When someone asks ChatGPT for a supplier, it does not hand back ten links. It names two or three companies and explains why. AI optimisation is the work of making sure yours is one of them, and that what the model says about you is actually true.
Search used to hand the buyer ten options and let them choose. That was a good deal for anyone sitting at position four: you were still in the room. An answer engine does not work that way. It reads the same web, forms a view, and returns a recommendation with two or three names in it.
There is no position four in an answer. You are either named or you are not, and the buyer never learns you existed.
This matters most for the businesses that have always won on reputation. Word of mouth is precisely the signal these models are worst at seeing. A dozen founders recommending you at dinner leaves no trace a model can read. Meanwhile a competitor with half your track record and twice your published output gets named instead, because they left evidence and you left conversations.
The gap is not quality. It is legibility.
Every engagement opens with a full visibility baseline across all three engines, so the first thing you get is an honest picture of where you actually stand.
You will be sold AEO by one agency, GEO by the next and AIO by the third. There is no agreed definition separating them, and practitioners use them interchangeably. Here is the honest version.
The umbrella, and the one we lead with. All the work that makes AI systems understand who you are, trust what you claim, and recommend you when it is warranted. Everything below sits inside it. Some people use AIO to mean Google AI Overviews specifically, which is a narrower reading of the same three letters.
The part concerned with being named in the answer. A buyer asks for a recommendation and gets two or three companies back. Applies to ChatGPT, Perplexity, voice assistants, featured snippets and Google AI Overviews alike. Also written answer engine optimization, which is how most of the global material spells it.
The part concerned with being cited inside a longer generated response. It comes from a 2023 academic paper and is currently the most written-about of the three. In practice the levers are the same ones AEO uses: entity clarity, corroboration, and material a model can quote.
Commercially, the distinction does not matter. The underlying work is close to identical, and anyone drawing a hard line between the three is selling you the label rather than the job. What matters is whether a buyer asking an AI hears your name, and whether what it says about you is accurate. That is the whole scope.
The model is not sure you are a single, real, specific business. Your name is ambiguous or your details conflict across sources. When a model is not confident who you are, it will not risk naming you. Cheapest failure to fix, and the most common.
Everything the model can find about you was published by you. Your site says you are excellent, which is what every site says. Without independent sources, the model reads your claims as marketing, not fact.
Your opinions, methods and numbers live in your head, in proposals and on calls. Models reach for material already written down in a form they can lift. If nothing of yours is quotable, nothing of yours gets quoted.
Buyers rarely ask for a supplier cold. They ask comparisons first: this versus that, what it costs, when it is wrong. If you have not answered those in public, the model uses whoever did.
Month one produces a full visibility report across ChatGPT, Perplexity and Gemini: recognition, share of voice against your named competitors, sentiment, and the specific sources each engine leans on. Every month after re-runs it against the same baseline, so movement is visible rather than asserted.
Making your business unambiguous to a machine. Consistent naming and detail across every property you control, structured data that says plainly what you do and who for, and the removal of the contradictions that make a model hedge. Fastest-moving lever, and the one most often left untouched.
Working the gaps the report exposes. Directory and listing accuracy, third-party profiles, industry publications, and the specific corroboration each engine indicated it was missing. Targeted at the sources those engines actually cited, not a generic list of places to be seen.
Turning what you know into material a model can lift: clear positions, named frameworks, real numbers, direct answers to the questions your buyers put to an AI before they ever put them to you. Written to be quoted, not to rank.
A short, honest summary of what moved, what did not, and what is next. Including the months where the answer is that a model refreshed its index and undid some of it, because that happens and pretending otherwise makes the number useless.
No lock-in contract. No setup fee. Cancel any time. If the numbers are not moving and we cannot tell you why, you should not be paying for it.
Full re-scan across all three engines against your baseline. Entity and schema maintenance. Source building against the live gaps. Quotable asset production. A written readout of what moved, what did not, and what is next.
Paid media, website builds, or a guarantee about a number neither of us controls. Design and development beyond copy and structured data gets scoped separately rather than quietly absorbed into the retainer.
Email, any time, effective at the end of the month. The schema, the pages and the report history are yours and stay live. Nothing about this is designed to be hard to cancel.
Where a buyer researches before contacting anyone, and being one of three names in an answer changes who gets the call. Professional services, B2B software, specialist trades.
Grown on referrals, with a reputation far larger than your published footprint. The gap between what you are worth and what a model can see is at its widest, and it closes fastest.
If buyers find you on a map, walk past you, or decide in under a minute, they are not asking an AI first. Fix your local listings and spend the money elsewhere.
With no clear ICP and no settled view on what you do better than the alternatives, this work makes an unclear message legible rather than fixing it. Start with a Strategy Sprint.
AIO (AI optimisation) is the umbrella: all the work that makes AI systems understand, trust and recommend your business. AEO (answer engine optimisation) is the part concerned with being named in a direct answer, whether that is ChatGPT, Perplexity, a voice assistant or a Google AI Overview. GEO (generative engine optimisation) is the part concerned with being cited inside a longer generated response.
There is no agreed definition separating them and practitioners use them interchangeably. The work underneath is close to identical.
It does not matter commercially. GEO is the most written-about term, AEO is preferred by several of the tooling vendors, and AIO is the broadest. Anyone selling you a hard distinction between them is selling the label rather than the work. What matters is whether your business gets named when a buyer asks, and whether what the model says about you is accurate.
SEO earns you a position in a list of links. AI optimisation earns you a place inside the answer itself. When someone asks ChatGPT or Perplexity to recommend a supplier, the model does not show ten blue links: it names two or three companies and explains why. There is no position four in an answer.
The plan is $997 per month, month to month, with no lock-in contract. That covers the monthly re-scan across ChatGPT, Perplexity and Gemini, the entity and schema work, source building against your specific gaps, quotable asset production, and the monthly readout.
Entity and schema corrections can land within one refresh cycle, often four to eight weeks. Source building moves more slowly because it depends on other people publishing. Expect early movement on recognition and confidence inside two to three months, and share-of-voice movement over six to twelve.
No. The two reinforce each other, and Gemini in particular leans heavily on Google results. AI optimisation adds the entity layer, the structured proof and the third-party corroboration that traditional SEO never had to care about.
No, and be careful of anyone who does. These models are retrained and re-indexed on schedules nobody outside the labs controls, and numbers move for reasons that have nothing to do with the work. What is committed is the work itself, the measurement against your own baseline, and a readout that tells you the truth in the months it goes sideways.
The first month starts with one. You get a full baseline across all three engines covering recognition, share of voice, sentiment and the sources each engine relies on, so every month after that is measured against a known starting point.
Impulse purchases, walk-in trade, and anything decided in under a minute. If buyers are not researching before they contact you, they are not asking an AI first either. Being honest about that upfront is cheaper for both of us than twelve months of a retainer that was never going to pay.
$997 per month, month to month, no lock-in. It starts with a full visibility baseline across ChatGPT, Perplexity and Gemini, so the first thing you get is an honest picture of where you stand.