The words this channel is argued in

Answer engine optimisation

Also: AEO

The work of being named, and named accurately, in the answers AI assistants give — rather than ranking in a list of links. It covers what your pages say, how they are structured, and which sources the engines reach for when they answer questions in your category.

Generative engine optimisation

Also: GEO

Used interchangeably with AEO by most teams. Where they are distinguished, GEO leans on the generation side — the content and technical signals you change — and AEO on the answer you are trying to appear in.

Tracked prompt

One question, kept in a fixed set and put to every engine on the same schedule. The set is what makes two weeks comparable: adding prompts you happen to win raises a score without anything changing in the world.

Visibility

How often AI answers name your brand at all, across your tracked prompts and every engine polled. A brand can be well known and invisible here, because the question is what the model says, not what people think.

Share of voice

Your share of all brand mentions in the same answers. It moves for two reasons — you gained ground, or the category got busier — which is why it is read beside visibility rather than instead of it.

Position

Where you land in an answer that names several brands. Fourth in a five-brand answer is a mention, and it is not a recommendation.

Sentiment

How the answers that mention you read — favourable, neutral or otherwise — scored consistently rather than sampled by hand. A mention that calls you expensive is not a win.

Citation

Also: Source

A page an engine drew on to build its answer, and named. Citations are where a mention is earned: they are usually not your own site, which is the point of tracking them.

Citation bot

An AI agent fetching your page because somebody asked a question and the answer is being written now. Distinct from a crawler, and recorded separately, because it means a reader is on the other end.

Crawler

An AI agent reading your site to learn it — for training or for the engine's own index. It tells you the model knows you exist; it does not tell you anyone was shown your page.

Hallucination

An untrue statement an engine makes about your pricing, features or status. Noma records the claim with the exact prompt and model that produced it, so it can be reproduced rather than argued about.

llms.txt

A plain-text file at the root of your site that gives models explicit context about who you are and what you sell. It is a proposal rather than a standard, and cheap enough to be worth having either way.

Grounding

When an engine checks live pages before answering rather than relying on what it memorised in training. It is why a page published this month can appear in an answer at all.

Prompt set drift

A prompt set quietly changing shape — questions added where you are strong, or dropped where you are weak. It moves every number on the page without anything changing in the world, which is why changes are recorded and dated.