Will an AI Overview name your business?

Someone in Bayswater wants a plan manager. They do not open ten tabs and compare. They type a question, read the answer at the top, and call one of the three businesses it names.

That is the part of local search that changed, and it changed quietly. The ranking did not disappear. It just stopped being the thing the buyer sees first.

The demand is already here, in Australia

When we were mapping this out for our own properties, we pulled Australian search volumes rather than guessing from American commentary. The figures below are our own, from DataForSEO Labs, pulled for the Australian market in August 2026. They were blunt enough to reorganise our roadmap around.

“AI overviews” gets roughly 9,900 searches a month in Australia. “Google AI overview” adds another 880. “Generative engine optimisation” is at 260 and climbing, “AI search optimisation” at 140, “LLM SEO” at 110. Those last few look small until you notice what advertisers pay for them: “LLM SEO” carries a cost per click of about $44, and “AI search optimisation” about $24.

Nobody pays $44 a click for a topic that is not converting.

The keyword difficulty numbers are the interesting half. “AI search optimisation” scores 8 out of 100. “LLM SEO” scores 20. For comparison, plain “local SEO” scores 70. There is a window here, and we do not expect difficulty that low to last.

How the answer actually gets assembled

An AI Overview is not a ranking. It is a summary written on demand, and the model writes it from sources it retrieves at that moment. Three things decide whether you are in it.

It has to be able to find a passage that answers the question directly. Not a page that is broadly about the topic. A passage. If the question is “what does a support coordinator cost in Perth”, the model wants a sentence that answers it, near a heading that asks it. Pages that bury the answer under six paragraphs of preamble lose to pages that lead with it.

It has to trust the source enough to cite it. Note that trust here is not the same as ranking, given how many cited pages sit outside the top 10. In practice the sources that get cited for local queries are rarely individual small business websites. They are the pages that already aggregate and compare: directories, comparison pages, review-led listings. This is the same dynamic that decides who wins the organic listing, which is worth understanding on its own terms in why a directory outranks the businesses inside it.

The entities have to agree. The model is assembling facts about a business from several places at once. If your name, category, service area and phone number differ between your site, your profile and the directories you appear in, you are a less confident answer than the operator whose details match everywhere. Ambiguity is a reason to leave you out, and there is no penalty message telling you it happened.

What we do about it on our own sites

We run both of our directories with this in mind, which is the only reason we can describe it rather than speculate.

Compare Brokers and Perth Disability Connect both publish how the rankings are produced: scored on real Google reviews, refreshed every month, recomputed as new reviews land, and dated so anyone can see when it last happened.

We do that deliberately, because it is the sort of thing a model can lift and a reader can check. A page that says “one of Perth’s leading directories” gives an assistant nothing to quote. A page that says on what basis the list was built, how often it is rebuilt, and when that last happened gives it a sentence.

The whole trick, and an unglamorous one. Publish your method. Date it. Say where the inputs came from.

Five moves that decide whether you get named

Numbered because the order matters. Doing five before one is how most businesses waste a quarter.

1. Find out where you actually stand

Write down the twenty questions a buyer asks before they call someone in your category. Not keywords, questions. Put each one into Google, ChatGPT and Perplexity and record who gets named. Most businesses have never done this and are surprised by who is showing up in their place. Usually it is a directory, a comparison page, or a competitor two suburbs away who wrote one good article. You cannot fix a gap you have not measured.

2. Answer the question in the first fifty words

Heading asks, first paragraph answers, detail follows. An assistant retrieving your page needs a passage it can lift whole. Pages that bury the answer under six paragraphs of preamble lose to pages that lead with it. This costs nothing and is the highest-return change on most sites.

3. Put real specifics in

Prices, timeframes, service areas, capacity, dates. Ranges are fine. Vagueness is what gets you skipped, because there is nothing to quote. “We offer fast turnaround” is unquotable. “Most jobs are booked within five working days” is a sentence an assistant can use.

