It’s twenty to ten on a Tuesday night, and someone who needs exactly what you sell is on the couch with their phone. A year ago they would have typed a few words into Google, scrolled past the ads, opened four tabs and made up their own mind. Tonight they open ChatGPT and type the question the way they’d ask a friend.
“Who’s a good bookkeeper in Parramatta who actually understands hospitality businesses?”
The answer takes a few seconds. It names three businesses, gives a sentence on why each one fits, and suggests getting in touch. If you’re one of the three, a stranger has just been referred to you by a machine. If you’re not, nothing happens at all. There’s no page two to scroll to, and you can’t buy your way into the answer itself. Nobody ever tells you that you weren’t in the room.
How often is that happening to your business right now? Would you even know? And if the machine did mention you, would it describe the business you actually run, or the one you ran three years and two phone numbers ago?
Those questions sat underneath an episode I recorded recently with Alice Piper for REA Group‘s Advantage AI Power Up series. The brief was real estate agents, and how they stay findable when buyers and sellers start asking AI for a recommendation instead of searching for one. But almost everything we covered applies to every small business owner I know. It also applies to every marketing and communications person with “SEO” somewhere in their job description who has quietly started wondering what that means now.
The alphabet soup, briefly
I have a computer science degree AND have completed advanced study at MIT Sloan School of Management. But you don’t need to be techy to understand the acronyms, because they are simpler than the people selling them would like you to believe.
Experts will happily argue about where one ends and the other begins. You can leave them to it, because the practical work is the same.
What has changed is the shape of the result. Search used to hand you ten links and let you choose. Now it increasingly hands you a recommendation, and a recommendation only has room for a handful of names.
This is already an Australian story, not an American one we’re waiting to catch up with. Roy Morgan estimates 13.6 million Australians aged 14 and over use AI tools in an average four weeks. Telsyte found that 12 per cent of Australians now say AI tools are their main way of finding information online, up from 5 per cent a year earlier. In October last year, the ABC reported that more than a third of Google searches here already came back with an AI-generated summary above the links. Your customers haven’t stopped searching, but a growing share of them are reading an answer instead of choosing from a list.
The engines don’t agree with each other
The research that stopped me in my tracks while preparing for the episode came from Cited, an AI citation agency that sells this kind of work. Earlier this year they asked ChatGPT, Perplexity and Gemini 324 questions about real estate agents across six Australian cities. Fewer than one per cent of the agencies they recommended were named by all three engines. In four of the six cities, including Sydney and Melbourne, not a single agency made all three lists. Being visible on one engine told you very little about the other two.
That sounds like chaos, and partly it is. But the same study had a finding that should cheer up every small business owner reading this. Between 85 and 94 per cent of the recommendations went to boutique or independent agencies. In Cited’s data, the engines favoured agencies that clearly stated who they served and where, and generic “full-service” descriptions earned no mentions at all. It’s one experiment, but it points somewhere encouraging: being specific mattered more than being big.
What a machine can actually quote
REA’s own research found that sellers choose an agent on five things: local market knowledge, their individual sales track record, the agency’s sales, current listings and client references. When I looked at that list, the whole topic clicked for me, because four of those five are public, verifiable facts. AI tools don’t reward clever copy. They repeat dull, checkable detail they can be confident about.
Your version of that list will be different, but it will look similar. What exactly do you do, where, and for whom? How long have you been doing it? What are you qualified, accredited or licensed to do? What specific results can you point to? If those answers aren’t written down somewhere public, in plain words, the machine has nothing to repeat about you except what other people happen to have said.
Reviews work the same way. “Great service, five stars” gives an AI tool nothing to work with. Compare that with “Sorted out three years of messy BAS for our café in Harris Park in a fortnight.” That one sentence gives the suburb, the problem and the result, which is far more useful to anything trying to describe what you do.
Then comes the part almost everybody skips. Your business name, address and phone number need to be identical everywhere they appear. That means your website, your Google Business Profile, LinkedIn, the industry directories, and whichever platform your customers already trust. For agents, that platform is realestate.com.au. For you it might be a trade directory, an association listing or a booking site. The description of what you do doesn’t need to be word-for-word the same in each place, but it needs to tell the same story. In my REA piece, I said that most agents haven’t published too little about themselves. They’ve published five versions of themselves that don’t match. I could have been describing almost any small business in the country.
If you own “the website”
If you’re the marketing or comms person in your organization, this shift quietly changes what your job is. For twenty years, success was measured in rankings and clicks. The question now is whether your organization gets named when someone asks a machine about the thing you do, and whether it gets described accurately. That is much less about keywords and much more about governance. You need one agreed source of truth for the facts, and a habit of answering real customer questions plainly on your own site. You also need someone with the authority to make sure those facts match across every channel, including the ones another team controls.
This is where it stops being a technology story and becomes a people story, which is where I spend most of my working life. Who in your organization actually owns this? Is it marketing, the agency, IT, or the person on reception who knows the opening hours on Google are wrong and has never been asked? Who has permission to fix the description on a directory listing that someone set up in 2019? If the honest answer is “nobody in particular”, that’s the first thing to fix, and it costs nothing.
Before you spend a dollar
There are a lot of cowboys selling AEO right now, so be careful who you pay and what you pay for. Nobody can guarantee you a citation. Someone may offer to add a special file to your website so that AI can find you, but Google’s own guidance says you don’t need one to appear in its AI search results. A genuine page for each area you actually serve helps. Hundreds of AI-generated ones that contradict your real details will make you look less reliable, not more. And ask any vendor one question before you sign: which of these numbers were measured in Australia, and can you show me the method?
It’s also worth remembering that handing your marketing to an AI tool doesn’t hand over the responsibility. The ACCC is clear that misleading a consumer is illegal whether it was deliberate or not, so “the tool wrote it” won’t get you very far.
Why this is really about experience
Look back at everything on that list. Publish what you actually do, keep it consistent, answer real questions plainly, and collect reviews with something in them. None of it is gaming a machine. Every item is about making the experience you already deliver legible to whoever, or whatever, is asking. Businesses that have built a genuinely good experience now have a way for it to be repeated at scale.
Your four minutes this week
Here’s the test I gave agents, and it works for any business. Log out of ChatGPT first, so you see something closer to what a stranger sees, and ask it the question your best customer would ask. Then ask Perplexity, then Gemini. Try three or four questions your customers would really ask, note the date, and look at which sources each answer leans on. You’re taking a snapshot, not a final verdict, because the answers will shift. Write down who gets named, and whether you’re described the way you’d describe yourself.
After that, the first month of work costs nothing but your time. Claim and complete your own Google Business Profile. Make your name, address and phone number identical everywhere, and make sure your description tells the same story wherever it appears. Then ask ten recent customers for a review that names where you are and the problem you solved.
The full episode with Alice, my practical article for REA and the Findability Audit prompt I built for agents are on the REA page (I’ll link to it in the comments).
So I’m curious. When you asked the engines about your business, who did they name, and how close was their description to the business you’d describe?