Ask Claude or ChatGPT to recommend a bookkeeper in Parramatta and you get a short list of three to five names, not ten blue links. If your business is not one of those names, you are invisible to that buyer. Australian SMBs are starting to notice the shift: enquiries now arrive saying "Claude suggested you" or "ChatGPT put you on a shortlist", and the businesses collecting those enquiries are often not the ones ranking first on Google.
The practice of earning those mentions has a name: generative engine optimization, or GEO. This guide explains how AI assistants decide which businesses to name, what that means for generative engine optimization in Australia, and the specific checklist we work through with clients.
GEO is not SEO with a new name
Traditional SEO is about ranking a page for a query. GEO is about making your business retrievable and citable as an entity. When an assistant answers "who does Claude training in Melbourne?", it is not ranking pages. It is assembling an answer from what it already knows, what it can fetch through live web search, and what it can verify across multiple sources. A page that ranks fourth on Google can be the first business named in an AI answer if it states its facts more plainly.
The second difference is that AI answers are zero-click by default. The user may never visit your site. That means the assistant has to be able to describe you accurately from the text it retrieved: what you do, where you operate, roughly what you charge. Vague brand copy gives it nothing to work with, so it names a competitor whose pricing page says "fixed fee, $3,500, Sydney metro" instead.
The third difference is competition. Almost no Australian small business is doing GEO deliberately in 2026. The same effort that barely moves a Google ranking can take you from absent to consistently mentioned in AI answers, because the field is close to empty.
How AI assistants decide which businesses to name
Every assistant works a little differently, but the pattern across Claude, ChatGPT, Gemini and Perplexity is consistent. When a question has commercial intent, the assistant runs a web search, reads a handful of results, and synthesises an answer with citations. The businesses that get named share traits: their websites answer the exact question in plain language, their name and location appear consistently across independent sources, and third parties corroborate that they exist and are credible.
Corroboration matters more than polish. An assistant deciding whether to recommend your firm is effectively asking: can I verify this business from more than one source? Industry association directories, Google Business Profile, local press, supplier case studies and detailed customer reviews all count. A beautiful website that no other site mentions is a weak signal on its own.
Structure also matters. Assistants and the search indexes behind them parse schema markup, headings phrased as questions, and pages with concrete figures. Content written the way a buyer asks the question ("how much does a Claude setup cost in Brisbane?") is far more retrievable than content written the way a brand talks about itself.
The GEO checklist for Australian SMBs
Here is the sequence we use, ordered by effort against payoff. Most of it is achievable in-house over a quarter.
Write pages that answer buying questions directly. One page per real question, with the answer in the first paragraph, prices in AUD, and the suburbs or states you actually serve.
State your entity facts on every key page: business name, what you do in one sentence, locations, service area and ABN. Assistants can only repeat what they can read.
Add schema markup. LocalBusiness, Service and FAQ schema turn your facts into machine-readable statements, which is exactly the format retrieval systems prefer.
Get listed where assistants look: your industry association directory, Google Business Profile, and one or two reputable Australian directories. Keep name, address and phone identical everywhere.
Collect reviews with detail. "Fixed our BAS backlog in a week, Wollongong" is retrievable; a bare five-star rating is not.
Publish comparison and pricing content. Assistants reach for pages with concrete numbers when a user asks about cost, and most of your competitors refuse to publish any.
Earn a few third-party mentions: local press, a chamber of commerce profile, a supplier case study. Two or three independent citations shift you from unverifiable to recommendable.
What it costs and how to measure it
Measurement is refreshingly manual right now. Pick five to ten prompts your buyers would plausibly ask ("best NDIS software consultant in Adelaide", "who can set up Claude for my accounting firm") and run them through Claude, ChatGPT and Gemini monthly. Record whether you are mentioned, who is, and which sources get cited. Add a "how did you hear about us?" field to your enquiry form; AI-referred leads will tell you.
On cost, the checklist above is mostly labour. Doing it in-house takes a few hours a week for a quarter. Engaging an agency for the schema, entity and content work typically runs $3,000 to $8,000 for an SMB site. Against that, weigh what one AI-referred client is worth: for a commercial fit-out firm or a mid-tier accounting practice, a single new engagement can be worth $45,000 or more. The maths does not need many wins to work.
Be realistic about volume. AI-referred enquiries are a trickle today, not a flood. But they arrive pre-qualified, because the assistant has already matched your services to the buyer's stated need, and the trickle is growing every quarter while the cost of showing up stays low.
If you want a second pair of eyes on where your business currently stands in AI answers, book a free brainstorming session and we will run your priority prompts with you and map the gaps.
