Home · Knowledge base
The whole subject, explained plainly
How AI systems decide which companies to name, how that gets measured, and what can actually be done about it. Written for someone running a business, not for someone who works in search. Organised by subject rather than by date, because nobody reads a library backwards.
- Start here
- Why you are not being named
- How it works
- What to expect
Twelve pieces, each answering one question
Each one answers a single question completely. Where a piece belongs to a step of the method, it says so and links there. Twelve is where the library starts, not where it stops — a piece is added when it has a question of its own that nothing else on this site answers.
Start here — what this is
The vocabulary, and what the terms actually mean once you strip the marketing out.
- What AI visibility actually meansThe six things that get measured, and why “visibility” on its own is not one of them.
- How AI builds supplier shortlistsWhat happens between a buyer’s question and the three names that come back.
- What is generative engine optimisationThe plain definition, what it is not, and why it is a discipline rather than a tactic.
Why you are not being named
The symptoms most companies notice first, and what is actually causing them.
- Why your competitors appear in ChatGPT answers and you do notUsually not a judgement about quality. Usually a question of what the system is confident about.
- Your B2B website stopped producing enquiriesWhere the enquiries went, why the drop is uneven, and what it does and does not tell you.
- When AI describes your company wronglyHow out-of-date descriptions take hold, and why correcting them often moves the number first.
How it works, and how it differs from search
The mechanics. What an answer is assembled from, and what actually changes the outcome.
- GEO vs AEO vs SEOThree acronyms, a lot of overlap, and the one distinction that matters commercially.
- Is SEO dead in 2026?No. But what it is for has changed, and the honest version of that answer is more useful than either extreme.
- Does schema markup help you get cited by AINot in the way it is usually sold. What it does do is worth understanding properly.
- Query fan-out: why one question becomes severalThe questions a system asks itself before answering yours, and why they decide whether you appear.
What to expect
Timelines and trade-offs, including the case for not hiring anyone.
- How long this takes to workFour to six months, why that floor exists, and what the months before it look like.
- Can you do this yourselfYes. Here is exactly what it takes, and the four places in-house attempts usually stop.
One question per piece, decided before a word was written. That is the discipline we ask of a client’s own pages, so it is the discipline this library was built under.
A wheel, not a blog
A blog is a pile sorted by date, where the most useful thing you ever wrote sinks out of sight in a year. This is a library sorted by subject, and it is built the way we build a client’s.
Each article links up to the method step that owns its concept, and across to at most two siblings. The clusters above are a reading order, not that structure — a cluster can hold articles belonging to three different steps. A dense web of cross-links dilutes the signal rather than strengthening it.
Every piece has one assigned question, and no second page competes with it. That rule was written down before the first page of this site existed — we are running the method on ourselves, and this library is where it shows most plainly.
Reading about it will only take you so far
At some point the useful question stops being how any of this works and becomes whether it is happening to you. That one has a measurable answer.