AEO agencies in Melbourne — comparing on evidence rather than postcode

Compare Melbourne AEO agencies on artefacts you can open yourself — a citation report with its engine, window and prompt count, your own server log, a live prompt run. The postcode is not a criterion. Across 86 DuckDuckGo polls of “lead generation agency Melbourne” between 19 May and 15 September 2026, individual agency sites took the top four places and Sortlist appeared on only 31.4%.

At a glance — the evidence that separates Melbourne AEO agencies:

• The artefact rule. Any criterion you cannot check in a file, a log line or a live answer before you sign is a preference, not a comparison. Office location, team size and “proprietary framework” are preferences.
• Five artefacts do the whole job: a citation report, your server log filtered to OAI-SearchBot, the robots.txt they propose, a dated page diff, and a live prompt run.
• The Melbourne retrieval pool is agency sites, not directories. In our Sydney polls the three international directories took first, second and third; in Melbourne they came fifth, seventh and thirteenth — so an agency’s own retrievability is visible to you here.
• Melbourne place names are genuinely ambiguous. Australia Post lists four suburbs called Cremorne and six called Richmond; Melbourne’s share postcode 3121.

What can I actually compare AEO agencies in Melbourne on?

On four things, and location is not among them: what they can show you about your domain, what they can show you about their own, what they propose to change, and how they propose to measure it. Everything else on a Melbourne pitch deck — the Cremorne office, the team photo, the “GEO-first methodology” — is unfalsifiable from the outside, so it cannot separate two shortlisted agencies even in principle.

Call it the artefact rule: if a claim cannot be checked in a file, a log line or a live answer before you sign, it is not a comparison criterion — it is a preference. Preferences are allowed; they just should not be scored as evidence. The practical effect is that a shortlist meeting stops being a presentation and becomes a request for documents.

The five artefacts to ask for, and what a pass looks like

An agency actually doing the work can produce all five in a day, and none asks you to take anything on trust. Request them before the second meeting.

The fifth artefact is the one that ends most shortlists. An agency that will not run an unrehearsed prompt for a prospect is telling you it has not run many.

What our own polling says about who an engine can even choose from in Melbourne

We run a citation-monitoring rig against a registry of real buyer prompts; two are the literal Melbourne and Sydney commercial queries. Across 86 successful DuckDuckGo result-page polls of “lead generation agency Melbourne” and 91 of the Sydney equivalent, both between 19 May and 15 September 2026 (one prompt per city, one engine, our own database), 44 distinct domains appeared at least once in the Melbourne pool. Our companion page on choosing an AEO agency in Sydney reports the same Sydney prompt on an 89-poll denominator because it was written two days earlier; the polls of 14 and 15 September are the whole difference, and the shares move by less than a point. The two cities behave differently:

Two readings, the second unflattering to us. First, the retrieval pool an engine draws on for a Melbourne agency query is made of agency sites: the directories that own Sydney came fifth, seventh and thirteenth here, so a supplier’s own retrievability is observable rather than hidden behind a Clutch listing. Second, we appeared on none of the 86 Melbourne polls despite being a Melbourne business ourselves — we have written pages about AI search and not earned a place in the Melbourne pool, and an agency selling AEO while concealing that would not deserve a citation either. Result-page presence is a weaker signal than an in-answer citation; our first-party study of 5,051 AI citation polls across ChatGPT, Gemini and DuckDuckGo sets out how far apart the engines sit, and how a marketing agency gets itself cited by ChatGPT is the method we are applying to our own gap.

Melbourne suburb names that an AI answer can get wrong

This local signal is about your entity, not your agency’s. Melbourne’s commercial precincts carry names reused across the country, so a bare suburb string leaves a model a guess to make. From Australia Post’s own postcode search: there are four suburbs called Cremorne — Victoria 3121 (the tech precinct), New South Wales 2090, Queensland 4740 and Tasmania 7024. Richmond appears six times: Victoria 3121, New South Wales 2753, Queensland 4740 and 4822, South Australia 5033 and Tasmania 7025. Brighton appears four times: Victoria 3186, Queensland 4017, South Australia 5048 and Tasmania 7030.

The fix costs nothing: never let a Melbourne place name appear without a disambiguating token in the sentence — the state, the postcode, or “Melbourne” itself — and use that string in your structured data, Google Business Profile and invoices. “Our Cremorne studio” is a coin toss; “our studio in Cremorne, Victoria 3121” is an entity. On the same source, Preston appears four times and Hawthorn twice.

