A homeowner in Weston Creek opens ChatGPT and types "who's a good electrician near me". The assistant doesn't pause. It names two or three local sparkies, in a tidy sentence, with a reason for each. The homeowner picks one and gets on with their day. The trades who weren't named never find out the question was asked. The answer was given in a room they weren't standing in.
That room is the new front door. People used to scroll a page of links and make up their own mind. More of them now ask an assistant to make up its mind first. The assistant doesn't keep a secret list of businesses. When it's asked for an electrician in a suburb, it builds the answer on the spot from whatever it can read about the electricians around: their reviews, whether their details line up across the web, whether their website plainly says what they do and where. If what it finds about you is thin or missing, you're simply not in the answer.
You can't manage what you won't measure
Here's the trap. AI search is invisible by nature. You can't see yourself losing, the way you'd see a competitor's van in a driveway. So the first thing worth doing isn't optimisation. It's measurement.
The method is dull on purpose. A fixed set of questions, the same ones every time, asked across ChatGPT, Gemini and Perplexity. The plain "who's a good [trade] in [suburb]" question. A "recommend a [trade] near [suburb]" version. And a branded one: "what do you know about [business]". You screenshot every answer on day one. That's the baseline. Now there's a before. Now the work has something to be judged against.
What actually moves it
The fixes are not exotic. This is good local SEO made machine-readable: a new front door, the same lock.
- Plain sentences. Your site says, in words a person would say out loud, what you do and the suburbs you want the work from. Assistants quote plain language, not keyword soup.
- Structured data. The same facts again in the schema the machines read directly, so there's no guessing.
- Consistency. Your name, number and service area match across the website, the Google profile and every directory. Mismatched details read as a thin, uncertain business.
- Reviews. Genuine, recent, specific. This is the biggest lever, and the one no shortcut survives.
- An open door. The answer crawlers are allowed in rather than blocked by reflex, because an assistant can only quote what it's able to read.
None of it is wasted even if you're sceptical about the AI part, because the same work lifts you in ordinary Google too.
Why we won't promise you a result
Nobody controls what a model says. Anyone who guarantees ChatGPT will name your business is selling you a feeling, not a method. Be wary of them.
What can be done honestly is to optimise the signals, not the model of the month. And it tends to land in two stages: first you get described accurately when someone asks about you by name, then, later, you start getting recommended when nobody asked for you specifically. The first is a foundation. The second is the prize.
The re-test is the proof
Ninety days after the work, the same questions get asked again. The exact same battery. The change between the day-one screenshots and the ninety-day ones is the only claim worth making, because it's the only one you can see with your own eyes.
That's the whole promise: not that you'll be named, but that you'll be readable, and that we'll show you what moved. Be the name the search returns, because the machine could finally read you, not because anyone swore it would.