Who can check if ChatGPT recommends your business?

You can check it yourself for free, pay for software that tracks AI answers over time, or buy a one-off report, such as our A$29 AI Recommendation Report. When we put this exact question to AI assistants, 10 answers named 39 different providers, none of them us, so asking ChatGPT won't settle it. What decides whether a check is worth anything is what it measures: several assistants, your suburb, enough questions, and who gets named instead of you.

We asked AI this question ourselves

Among the eight questions we track about our own work, two are this one: "Who can check whether ChatGPT and other AI assistants recommend my business in Australia?" and "Is there an AI visibility report for local businesses in Australia, and what does it cost?" On 7 October 2026 we asked both in ChatGPT, Gemini, Perplexity and Claude. On 2 November we asked Claude again.

Those 10 answers named 39 different businesses and products. The one named most came up in 4 of the 10. The names were a mix of tracking software, marketing agencies and one-off checkers, and in November Claude didn't repeat a single name it had given for the same two questions in October. None of the 10 named sbw studio or cited our site.

Ten answers is a small sample, so read this as a sign, not a ranking. It does say something useful: the shortlist you get depends on which assistant you ask and when. That is true of the answer about checkers, and it is just as true of the answers your customers get about you.

What a check has to measure to mean anything

Our study of 12 local businesses, 733 answers plus 46 answers checking their details, shows why a quick look misleads.

• One assistant isn't enough. Compare any two assistants' five most-named businesses in an area and they typically share three.
• One question isn't enough. The typical business we tested was named in 23% of answers about its area and trade, so a single try will usually say you aren't there.
• The location has to be yours. A customer asks from somewhere, so we set each assistant's location to the business's suburb. A check with no location asks about nowhere in particular.
• The count matters less than the names around it. Who is named instead of you, and which pages the answer cites, tells you where to start.
• Your details need checking too. When the assistants stated a fact about a business they were right 94% of the time, but for 5 of the 12 at least one assistant gave the wrong opening hours.

Any check worth paying for, including a free one you run yourself, should cover most of that list.

The three ways to get it checked

Do it yourself

This costs nothing but your time. Eight questions in four assistants makes 32 answers, and you see what a customer sees. Ask each one in a new chat with memory off, and name your suburb in the question so the answer is about the right place. Our steps are in does your business show up in AI search?

Tracking software

This is built to ask the same questions again and again and chart the result, which suits the problem, since the names move between checks. Before you sign up, ask whether it can set the location to your suburb and whether it checks your details as well as your name.

A one-off report

This gives you a snapshot and someone to read it for you. Ours asks ChatGPT, Gemini, Perplexity and Claude ten questions locals ask, 64 times in all, with the location set to your suburb. It checks your address, phone, hours and prices in every assistant, lists your first fixes and tests again for free after 60 days. It has limits we state on the page: we ask through code rather than the apps, Gemini also gets Google's Maps tool, and your customers' answers can differ.

Questions to ask before you pay anyone

1. Which assistants does it ask? Four is better than one, because they disagree.
2. Is the location set to your suburb, or to Australia, or to nowhere?
3. How many answers is the result based on? A handful can swing on chance.
4. Does it show who is named instead of you, and the pages cited?
5. Does it check your hours, phone and address as well as your name?
6. Will it check again later on the same questions, so the two results compare?
7. Can you see the questions? They should be ones your customers ask, not ones built to flatter you.

What we are doing about it on sbw.studio

The assistants had no page from us to cite when they answered this question, so this article is that page. It is the same move we suggest to clients: when people ask a question and the answers skip you, publish a plain answer to it.

We don't know yet whether it will work. We ask our eight questions again on the first Monday of December, with Claude alone as in November, and the result goes in our monthly growth log, zero or not. The October log has the 0 of 8 we start from.

How we measured

The 10 answers come from our own AI check: the question above asked of four assistants with web search on, on 7 October 2026, and of Claude alone on 2 November. The 23% and the shared three of five come from our study of 733 answers about 12 local businesses in 11 Australian areas on 5 October 2026, and the 94% and the hours from 46 further answers about the same businesses' details. That is also a small sample. Full findings and method: score.sbw.studio/findings.

If you'd rather have it counted for you, get the AI Recommendation Report: four assistants, your suburb, who gets named instead of you and what to fix first.