What Melbourne buyers are actually typing

The hiring demand is real, small, and splits in two. Measured in Google Search Console for leadsnow.ai over 30 August – 12 September 2026, the agency-hiring, AI-search and vertical-marketing Melbourne queries we are visible on look like this. Head lead-generation terms are excluded and are an order of magnitude larger — “lead generation agency melbourne” alone drew 259 impressions over the same fortnight:

Two things are worth acting on. Every row earned zero clicks — so did every other Melbourne query we were visible on that fortnight, 879 impressions in total — which is what ranking outside the top ten on a Melbourne query is worth. And the vertical terms sit beside the generic ones: brokers, schools and dental practices are searching for this in Melbourne by name. Ask a shortlisted agency for a citation report in your vertical, not a citation report. A prompt set that contains no Melbourne mortgage-broker question cannot tell you anything about a Melbourne mortgage broker.

Which kind of supplier is wrong for you

Three shapes exist, and each is wrong for someone. A specialist AEO shop is wrong if your pipeline problem is the whole funnel rather than retrieval: they will fix your citations and you will still have no booked meetings. A full-service Melbourne agency is wrong if the narrow discipline is what you need — entity consistency, extractable page structure and citation measurement are usually subcontracted there, at a management margin. In-house plus a contractor is wrong below roughly 20 pages of substantive content or about 15 distinct buyer questions: at that size your constraint is source material, not execution, and no supplier can buy you depth you do not have.

The crossover is a publishing-rate question, not a budget one. If your business can produce fewer than two genuinely new, fact-carrying pages a month, hiring anyone is premature.

What running the evidence check yourself costs

You can generate all five artefacts without an agency, and some Melbourne teams should. The honest cost: someone reads the server logs weekly for AI crawler user agents, someone maintains a prompt list in your verticals and polls it on a fixed schedule, and someone owns entity consistency across your site, Business Profile and structured data. Measurement is the part that quietly breaks. Our own rig ran 5,051 polls across three engines in 77 days — roughly 65 a day — because a weekly manual spot-check cannot separate a real citation gain from engine flicker, and a 44-domain pool like Melbourne’s moves week to week. No polling schedule means no feedback loop, and that is usually where a team hands it over. Our AI SEO and AEO service for Australian businesses runs this method, our evidence-based comparison of Australian AI SEO and AEO agencies applies the same artefact tests nationally, and our comparison of Melbourne lead generation agencies covers the adjacent question of who books the meetings once the citations exist. Our Australia-wide AEO agency comparison holds the national shortlist this page sits under.

Frequently asked questions

Is there a leading AEO agency in Melbourne?

There is no measured leader, because nobody publishes a Melbourne AEO leaderboard with a method behind it. What you can measure yourself is retrievability, and in Melbourne that is unusually easy: across our 86 DuckDuckGo polls of the Melbourne agency query between 19 May and 15 September 2026, the top four domains were individual agency sites rather than directories. Put your own buyer’s question to ChatGPT and Gemini, write down every domain cited across ten prompts, and you have a defensible shortlist in an afternoon.

What is the difference between an LLM SEO agency and an AEO agency in Melbourne?

Nothing reliable — the labels are self-applied and used interchangeably, as the Melbourne search data shows: “llm seo agency melbourne”, “answer engine optimisation melbourne”, “best ai seo agency melbourne” and “leading chatgpt ranking agency melbourne” all appear as separate queries for the same job. Do not shortlist on the label. Shortlist on whether the deliverable is a citation report with an engine, a window and a prompt count, or a keyword-ranking report with an AI-sounding cover page.

How do I check a Melbourne agency can actually get me cited in ChatGPT?

Ask for your own server log, filtered to the user agents OpenAI publishes. Its crawler documentation names OAI-SearchBot, which it says “is used to surface websites in search results in ChatGPT’s search features”, alongside ChatGPT-User for user-initiated fetches and GPTBot for model training. If neither OAI-SearchBot nor ChatGPT-User has fetched your commercial pages in the last 30 days, nothing else on the proposal matters yet.

Can a Melbourne AEO agency guarantee I will be cited in ChatGPT?

No, and a guarantee is a disqualifier rather than a selling point — no agency controls an engine’s source selection, and answers vary between two runs of the same prompt. A defensible commitment is about inputs and measurement: a stated number of prompts polled on named engines at a stated cadence, the artefacts above delivered on a schedule, and a published baseline you can both read. Note also what blocking buys you nothing: Google documents that Google-Extended “does not impact a site’s inclusion in Google Search nor is it used as a ranking signal in Google Search”, while it does govern grounding in Gemini apps.

Is “Melbourne’s best AEO and GEO agency” a meaningful claim?

Only if it points at a list with a published methodology, a stated inclusion rule and a disclosure of who wrote it — and most do not. The phrase appears in Melbourne search data as a real query, which is why agencies put it in their title tags. Treat it as a keyword rather than an award: ask which list, who published it, what the criteria were, and whether the publisher appears on its own list. If nobody can answer that in a sentence, the claim is self-awarded.