4. Make your entity boring and consistent

Same business name, same categories, same service area, same phone number, everywhere you appear. Contradictions make you a less confident answer than the operator whose details match everywhere, and there is no penalty message telling you it happened. This overlaps almost entirely with the profile work in the map pack checks nobody in Perth is doing, which is convenient, because most businesses need to do that anyway.

5. Get onto the pages that already get cited

If the aggregator ranks and you do not, being inside it is the faster route to being named than out-ranking it. That is our honest argument for our network, and it is equally true of directories we have nothing to do with.

The assistants are not all the same

Most of the advice written about this treats “AI search” as one thing. Wrong, and the differences change what you do.

Start with Google AI Overviews, because the received wisdom about them is backwards. Ahrefs analysed 863,000 keywords and four million cited URLs and found only 37.1% of AI Overview citations also rank in the top 10 for the same query. Roughly two thirds come from pages that do not. Read that again if you have been told you need to rank before you can be cited, because it is the most useful fact in this article. Being named is not gated behind out-ranking everyone first, which is why a small operator with one genuinely clear page can appear beside businesses ten times its size.

Gemini shares Google’s index and much of its judgement, so effort spent on AI Overviews carries across. Treating it as a separate project wastes money. Copilot sits on Bing’s index, which is the practical argument for claiming Bing Places even though its own traffic is small.

ChatGPT behaves differently. It leans on its index plus live retrieval and cites fewer sources per answer than Google does, so being one of three named is harder and worth more. Perplexity sits at the opposite end: more citations, shown more visibly, updated faster. That makes it the cheapest place to check whether your content is retrievable at all, even if none of your buyers ever use it.

The work barely changes across them. Clear answers, real specifics, consistent details, presence on pages that already get cited. What changes is how quickly you see the result, which is why we check quarterly rather than trusting a checklist.

What not to bother with

Do not rewrite your site in bullet points because someone said models prefer lists. They do not. They prefer answers.

Do not chase “optimising for ChatGPT” as a separate discipline from making your pages clear, specific and consistent. Every practical thing on the list above is also just good local SEO. The difference is the reward has moved: you used to be optimising for a click, and now a share of the time you are optimising to be the business the answer names, whether or not anyone clicks through.

Do not assume this is settled. The mechanics will move. Measure quarterly rather than trusting a checklist that is eighteen months old, including this one.

Questions we get asked

What is AI search optimisation, and is it different from SEO? You will see it called AI search optimisation, answer engine optimisation (AEO) and generative engine optimisation (GEO). The acronyms are mostly marketing. The work is clear answers, real specifics, consistent entity data and presence on pages that already get cited, which is what good SEO always was. What changed is the reward: you are now often optimising to be the business an answer names, not only to be a link someone clicks.

Does structured data help? Yes, and it is underused. Schema markup does not force you into an answer, but it removes ambiguity about who you are, what you sell, where you work and what you charge. Organisation, LocalBusiness, Product, FAQ and Review markup all give a machine facts it would otherwise have to infer. Inference is where you get left out.

How do I measure it? Pick twenty real buyer questions, run them across Google, ChatGPT, Perplexity and Gemini, and record who gets named. Repeat quarterly. There is no equivalent of a rank tracker yet that we would trust on its own, so the manual baseline is still the honest measure.

How long does it take? Faster than classic rankings, because retrieval is not waiting on the same trust signals. Clean up your answers and specifics and you can see movement inside a quarter. Entity consistency takes longer, because the corrections have to propagate across the places that hold your details.

What does it cost? For most local businesses this is not a separate line item. It is a few days of content and profile work inside an existing programme, and our local SEO retainer starts at $1,999 a month plus GST, month to month, with no lock-in. If someone is quoting you a standalone AI package with its own retainer, ask what is in it that would not already be in competent SEO.

Where this leaves you

The businesses that will do well in AI answers over the next two years are not the ones with the cleverest tactics. They are the ones that publish checkable specifics, keep their details consistent, and appear on the pages that already get cited.

None of that is new advice. What is new is that the cost of ignoring it now shows up before the click, in an answer you never see, on a search you will never appear in a report for.

If you want to know whether you are being named in your own category, that is the first thing we check on a call. Tell us the category and the postcode and we will look before we speak to you: book a call.