at a stage where no click exists.
That is the false floor under most reporting. It is not that the numbers are wrong. It is that they only start counting after the important part is over.
Now hold this next set of numbers next to the panic, because they change the instruction completely.
Should you panic about buyers handing the sale to a robot?
No, and this is where most AI commentary loses the plot. The second study, from Ideally, published 14 September 2026, found just 3% of Australians would buy a product on an AI recommendation alone. People are using AI to think, not to check out.
Watch what they do when an assistant names a brand they do not know. 52% would search Google before deciding. Another 21% would check review sites first. Asked which source they trust most for a major purchase, the ranking is telling.
Source: Ideally, September 2026.
So the AI opens the door, then the buyer verifies everywhere else. That is the whole game in one sentence. AI shaped the shortlist, and 82% of people are more likely to trust a brand that appears in AI results, so your job is not to win a checkout inside ChatGPT. Your job is to be the brand that survives the verification that follows: strong Google presence, real reviews and a reason for a mate to say your name.
Stop trying to sell inside the AI. Start earning the trust the buyer goes looking for the moment they leave it.
What I would do about it
Four things, in order.
First, stop grading yourself on the last click. Split your reporting into two questions: are we being considered, and are we converting. Most businesses can answer the second and have no instrument for the first. That gap is the work.
Second, fix your measurement floor before you spend a cent chasing AI visibility. If your analytics is broken, your tagging is half-wired or every conversion is credited to the final touch, you cannot see upstream movement even when it happens. Get the basics honest first.
Third, build the assets that survive verification. Reviews you actually ask for. Comparison content that answers the real question a buyer types into an AI. A brand presence that shows up when the 52% who go back to Google check the name they just heard.
Fourth, treat word of mouth as a channel with a budget, not a happy accident. Friends and family sit at 25% for a major purchase, a whisker behind Google and streets ahead of any AI assistant. Referrals, partnerships and reputation are upstream trust you can build on purpose.
A note on how we measure this
Our read on the Australian market here is directional, not a dataset count. The New Rebellion benchmark work is mid-rebuild, so I am describing patterns we see across the businesses we score rather than a specific figure. Where our sample is thin, I say so. If you want the actual methodology, dimension by dimension, it is public at how we score. The external numbers in this piece are from WPP Media and Ideally, both from fieldwork completed in the second half of 2026.
Where I land on this
This is my opinion, held hard. The businesses that struggle over the next two years will not be the ones who failed to get cited by an AI. They will be the ones who kept optimising the last click while the decision quietly moved to a stage their reports never reached. Visibility is a symptom. Measurement is the disease, and the cure is unglamorous: see the whole journey, then earn trust at the top of it.
That is the work we do inside NR Studio.
Frequently asked questions
Does AI search mean Google is finished for my marketing?
No. Google search is still the most trusted source for a major purchase at 26%, and 52% of people go back to Google to verify a brand an AI names. AI shapes the shortlist earlier, Google still closes the trust gap. You need both.
If only 3% buy on an AI recommendation, why should I care about AI at all?
Because the influence is upstream, not at checkout. 94% of AI use in a purchase decision happens at the start of the journey, shaping which brands even make the consideration set. Low purchase share, high framing power.
How do I measure something that happens before the click?
You cannot attribute it perfectly yet, and anyone promising a clean number is selling you something. What you can track are consideration signals: branded search volume, direct traffic, review growth and share of shortlist in your category. Measure the movement, not the final touch.
What is the single biggest mistake businesses are making right now?
Pouring effort into getting cited in AI while their core measurement still credits everything to the final click. Fix the measurement floor first. You cannot manage an upstream stage you cannot see.