Entity SEO in Australia

Google stopped thinking purely in strings a long time ago. Its Knowledge Graph stores things, not words: a business with an address, a founder, a service, a set of relationships to other things. Structured data is how you tell Google which things your page is talking about, and schema.org is the shared vocabulary both sides agreed to use. None of that is new.

What is new is the cost of getting it wrong. A ranked link is forgiving. If Google is unsure who published a page, it can still show the page and let the reader work it out. A generative answer is not forgiving, because it has to make a claim in a sentence. When an assistant writes "for corporate video in Sydney, consider X", it is asserting that a specific entity exists, does that work, and is trustworthy. If the model cannot resolve your name to a confident record, the safest thing it can do is name someone it can resolve. That is the quiet mechanism behind a lot of missing citations, and it is upstream of everything in how to get cited by AI and generative engine optimisation.

Australian business names are not unique identifiers and were never meant to be. ASIC's business names register exists precisely because the same name can be held, or nearly held, in different forms and different places. Add the rest of the English-speaking world and a distinctive-sounding name becomes a crowd. Every one of those look-alikes publishes a website, collects a directory listing and accumulates mentions, and every one of them is competing to be the record the engine attaches your name to.

When that happens, the engine does what the research on entity resolution says it will: it falls back on secondary signals. Which record has corroboration from independent sources? Which one has a verified place? Which one has details that agree wherever they appear? The business with the most consistent, hardest-to-fake evidence wins the record, and the answer engine describes that one. This is not a penalty, and no algorithm is punishing you. It is a confidence problem, and confidence is something you can supply.

The stakes are visible in our own field data. Across the thirty days to 13 July 2026, the five answer engines we monitor cited 487 different domains on Australian agency and search questions. When the field is that fragmented, no entity is established enough to be the default, so the engine leans harder than usual on whatever evidence it can verify. That is a threat if your evidence is thin and an opening if it is not.

Entity resolution rewards evidence that cannot be manufactured. A rival can copy your name, your tagline, your service list and your page structure. What they cannot copy is a government-issued identifier that resolves to you on a public register.

In Australia there are two, and both are free and machine-readable. The first is your ABN, recorded on the Australian Business Register and published through ABN Lookup, which is operated by the Australian Taxation Office. It carries your entity name, type, status, registration date and location. The second is ASIC, which runs the company register and the business names register: the canonical public source for Australian companies and for who holds a given trading name. Between them, they are exactly the kind of authoritative, independent, non-replicable corroboration an entity-resolution system is built to trust.

The practical move is small and almost nobody makes it. In your Organization schema, add your ABN Lookup record and your ASIC record to the sameAs array alongside your social profiles, and expose the ABN as an identifier on the page where a human can see it too. The sameAs property is defined by schema.org as a link to a reference web page that unambiguously indicates the item's identity, which is exactly what a register entry is. Most Australian sites fill sameAs with Facebook and LinkedIn and stop, which means they hand the engine only signals a look-alike could also produce.

You cannot fix an entity problem you have not measured. This is the audit we run before any generative engine optimisation engagement, and you can run it yourself today. Score every check 0 (clean), 1 (partly) or 2 (a problem), and add them up.

Reading the score. 0 to 3, your entity is resolved and your effort belongs in content and proof. 4 to 6, you are contested, and the fixes in the playbook below will pay for themselves faster than another article will. 7 to 10, you are at risk, and no amount of publishing will help until the engine knows which business is doing the publishing. Assumption: these bands are our working rubric rather than a measured benchmark, offered as a practical way to prioritise.

There is nothing to buy here. ABN Lookup and the ASIC registers are public and free, schema.org costs nothing, and consistency is an editorial discipline rather than a licence fee. As a working assumption, entity work belongs inside an SEO or generative engine optimisation retainer rather than as a line item, because it is the foundation the rest of that retainer stands on. The wider numbers are in how much SEO costs in Australia.

What it does cost is attention, and this is where the leverage sits. Our Snowball SEO platform automates the schema, the entity wiring and the AI-visibility monitoring that other agencies bill by the hour, so more of your budget goes into the creative storytelling that earns the corroboration in the first place: the video, photography and written content that gives other people something true to say about you. The machine needs a record. People need a reason to talk about you. Both are worth paying for, and only one of them should be charged by the hour.