Your next customer might never see your website. They will ask ChatGPT for a recommendation, get three business names back, and ring one of them. If your name is not in that list of three, you were never in the running, and nothing in your Google Analytics will tell you it happened.
This is not a prediction. BrightLocal’s 2026 Local Consumer Review Survey found that 45% of consumers have used AI tools such as ChatGPT, Gemini or Perplexity to find local business recommendations in the past year, up from 6% the year before, with 30- to 44-year-olds leading adoption at 64%. That is the demographic that books trades, changes accountants, and signs commercial contracts. So, it pays to explore how AI assistants decide who to name, how to check what they currently say about you, and what to fix first.
What does it mean to be “recommended” by an AI assistant?
Being recommended by an AI assistant means the model names your business in its answer, usually with a short reason and a link. Unlike a Google result, there’s no page two. Assistants return a shortlist of two or three names, so visibility is binary. You’re either in the answer or invisible, regardless of where you rank on Search Engine Results Pages (SERPs).
That last point is the one that catches experienced marketers out. Ranking well is no longer a reliable proxy for being cited. Ahrefs found that only 37.9% of URLs cited in AI Overviews also rank in the top 10, down from around 76% in July 2025. The overlap between “ranks well” and “gets quoted” is shrinking every quarter.
Is this actually happening in Australia yet?
The honest answer is yes for discovery, not yet for traffic volume, and the gap between those two things is where the opportunity sits.
The adoption numbers are large. Roy Morgan data from the first quarter of 2026 puts 58% of Australians aged 14 and over using AI tools at least once every four weeks, which is 13.6 million people, and Telsyte reports that 12% of Australians now name AI tools as their main way of finding information online, up from 5% a year earlier.
The referral traffic numbers are still small. Statcounter had AI chatbots at 0.39% of Australian web traffic in May 2026, with search still at 90.72%. Anyone telling you SEO is dead is selling something.
But look at the direction rather than the volume. Optimising’s Australian AI Traffic Study, covering 24.8 million sessions across 115 Australian businesses over 12 months, found median year-on-year AI traffic growth of 1,200%, with 88.5% of sites having already received some AI-referred traffic. A small number growing at that rate is a planning problem, not a curiosity.
The more useful number for a local business is the one that measures absence. A July 2026 study by Snowball Productions, titled The Australian AI Answer Gap, tested 50 everyday Australian consumer questions across ChatGPT, Perplexity, Gemini and Google AI Mode, and found that 23 of the 50 answers named no Australian business at all. Treat that as vendor research rather than peer-reviewed work. However, it matches what we see when we run audits: the shortlist is often filled by directories, national aggregators and overseas sites simply because no local business gave the model anything to work with.
The businesses missing from AI answers are rarely the worst businesses in their category. They are the ones whose information is hardest to verify.
Can ChatGPT see my Google reviews?
No. ChatGPT does not have direct access to Google reviews or Google Business Profile data, because that is Google’s own ecosystem. Reporting on ChatGPT’s local behaviour indicates it draws on Bing-indexed web results, Foursquare, Yelp, the BBB, industry directories and your own website instead. Gemini can use Google’s local data. ChatGPT cannot.
This single fact undoes the standard Australian playbook. Most local businesses here have poured a decade of effort into one platform: Google Business Profile, Google reviews, Google Maps. That work still matters enormously for Gemini, AI Overviews, and the map pack. Google Business Profile is the primary data layer feeding Gemini, Search and Maps in 2026.
But for the assistant with the largest share of the Australian market, that entire investment is second-hand information at best. ChatGPT holds 67.7% of Australian chatbot share, with Microsoft Copilot second at 14.2% as at June 2026. Copilot runs on Bing. So does much of ChatGPT’s search layer. If your Bing Places listing is empty or wrong, you have handed two-thirds of the AI market a reason to skip you.
The Answer Gap Audit: 5 prompts, 20 minutes
Before you change anything, find out what the machines already say. Open a fresh chat in each assistant, logged out where you can, so you are not seeing results shaped by your own history.
Run these five prompts, substituting your category and your suburb:
- The shortlist prompt. “Who are the best [electricians] in [Wollongong]?” Note every business named and in what order.
- The comparison prompt. “Compare [your business name] with [your two closest competitors].” This shows you what the model believes about you, which is often out of date.
- The direct prompt. “Tell me about [your business name] in [your suburb].” Check the facts it returns: hours, services, address, who runs it.
- The buying-question prompt. “How much does [your service] cost in [your region] and who should I call?” This is the query with money attached.
- The objection prompt. “Is [your business name] any good?” You want to know which reviews and third-party pages it reaches.
Score yourself honestly. Named in four or five prompts means you are in good shape and should focus on accuracy. Named in one to three means you exist, but you are not the default. Named in none means you have an entity problem, not a marketing problem, and the fix below is sequential rather than optional.
Two warnings. Results vary between runs, so treat this as a temperature check rather than a rank tracker. And check the sources the assistant cites, not just the answer. The citation list tells you which layer is doing the work.
