Methodology
What AI visibility is, and how AnswerSignal measures it
This page documents the measurement itself: the terminology, the prompt set, the metrics, the evidence behind each number and the limits of what the measurement can prove.
Terminology
- AI visibility
- How often and how prominently an AI assistant names a business when a customer asks a buying question in its category and location.
- Answer-engine visibility (AEO)
- Visibility inside a generated answer rather than inside a list of links. The unit of competition is a sentence naming a business, not a blue link at position four.
- Generative engine optimisation (GEO)
- The practice of making a business easier for a generative system to discover, understand accurately and cite. In practice this is evidence work: stating plainly what you do, for whom, where, and with what proof.
- Citation readiness
- Whether the claims a business wants repeated exist somewhere a machine can read, attribute and quote.
How the measurement runs
- 1. Build the question set. The business profile (services, city, state, country) is substituted into 23 customer-intent templates spanning discovery, comparison, urgency, trust, price and segment intent, producing 30–100 de-duplicated questions.
- 2. Ask and record. The 30 highest-intent questions are put to an assistant and the full answers are stored, together with the model used, latency and any error.
- 3. Extract named businesses. Each answer is parsed into an ordered list of the businesses it names, with the quoted text that names them, plus any URLs cited.
- 4. Score deterministically. Presence, position quality and share of recommendations are combined with fixed weights. The same recorded answers always produce the same score.
- 5. Combine with website evidence. The headline score blends recommendation performance with what is measurably present on the public website, so the report says not only where you stand but which missing evidence is likely holding you back.
What is measured
Mention rate
The share of answered questions in which the business is named at all. Derived by matching the business name against every business named in the recorded answer.
Recommendation position
Where the business appears in the order of businesses named. First place counts 100, second 84, third 70, and later positions decay — because the first name in an answer is the one most customers act on.
Share of recommendations
The business's mentions as a share of all business mentions across the answered set. It shows how crowded the category answer is.
Competitors named
Every other business the assistant recommended, with how often and how highly each was named.
Cited sources
The domains and URLs the assistant referenced when it answered, aggregated so you can see which sources shape the category answer.
Website evidence
Measured facts read from public pages — what is stated, what is missing, and what a machine cannot confirm about the business.
Limitations
- Assistant answers vary by model version, phrasing, retrieval and time. A measurement is a dated observation, not a permanent state.
- No one — including AnswerSignal — can guarantee that an assistant will recommend a business, or place it first. We measure and we show what evidence is missing.
- "Not recommended" is reported as a factual state, never disguised as a low score. If a business was not named, the report says so and shows who was named instead.
- Revenue impact is only ever shown when the underlying operating inputs are supplied. We do not fabricate financial figures.
- The prompt set is generated from the profile you give us. A wrong category or city produces a technically correct measurement of the wrong market.
Common questions
- What is AI visibility?
- AI visibility is how often, how prominently and how accurately an AI assistant names a business when a potential customer asks a buying question. It is measured from recorded assistant answers, not from search rankings.
- What is the difference between GEO, AEO and SEO?
- SEO concerns ranking in a list of links. Answer-engine optimisation (AEO) and generative engine optimisation (GEO) concern being named, described correctly and cited inside a generated answer. AnswerSignal measures the answer, not the link list.
- How does AnswerSignal generate the questions it asks?
- It builds a prompt set from the business profile — services, city, state and country substituted into intent templates covering discovery, comparison, urgency, trust, price and segment intent — then de-duplicates and asks the highest-intent subset.
- Can AnswerSignal guarantee that an AI assistant will recommend my business?
- No. Assistant answers change with model versions, retrieval and phrasing. AnswerSignal measures current recommendation behaviour and shows which evidence is missing; it does not guarantee any placement or recommendation.
- What evidence does AnswerSignal read from a website?
- Publicly available pages only: what the business does, who it serves, where it operates, proof such as service and location pages, and machine-readable structure. Nothing requires site access or credentials.