Would ChatGPT recommend your business?

AI visibility is whether ChatGPT, Gemini or Google’s AI Overviews would name your business when a serious buyer asks who to hire.

I am Adam Stewart. I work with founder-led Australian businesses through Digital Bond, and this is the commercial test I want you to run this week. Not another definition. A simpler question. If a buyer asked a chat window for help, would you be in the answer?

People are not only asking ChatGPT to rewrite emails anymore. They are asking it what to buy, where to go, and which business to choose.

If that sounds early, it is not. Shopify told investors that AI traffic to merchants is up 3x year on year, orders from AI searches are up 3x, and new-buyer order rates from that traffic are twice other channels. I wrote that up in August 2026. It is worth paying attention to now rather than later.

Why is this a sales channel, not a parlour trick?

For a long time, “being online” meant a website, Google, maybe ads, maybe a marketplace.

The new behaviour sits beside those, not instead of them.

A shopper asks Gemini for a stroller that fits in a small Melbourne apartment. A facilities manager asks ChatGPT for a commercial cleaner who can do medical clinics on the Gold Coast. A founder asks for a payroll tool that handles Australian super without a six-month implementation.

They get a shortlist. They click one or two. Sometimes they buy from the chat. Sometimes they walk into Google already half decided.

That last part matters for local businesses as much as ecommerce. Google Maps is getting better at taking the whole request (“dinner near the hotel, quiet, we have a nut allergy”) instead of five separate searches. If your Google Business Profile is thin, late, or vague, you are harder to recommend.

Some people call this ChatGPT SEO. Some say AI SEO. I do not love the fathead version, because it sounds like a plugin. It is closer to “can a careful stranger describe us, and would they send us the job?”

If you want the plain-English definition of the wider shift, I already wrote what is answer engine optimisation. This piece is the buying-question test.

What are the Shopify numbers actually saying?

I am not going to invent a bigger story than the one Shopify published.

AI traffic to merchants, 3x year on year.

Orders originating from AI searches, 3x.

New-buyer order rates from AI traffic, twice other channels.

Read that last line again. The people arriving from AI are not just browsing. They are more likely to be first-time buyers, and they are converting harder than the average channel.

That is a recommendation effect. Someone already did the shortlisting.

If you sell online, this is a reason to look at product pages the way you used to look at Google Shopping titles. Specific. Comparable. Honest about who the product is for.

If you sell locally, the equivalent is your profile, your reviews, your service-area pages, and the third-party mentions a model can find.

Diversify that visibility. Own site. Reviews. Press. Directories. Google Business Profile. One accurate homepage is not enough if every other public description of you is a ghost listing from 2019.

How do you run the customer-question test?

Do not ask “is my website good?”

It will flatter you.

Ask the buying question.

Try this

I need [product or service] for [situation] in [Australian city or region]. Budget is [range, GST inclusive if relevant].

Recommend 5 real options. For each, say:
- who they are best for
- what evidence you found
- what you could not verify
- why someone might choose a competitor instead

If you do not have enough information to recommend a business, say so. Do not invent names.

Run it in ChatGPT. Run it in Gemini. Try a second phrasing, the way a tired customer would type it.

Then do the local version if you have a physical presence.

I am in [suburb]. Who would you send me to for [job], and what should I check on their Google profile before I call?

Three outcomes are useful.

You appear, and the reason is right. Good. Protect that clarity. Do not let the next website redesign turn it into slogans.

You appear, and the reason is wrong. You have a message problem. The model found you and still misunderstood the offer.

You do not appear. Look at who did. Copy the structure of their public facts, not their adjectives.

What should you fix when you are invisible?

I teach the same public playbook I use when I check this for a business.

Write one sentence a model can repeat: what you sell, who it is for, where you do it, what a good outcome looks like.

Put that sentence on the homepage, the key service or collection page, and your Google Business Profile.

Add proof a stranger can see. Reviews that mention the actual job. A page that names a problem and a result. A FAQ in the customer’s words.

Get listed, accurately, in the two or three directories your industry still uses. Kill or claim the old listings that contradict you.

Make sure the site can be read. If the page is blocked, no amount of “ChatGPT SEO” copy will save it.

Do this for one query family first. The question that sits right before money moves. Then expand.

I have seen a plain services page outperform a prettier one because it named suburbs, prices, and who should not buy.

Local and ecommerce are different surfaces

A cafe and a Shopify store can fail the same test for different reasons.

The cafe fails because the profile is missing hours, the menu is a photo, and the last review that mentioned a dish is fourteen months old. Maps cannot recommend what it cannot parse.

The store fails because the product title is internal jargon, the description is lifestyle fog, and there is no comparison a model can use. “Best for cold sleepers who hate bulky doonas” is a recommendation hook. “Premium sleep system” is not.

A trades business fails when every page says “Melbourne and surrounds” and none of them say the actual councils or job types.

A B2B firm fails when the only public description is “we partner with ambitious teams to unlock growth.” Nobody gets recommended for that sentence. Nobody should.

Same discipline. Different fields.

What not to do with this

Do not publish forty AI blogs hoping a model will cite them.

Do not stuff “as recommended by ChatGPT” on your homepage. You will look like you are lying, and you might be.

Do not treat this as a reason to abandon Google. People still search. They still click. The AI Overview sits on top of that behaviour. It does not delete it.

Do not hand an AI the right to speak as you in public. Use the tools to find the gap. You decide what the internet should say about you.

For most businesses the recurring task is a monthly recommendation test plus one page update. Not a six-month “AI transformation.”

If the recommendation test exposes a wider AI implementation problem, the AI consultant Australia page explains how Digital Bond turns a useful AI idea into a governed workflow.

My honest read

My honest read: if ChatGPT cannot name you for the job you actually want, you have an offer-clarity and proof problem, not a missing tool. Do not spend a dollar on “getting into AI search” until the customer-question test produces an accurate sentence about you.

Treat AI as a recommendation channel you can inspect, not a slot you can buy. Measure whether you get named, and whether the reason is true.

FAQ

How do I check my AI visibility in ChatGPT?

Ask the question a buyer would ask, not “tell me about my website”. Name the city, the job, the budget and the GST posture. Run the same prompt in Gemini. Write down who got named, why, and what the model could not verify. That list is your brief for the next page or profile update.

Does this replace Google SEO?

No. People still search and still click. AI Overviews sit on top of that behaviour. Rankings still matter. The extra job is being selectable when the customer never sees ten blue links. Keep the Google Business Profile accurate either way, because Maps and AI answers both read it.

What if I appear but the description is wrong?

That is a message problem, not a visibility problem. The model found you and still guessed. Make the offer sentence identical on the homepage, the service page and your Google Business Profile. Kill old directory listings that contradict you. Then run the test again in a fortnight.

What to do this week

Run the prompt today. Two models. One buying question.

Write down who got named and why.

Fix one public page, one profile field, or one review request so the missing fact exists by Friday.

If you would rather build the workflow with other founders, AI Systems Lab is the implementation room. The scorecard can wait. Run the test first.

The Digital Bond scorecard gives you the wider acquisition, conversion, follow-up and measurement context around that visibility gap.

I still want you to rank.

I also want you to get chosen when the customer never sees the ranking.