Nostimates

Glossary

AI search, defined precisely

The vocabulary of AI visibility is young and used loosely. These are the definitions we use across Nostimates, written so you can paste them into a report or a deck without hedging.

AEO (Answer Engine Optimisation)

Optimising to be cited inside AI answers rather than ranked in blue links.

GEO (Generative Engine Optimisation)

The same discipline as AEO, named after the generative engine rather than the answer.

AI Overviews

Google's AI summary box above the classic results.

AI Mode

Google's conversational tab with no blue links.

Citation share

The share of sampled AI answers in which a brand or domain is cited.

n-sampling

Running the same prompt n times and reporting the distribution.

Query fan-out

Decomposing one question into many parallel searches before answering.

Grounding

Attaching a model's answer to retrieved sources it can cite.

Share of voice (AI)

A brand's citation share relative to its named competitors.

Extraction success rate

The share of collection attempts that produce a correctly parsed answer.

Surface

An interface a human actually looks at, as opposed to a model or an API.

Answer engine

Any system that returns a synthesised answer instead of a list of links.

Citation

A linked source attached to an AI-generated answer.

Brand mention

A brand named in the answer text, whether or not it is linked.

Mention rate

The proportion of samples in which a brand is named.

Confidence interval

The range a measured rate would plausibly fall in on repeat measurement.

Statistical significance

Whether an observed change is larger than the measurement noise.

Citation volatility

How much an engine's citation set changes between identical prompts.

Non-determinism

The property that the same input can produce different outputs.

Reproducibility

Whether someone else running your method would get your number.

Identity-free collection

Measuring logged-out, so no account history contaminates the result.

Personalisation

Answer changes driven by the user's history rather than the query.

Locale

The country and language a measurement was collected in.

Residential proxy

Egress through consumer IP space so answers reflect a real market.

Bot detection

The systems engines use to identify and block automated collection.

Collection cost

The real per-sample cost of getting one answer out of one engine.

Credit

One collection: one prompt, on one engine, sampled once.

Evidence retention

Keeping the raw answers behind every reported number.

Audit trail

The record of how a number was produced.

Parser version

The identifier of the extraction logic used for a given result.

Canary prompt

A known-answer prompt run continuously to detect extraction breakage.

Data freshness

How recently a reported figure was actually collected.

Cadence

How often a prompt is re-run.

Prompt set

The list of questions a brand is measured against.

Head term

A short, high-volume query with heavy competition.

Long-tail query

A specific, lower-volume prompt that is easier to win.

Competitive set

The declared list of brands measured alongside yours.

llms.txt

A proposed root file describing a site to AI systems.

robots.txt

The file controlling which crawlers may access your site.

AI crawler

Any bot collecting web content for an AI system.

Training crawler

A bot collecting content to train models, not to answer queries.

Retrieval-augmented generation (RAG)

Generating an answer from documents retrieved at query time.

Hallucination

A confident answer that is not supported by any source.

Answer box

A direct answer rendered above classic results.

Featured snippet

The pre-AI answer box, drawn verbatim from a single page.

Zero-click search

A search satisfied on the results page, with no visit.

Entity

A distinct thing a model can reason about, such as a brand.

Entity salience

How strongly a brand is associated with a topic.

Knowledge graph

A structured store of entities and their relationships.

Structured data

Machine-readable markup describing page content.

JSON-LD

The preferred syntax for embedding structured data.

schema.org

The shared vocabulary for structured data.

FAQ schema

Structured markup for question-and-answer content.

Answer-first content

Leading with the direct answer before the context.

Chunk

The passage-sized unit a retrieval system actually indexes.

Embedding

A numeric vector representing the meaning of text.

Semantic search

Retrieval based on meaning rather than keyword match.

Prompt injection

Hiding instructions in content to manipulate an AI system.

Market coverage

Which countries and languages an engine is measured in.

Sentiment analysis

How favourably an answer describes a brand.

Position tracking

Classic rank monitoring, and why it does not transfer.

White-label data

Data resold under the reseller's own brand.