GEO Optimisation in Melbourne: How to Get Cited in AI Answers

What is GEO, and how is it different from SEO?

SEO is about ranking. You want to appear high in a list of links, and the reader clicks one of them.

GEO is about being quoted. Someone asks ChatGPT or Perplexity a question, and the engine writes a paragraph of prose. Your business is either named in that paragraph or it is not. There is no second page to be on.

That difference matters more than it sounds. A search result gives you a chance to earn the click. An AI answer has often already made the decision for the reader. If the engine recommends three firms and you are not one of them, you were never in the running, and you will never see it in your analytics.

The two disciplines overlap. Clean, fast, crawlable pages help both. But the target is different. SEO optimises for a position. GEO optimises for a citation.

Most small firms in Melbourne have never checked whether they appear in AI answers at all. That is not negligence. Until recently there was nothing to check.

If you want the ranking side of the picture, we coveredhow to find the right SEO consultant in Melbourneseparately, and ourSEO consultant servicepage explains how we approach it.

Why is there no “ranking” inside an AI answer?

Ask ten people to search the same keyword and they see roughly the same ten results. Ask ten people the same question in ChatGPT and they get ten slightly different paragraphs.

There is no position one. There is no stable list. What exists instead is citation frequency: across many phrasings of the same question, how often does the engine name you.

This is why a single check tells you almost nothing. If you type your business name into ChatGPT once and it says something flattering, that is not visibility. That is one roll of the dice.

Measuring GEO properly means fixing a set of prompts, running the same set every time, and counting. Two things per answer: were you mentioned, and were you cited with a link. Everything else is anecdote.

What happened when we ran the baseline on ourselves?

We ran ours in June 2026, before doing any of the work described below.

Eighteen prompts, grouped five ways: branded questions about Newaurizon itself, category questions like “best fractional CMO in Melbourne”, problem-aware questions where the buyer describes the pain rather than the solution, comparison questions, and one aimed at a specific type of client. Each prompt went through five engines: ChatGPT with search on, Perplexity, Google AI Overviews, Gemini and Claude.

Newaurizon appeared nowhere.

Not a mention. Not a citation. Not once, across any prompt, in any engine.

What did appear were six local firms and a directory listicle. The listicle matters more than any single firm, and I will come back to why.

That is the honest starting line, and it is the correct one for a business this new. The engines had no reason to know we existed. Every citation from here is clean, measurable progress, and I would rather publish the zero than a number I cannot show you the working for.

If you are reading this with a fifteen-year reputation and a book of word-of-mouth clients, run the same test on yourself. Most owner-operators I have done this with are surprised, and not in a good way. Being excellent and being cited are unrelated problems.

What actually makes a business citable?

Nothing here is a trade secret. It is a checklist, and the order matters.

Let the AI crawlers in. An engine can only cite a page it is allowed to read. Six crawlers do most of the work today: OAI-SearchBot and ChatGPT-User for OpenAI, PerplexityBot, Google-Extended, ClaudeBot, and CCBot for Common Crawl. Check your robots.txt. If a developer once blocked bots as a blanket rule, you are invisible by configuration, and no amount of writing will fix that.

Publish an llms.txt. It is a plain index of your key pages with a one-line description of each, written for machines rather than people. Think of it as a sitemap for language models. Most of your competitors do not have one, which makes it cheap ground to take.

Make one page the fact-dense source of truth. Legal name, founder, credentials, prices, coverage area, positioning, all in plain language. Engines parse a page like this once and rely on it for months.

Open every page with a 40 to 60 word answer. Ask the page’s core question, then answer it immediately in a short block. Engines lift these almost verbatim. On most sites this is the highest-return change available, and it costs an afternoon.

Add the schema. Organization, ProfessionalService, Service, FAQPage, Person and BreadcrumbList are the types that tell a machine what it is looking at. It is unglamorous plumbing and it works.

Build FAQs from real prompts. Not the questions you wish people asked. The exact questions they type into an AI, answered in two to four sentences each, four to eight per key page.

State your facts plainly. “Newaurizon Advisory offers three packages from $1,500 per month” is easy for a machine to lift. A sentence with three clauses and a metaphor is not.

Google’sdocumentation on AI featuresexplains how its AI surfaces choose sources, OpenAI publishesa list of its crawlers, and the FAQ markup itself is specified atschema.org.

Why do AI answers quote directories instead of your website?

Because the engine is answering a question, not selling your business.

Ask an AI for the best fractional CMO in Melbourne and it often reaches for a page that has already compared several: a listicle, a directory, a roundup. Those pages are built to answer exactly that question. Your homepage is built to sell to one visitor. The engine picks the one that fits the job.

This is the part most people miss, and it is frequently the fastest mover. Getting onto the pages an AI already trusts can do more in a month than a year of tidying your own site. The directories worth completing are the ones the engines quote: Clutch, GoodFirms, The Manifest. Keep your name, address and phone identical everywhere, because consistency is itself a signal a machine can weigh.

None of that requires new content. It requires filling in forms properly, which is precisely why almost nobody does it.

How do you know whether any of it worked?

You count.

Fix your prompt set. Run it across the engines. Record two things per answer: mentioned, and cited. Add up the citations. That total is your score, and it is the only number that matters.

Then re-run the identical set at day 30 and day 60. Same prompts, same engines, same method. If the number moves, the work is working. If it does not, you change the work, not the measurement.

Here is where I have to be straight with you. We have run our baseline. We have not published an after. The day 30 and day 60 re-measures are scheduled, and when they land we will publish them whether they flatter us or not.

So this post offers you a method and a starting line. It does not offer a result, because we do not have one yet, and I am not going to borrow somebody else’s statistic to make our zero look better.

When GEO is not your problem

My father ran a small business. No marketing, no strategy, just hard work and word of mouth, and he made it work. He built a reputation one client at a time.

I have always thought about what he could have done with even a basic marketing engine. Not to replace the hard work. Nothing replaces that. Only to make sure the right people could find him in the first place.

That is the gap in most of the businesses I work with. They have done the hard part. Their clients love them. They simply have not worked out how the next client finds them. Word of mouth is real, and it has a ceiling, and the ceiling arrives sooner than anyone expects.

But visibility is not always the thing holding you back, and I would rather say so than take your money.

If you cannot service the clients you already have, GEO is not your problem. If your pricing does not work, more enquiries will lose you money faster. If people meet you and do not buy, the issue sits upstream of marketing, and no amount of citation frequency will fix it.

GEO is worth doing when you are good, you are known by the people who already know you, and you are invisible to everyone else. That is a visibility problem. It is entirely fixable. If that is not you, keep your money and come back when it is.

If the constraint really is visibility, the next question is which levers to pull and in what order. That is what amarketing strategy for a small Australian businessis for, and it is the work behind ourdigital marketing consultingengagements.

Frequently asked questions