AI search
Generative Engine Optimisation
A generative engine does not hand back a list. It reads a handful of sources, writes an answer in its own words and names the businesses it is confident about. Generative engine optimisation is the practice of being one of those businesses. It is narrower than SEO and more decisive, because the answer usually mentions three names, not ten.
- Built around the prompts your buyers actually use
- Tracked across AI Overviews, ChatGPT, Gemini and Perplexity
- One client per industry
Why it matters now
Ten links became three names.
In a list of links, tenth place still gets a trickle. In a generated answer there is no tenth place. The engine names two or three businesses, and everyone else is simply not in the room while the decision is made.
That makes position much sharper, and it also makes the inputs more legible. An engine tends to quote sources that are specific, current, structured and corroborated elsewhere. Those are all things you can go and build.
The work is to become the obvious source for a defined set of questions, then widen the set.
The mechanics
An engine quotes what it can verify, not what it likes.
Why specificity beats polish.
Read a generated answer closely and you notice what gets quoted: a number, a timeframe, a named process, a condition. Engines reach for the passage that resolves the question, and vague marketing copy resolves nothing.
So GEO pushes in a direction most websites find uncomfortable. Fewer adjectives, more facts. Say what is included, what it costs to start, how long it takes, who it is not for. Those are the sentences that get lifted.
The side effect is that the pages convert better for humans too, who were asking the same questions and getting the same non-answers.
What this includes
What GEO work looks like in practice.
Prompt research
We build the list of prompts a buyer in your category actually types, including the messy conversational ones that never appear in a keyword tool.
Answer auditing
For each prompt we record what the engines say now, who they name and which pages they pull from. That is the scoreboard.
Source building
We create or rework the pages that deserve to be quoted for those prompts: specific, sourced, current, and structured so a passage can be lifted cleanly.
Corroboration
Engines prefer facts they can see in more than one place. We get your claims, credentials and specifics repeated on third-party sources that models trust.
Structured data
Schema that matches the copy exactly, so there is no gap between what you say and what a machine reads.
Share of answer
Monthly tracking of how often you are named across the prompt set, and which competitor you took it from.
Scope
What the service can include.
Every engagement is scoped to what the site actually needs. Nothing on this list is billed because it appears on a list.
- Commercial prompt set research and sign-off
- Baseline answer audit across every major engine
- Named competitor and cited source mapping
- Source pages built to be quoted
- Fact corroboration on third-party sources
- Schema aligned to the published copy
- Share of answer tracking by prompt
- Monthly re-run and gap plan
Straight answers
Where GEO advice goes wrong.
ClaimedYou can submit your site to AI engines.
ActuallyThere is no submission queue. Engines retrieve from the open web and from their own indexes. Being retrievable, specific and corroborated is the whole submission process.
ClaimedMore content means more mentions.
ActuallyMentions follow the best available source for a question, not the biggest library. One excellent page for a prompt beats twenty adjacent ones.
ClaimedGEO is a one-off project.
ActuallyGenerated answers change month to month as sources and models change. Without re-measurement you have no idea whether you still hold the position you paid for.
ClaimedHidden text can steer the answer.
ActuallyIt gets detected, it does not survive, and it puts the site itself at risk. We do not do it, and we would be wary of anyone offering it.
How we work
From first audit to measured result
- 01
Prompt set
Agree the fifty or so prompts that matter commercially. Everything after this is measured against them.
- 02
Baseline
Run the set across the major engines and record every named business and cited source.
- 03
Gap plan
Rank the gaps by how close you already are and how much the question is worth. Cheap wins first.
- 04
Build and corroborate
Publish the sources, then get the facts inside them repeated elsewhere.
- 05
Re-run
Re-run the prompt set monthly. Generative answers drift, so measurement has to be regular to be worth anything.
What better performance looks like
The outcome we are actually working toward.
A generated answer names two or three businesses. Holding one of those places for the questions that precede a sale is worth more than a page of links you are halfway down.
Plain english
The words, defined.
- Generative engine
- A search tool that writes an answer rather than returning a list: AI Overviews, ChatGPT search, Gemini, Perplexity and Copilot.
- Share of answer
- How often your business is named across a fixed set of prompts, expressed against your competitors. The working metric for GEO.
- Grounding
- The sources an engine fetches and leans on while writing an answer. Being well grounded is how you end up quoted.
- Corroboration
- The same fact appearing on independent sources, which gives an engine enough confidence to state it.
- Prompt set
- The agreed list of buyer questions used as the measurement baseline, so progress is comparable month to month.
Proof, not promises
Thirty-two Australian businesses, rebuilt and measured.
Every build in our showcase is a live site you can open, scroll and judge for yourself — with the brief, the approach and the result written up beside it.
- 32client builds in the showcase
- 6answer engines we optimise for
- GooglePartner certified agency
Honest fit
We would rather tell you now than in month three.
A good fit if
- Your buyers research before they commit
- You have facts worth quoting, or are willing to publish them
- You want a measurement baseline rather than a feeling
- You can commit to a few months of source building
Probably not if
- You want a guaranteed position in an AI answer
- You cannot publish specifics like inclusions or timeframes
- You need results inside a month
- You are not prepared to touch the existing site
Questions
Generative engine optimisation, answered
Is GEO just a new name for SEO?
No, but it sits on top of it. SEO decides a lot of what gets retrieved. GEO decides whether what is retrieved gets used in the answer. You need both, and the second does not work without the first.
What is the difference between GEO and AEO?
AEO is about being the direct answer to a specific question. GEO is about being named in a longer composed answer that weighs several options. The work overlaps heavily, so we usually run them together.
Which engines do you optimise for?
Google AI Overviews and Gemini, ChatGPT search, Perplexity and Copilot. They behave differently, so we report them separately. A win in one does not mean a win in all.
How do you measure it?
A fixed prompt set, run monthly, with your share of named mentions and cited sources tracked over time. No proxy metrics standing in for the real thing.
How long does it take?
Structural and schema changes can show up within weeks. Going from unmentioned to regularly named in a competitive category usually takes a few months of source and corroboration work.
Can this work for a local service business?
Yes, and often faster, because local prompts have fewer credible sources competing. Your Google Business Profile and local citations do a lot of the heavy lifting here.
Works well with
Where this fits in the rest of the work.
Find out what an AI search engine says about you today.
We will run your brand through the answer engines your customers actually use, show you what comes back, and tell you plainly whether we can improve it.