The five source layers AI assistants check before naming you
Assistants assemble a local recommendation from public sources, and they weight those sources very differently. This is the order we work through in a free GEO audit, because fixing layer four before layer one wastes a month.
| Layer | What it covers | Who leans on it | Effort to fix | Typical time to land |
|---|---|---|---|---|
| 1. Entity consistency | Business name, address, phone, hours and services identical everywhere | All assistants | Low effort, high tedium | 2 to 4 weeks |
| 2. Google Business Profile | Categories, services, photos, Q&A, posts | Gemini, AI Overviews, Maps | Low | Days to index |
| 3. Bing Places and directories | Bing Places, Foursquare, Yelp, ProductReview, industry bodies | ChatGPT, Copilot, Perplexity | Low to medium | 2 to 6 weeks |
| 4. Answer-shaped content | Pages that answer one question in one extractable paragraph | All assistants | Medium, ongoing | 4 to 12 weeks |
| 5. Third-party corroboration | Reviews, press coverage, local media, mentions on sites you do not own | All assistants, heavily | High, slowest | 3 to 6 months |
The pattern is worth naming. The layers that cost the least are the ones most Australian businesses have never touched, and the layer everyone obsesses over, their own website copy, sits fourth.
Layer 1: Make your entity impossible to misread.
If your website says “Canty Plumbing,” your Google profile says “Canty Plumbing and Gas Fitting Wollongong,” and an old directory says “M. Canty Plumbing Pty Ltd,” you’ve given the model three candidate businesses and no way to merge them. Ambiguity is not neutral. It is a reason to recommend someone clearer.
Pick one legal trading name and one format for your address and phone number, then push that exact string everywhere. Include your ABN on your contact page. Add sameAs links in your schema pointing to every verified profile you own, which is the machine-readable way of saying “these are all me.”
Layer 2: Treat your Google Business Profile as a data feed.
Fill every field, not just the ones that feel worth it. Categories and services drive category matching. Photos with real timestamps signal an operating business. The Q&A section is one of the few places you can publish a plain question and a plain answer on Google’s own property, which is exactly the shape an assistant wants to lift.
Layer 3: Claim the Bing side of the map.
Bing Places, Apple Maps, Foursquare, ProductReview.com.au and the directories specific to your trade or profession. For most Australian SMBs, this is an afternoon of work that has never been done, and it is the layer that feeds the assistant your customers are most likely to use. If you have outsourced local SEO and only ever received Google reporting, ask what happened to Bing.
Layer 4: Write pages that can be quoted.
Assistants extract passages, not pages. A heading phrased as a real question, followed immediately by a 40- to 60-word answer that makes sense on its own, is far more citable than three paragraphs of preamble. Restate the subject in the answer rather than starting with “it”, because the extracted snippet will not have your heading attached.
Then add the schema that labels it: LocalBusiness for who you are, Organization for your brand identity, FAQPage only where the questions are genuinely visible on the page. This is the bulk of what ongoing content creation should be producing in 2026, and it is a different craft from writing 1,000 words for a keyword.
If a paragraph on your site cannot be copied into a text message and still make sense, an AI assistant will not quote it either.
Layer 5: Be talked about somewhere you do not own.
This is the slow layer and the one that decides ties. Assistants weight corroboration from sources you do not control: reviews, local news, industry publications, association listings, genuine mentions. Steady review velocity beats a burst of ten reviews in a week, which reads as manipulation to every platform involved. Never gate reviews or buy them; it breaches platform policy, and the recovery costs more than the reviews were worth.
Earned coverage counts here too. A press release that lands on real publications gives the model something to cite that is not your own marketing, which is precisely the kind of evidence it prefers.
How long before an AI assistant starts naming my business?
Expect 4 to 12 weeks for the fast layers and 3 to 6 months for meaningful shortlist presence. Entity fixes and directory claims propagate within weeks. Answer-shaped content needs to be crawled and re-crawled. Third-party corroboration accumulates slowly. Businesses starting from zero citations should plan on two quarters, not two weeks, before AI answers change.
What to measure when clicks stop being the metric
Traditional reporting will understate all of this because much of the value never produces a session. SparkToro’s 2026 analysis found 68% of Google searches now end without a click, with AI Overviews appearing on more than 20% of searches and cutting click-through rates by close to 60% when present.
So measure three things instead:
- Citation presence. Re-run the Answer Gap Audit monthly and log which prompts name you. It is crude and manual, but it is still the closest thing to a rank tracker for AI answers.
- Quality of the clicks you do get. Being cited is worth more per impression than being ignored. Seer Interactive’s 2026 update found brands cited in AI Overviews earn roughly 120% more organic clicks per impression than uncited brands on those queries. Fewer visitors, yet further through the decision.
- Phone calls and form fills you cannot attribute. Watch the trend line, not the source. Direct and “unassigned” traffic growing while sessions flatten is often AI discovery working, not a tracking fault.
Start here on Monday
Run the five prompts. Write down which of the five layers explains each gap. Then fix them in order: entity consistency, Google profile, Bing and directories, answer-shaped pages, and the long game of being talked about elsewhere.
If the audit shows your name missing from every answer in your own suburb, that is worth knowing this week rather than at the end of the financial year. We run a free GEO audit for exactly this, and if you would rather hand the whole thing over, our Generative Engine Optimisation service picks up from layer one.
