Every business owner asks whether they show up in AI search. The costlier question is whether the machine is describing them accurately. The IAB's own head of AI says factual inaccuracy is the real risk, and almost no Australian business checks what ChatGPT or Gemini is telling buyers about it.
A confident wrong answer is worse than silence. Silence costs you a lead. A wrong answer costs you a lead and hands your reputation to the buyer in the wrong shape.
The Take: Every business owner I talk to about AI search asks the same question. Am I showing up? It is the wrong question. The one that will cost you money is a different one. What is the machine actually saying about me, and is it true.
Should you worry about being unseen by AI, or being described wrongly?
The bigger risk is being described wrongly. In Australia 12% of people now use AI tools as their main way to find information, and almost none of the businesses they ask about check the answer.
The IAB's own head of AI says factual inaccuracies about brands are the thing that keeps her up at night, not whether a brand appears at all. Presence is the metric most people chase because it is the easiest one to count. It feels like the old game of ranking on a search page. Show up, win. That instinct is out of date. An AI answer does not just list you. It characterises you, and a confident wrong sentence about your pricing or your service area does more damage than not appearing would.
What did the IAB actually publish?
On Monday the IAB released Measuring Visibility in the AI Era, guidance for how brands should track their AI search analytics. It proposes a hierarchy of four metrics it calls the 4 P's: presence, prominence, portrayal, persuasion. Most of the chatter this week has been about the first P. The one that matters is the third.
Portrayal, as the IAB defines it, is not whether you show up but how. It splits into sentiment and accuracy, and accuracy splits again into hallucinations and factual inaccuracies. Hallucination rates have declined as the leading models improved, which is the good news. The bad news is the quieter failure. An AI draws on outdated or misrepresentative data about you and states it with total confidence.
Caroline Giegerich, the IAB's VP of AI, was asked what marketers should worry about most. Her answer was factual inaccuracy, the thing that "would keep me up at night". That is a trade body telling you where to point your attention, and it is not at your mention count.
Why does this hit Australian businesses harder?
Because the buyers are already there and the businesses are not watching. In Australia 12% of people now name AI tools as their primary way to find information online, up from 5% a year earlier. Telsyte's 2026 figures put ChatGPT at 13.8 million users in Australia, ahead of Google Gemini at 9.1 million.
Australians now using ChatGPT, any one of whom can be handed a wrong description of your business
Sit with that number for a second. That is the size of the room in which a machine is describing your business to people who have never visited your website. We score Australian companies across six marketing dimensions and benchmark them against a large panel of local businesses, and brand clarity is one of the weaker dimensions across that panel. So the national number and our number point the same way. More Australians are asking machines to describe local businesses, and those businesses are, on the whole, poor at controlling how they are described anywhere.
How would you even check portrayal?
You do it by hand before you pay anyone to do it for you. Open ChatGPT and Gemini, ask the ten questions a real buyer asks before choosing someone like you, then read the answers as a customer would. Where does the price look wrong. Where is your service area misstated. Where does a competitor get described in warmer terms than you.
The IAB is blunt about how much of this you can trust. It classifies fewer than 50 queries as exploratory, a rung below even directional. It warns that a sample that thin cannot characterise a category. So a single bad answer is a prompt to investigate, not proof of a pattern. Ask the same question several times across several days before you decide the machine has a fixed and wrong opinion of you.
What I would do about it
If you run an Australian business, work in this order.
First, run the portrayal audit yourself this week. Twenty minutes with ChatGPT and Gemini and your ten buyer questions will tell you more than any dashboard demo.
Second, treat every wrong answer as a content brief. If the machine thinks you only service one suburb, or lists a price you retired two years ago, the fix is on your own website and your own listings, where the model is drawing its picture from. Correct the source and the description tends to follow.
Third, only then weigh a tool. A portrayal monitor that samples enough queries to clear the directional bar and flags sentiment and accuracy might earn its fee. A tool that sells you a vanity count of mentions will not.
Fourth, do the boring hygiene the model rewards. Accurate, current, well-structured information about your business on pages you control is what an AI reads. Neglected pages produce neglected answers.
How we score this
New Rebellion scores Australian businesses across six marketing dimensions and benchmarks each one against a large panel of local companies. Brand clarity and data are two of those six. Where the evidence for a business is thin, we mark the score as directional rather than dress a guess as certainty. That same methodology is exactly what I am asking you to bring to AI portrayal. You can read the full method at how we score.
This is my opinion, not a forecast. AI search is early and the tools will sharpen. My position is that the Australian businesses who come out ahead will not be the ones who won a mention-count leaderboard. They will be the ones who noticed, early, what the machine was telling buyers about them and fixed the parts that were wrong. If you want to see how clearly your business reads today, start with your own scorecard.
Frequently asked questions
What is AI portrayal in search?
It is how an AI tool describes your business when it answers a buyer's question, covering both sentiment and factual accuracy. It goes beyond whether you are mentioned to what is actually said about your pricing, service and reputation.
How do I check what AI says about my business?
Ask ChatGPT and Gemini the questions a real buyer would ask before choosing you, and read the answers for accuracy. Repeat each question across several days, because a single response is not a reliable pattern.
Is AI search big enough in Australia to bother with?
Yes and growing. 12% of Australians now name AI tools as their main way to find information online, double the share a year earlier. ChatGPT alone has 13.8 million Australian users.
What are the IAB's 4 P's of AI visibility?
Presence, prominence, portrayal and persuasion. Presence is whether you appear. Prominence is how high. Portrayal is whether the description is accurate and fair. Persuasion is whether it changes a buyer's decision.