AI visibility
A brand's presence in AI assistant answers: whether a model mentions it, how it describes it, where it places it and what it backs it with. Measured on buyer questions, not brand prompts.
Glossary
The AI visibility glossary.
The twenty-six terms this site, the methodology and the report use, each in two sentences. The definitions are written to be quotable in full.
Core terms
Visibility and measurement
Discovery question
A buyer question without the brand name, e.g. about the best option in a category. It measures organic discovery; only these questions count toward the index.
Brand-aware question
A question naming the brand. It measures recognition and description accuracy, and is never counted as organic discovery.
AI Visibility Index
A 0-100 composite computed from scoreable discovery observations with published weights: mention 20%, recommendation or alternative 30%, prominence 20%, citations 15%, description accuracy 15%.
Question-provider observation
One question asked of one provider with the outcome recorded. A standard audit has up to 100: 20 questions times five providers.
Sealed test set
The approved question set frozen with a SHA-256 hash before the first test. The seal makes the result repeatable and comparable, and impossible to rewrite quietly.
Snapshot
One audit run: a snapshot of visibility on a given day, market and configuration. Later Snapshots on the same set show the change.
Mention
The brand appears in the answer in any form. Appearing at all is the lowest tier of visibility.
Active recommendation
The model explicitly presents the brand as a recommended choice rather than merely naming it. A different class from a mention, counted separately.
Prominence
The brand's position in an ordered recommendation: first, second, third or lower. A short use-X directive can be stronger than a paragraph of description.
Citation support
The answer points to sources, and the brand is or is not among them. An answer with citations that exclude you is a specific, measurable gap.
Description accuracy
Whether the model understands what the company is, for whom and what it offers. A wrong description can be worse than no mention, because it cements the wrong association.
Benchmark and diagnostics
Comparison and conditions
Competitive benchmark
A comparison with up to three commercially substitutable competitors chosen from market research, not from whoever a model happened to name. Both sides get the identical research plan.
Six benchmark dimensions
Category association, third-party authority, recommendation footprint, attribute association, proof and verifiability, and reputation.
Five comparison states
Client stronger, comparable, benchmark stronger, mixed, indeterminate. Fixed scoring rules set the direction, not analyst discretion.
Substitutability
The entry condition for the benchmark: a buyer could realistically choose that company instead of yours. A tool is not a substitute for an agency just because an answer named it.
Technical readiness
Thirty signals across four areas: technical accessibility, entity and product clarity, buyer content coverage, and trust and authority evidence. Supporting conditions, not ranking factors.
First-party and third-party evidence
What a company says about itself versus what independent sources say. The audit keeps them apart, because models weigh them differently.
Evidence state
Every research finding is validated: confirmed, partial, unverified or conflicting. Weak evidence is kept for review but never builds conclusions.
Unavailable is not zero
A failed observation after three retries is recorded as unavailable and excluded from denominators. A provider outage is not zero visibility.
Confidence
The published level of evidence support for a finding: high, medium or low. Separate from the score itself, because 34/100 at high confidence means something different than at low.
Strategic priority
An audit conclusion saying what should improve and why, with confidence and the competitive gap. Deliberately not an implementation instruction.
Names of the discipline
GEO, AEO and relatives
GEO
Generative Engine Optimization: work on a company's visibility in AI-generated answers. In practice it overlaps heavily with good SEO, but is measured by presence in the answer, not a list position.
AEO
Answer Engine Optimization: an older name for the same work, stressing answer engines. The differences between GEO and AEO are mostly vocabulary.
LLMO
Large Language Model Optimization: another name for the same discipline, from the language-model side.
AI SEO
A colloquial term joining classic SEO with work for AI answers. Here it is one body of work at one price, because the same foundations serve both results.
Terms are the start. Measurement is the proof.
The $49 audit turns this glossary into your numbers: 20 questions, five providers, a report with priorities.