Ask ChatGPT About Your Own Business. The Answer Should Worry You

Try this before you read any further. Open ChatGPT and ask it to recommend three businesses like yours in your suburb. Then ask it what your company does, who it's for, and what it costs.

In our experience running this exercise with Australian service businesses, the result lands in one of four buckets: the assistant has never heard of you, it describes a version of you that's two or three years out of date, it confidently blends you with a similarly named competitor, or it recommends three rivals and never mentions you at all.

None of those are search problems in the way most owners think about search. There's no ranking to check, no position to climb. There's just an answer, and either you're in it or you aren't.

Why the answers go wrong

AI assistants don't hold a tidy database of Australian businesses. They assemble an answer at query time from whatever they can retrieve and whatever the model already absorbed during training. That means your reputation inside an LLM is built from sources you probably haven't looked at in years.

• Directory listings — Yellow Pages, Hotfrog, industry associations — that still carry an old address, an old phone number or a service you dropped.
• Your own website copy, which for many businesses hasn't been rewritten since the services and audience changed.
• Third-party roundups and 'best in Melbourne' listicles, which the model treats as social proof whether they're accurate or paid placement.
• Review platforms, where a handful of specific complaints can end up characterising the whole business in a summary answer.
• Nothing at all — in which case the model fills the gap with the nearest plausible business it does know.

Run the audit properly

A five-minute poke around isn't an audit. Do it systematically and you'll get something you can act on.

Write down the ten questions a genuine prospect would ask — not the ones you'd like to rank for. 'Who does commercial electrical work in Dandenong?' 'Is X Physio any good?' 'How much does a small business website cost in Australia?' Then run each one through ChatGPT, Gemini, Perplexity and Google AI Overviews, in a fresh session with no personalised history, and record two things for each: whether you appeared, and which sources were cited.

The citation list is the important part. It tells you exactly which pages the machine trusts on your topic. If your competitor's blog post is cited five times and your service page appears nowhere, you now know what the gap actually is — and it's rarely 'more keywords'.

What actually changes the answer

Once you can see the sources, the fixes stop being abstract. In order of how much they move the needle for a typical service business:

• Make the facts consistent everywhere. Name, address, phone, hours, service list and service area should match exactly across your site, Google Business Profile and every directory you can still log into. Contradictions make a model hedge, and a hedging model recommends someone else.
• State plainly, in text, what you do and who for. Models can't infer positioning from a hero image and a three-word tagline. One clear paragraph near the top of each service page does more than a whole page of atmosphere.
• Publish the specifics only you have — real project outcomes, real constraints, real numbers, real local knowledge. Generic advice is filtered out as low-signal because a hundred other pages already say it.
• Add structured data that mirrors your visible content, so the retrieval layer gets the facts in machine-readable form rather than guessing from prose.
• Answer the actual questions in your own words. Pages with clear headings and direct answers get extracted far more often than pages that build to a point over eight paragraphs.

Treat it as a monthly check, not a project

AI answers move. A model update, a new citation, one competitor publishing a genuinely useful page — any of these can change what a prospect is told about you next week. The businesses handling this well keep a short standing list of prompts and re-run them monthly, the same way they'd glance at rankings or Google Business Profile insights.

It takes fifteen minutes. It's the cheapest early warning system available for a channel that is quietly becoming the first thing a buyer reads about you.

The honest caveat

Nobody can guarantee placement in an AI answer, and any agency promising it is selling you something it can't control. There's no submission form, no bid, no dashboard. What you can control is whether the machine has clean, distinctive, verifiable material to work with — and whether that material is better than your competitor's.

That's not a new game. It's the same game with a different reader.