What AI visibility means in simplyBrand

AI visibility is the share of eligible category-question answers that mention the monitored brand. It answers a narrow question: when people ask AI about the category rather than naming the brand, how often does the brand appear? Every result is tied to its sample size and measurement scope.

The metrics and their denominators

Metrics are computed within a scope that can include project, period, engine, market, geography, and topic. Filters change the eligible answer set, so each result is interpreted with its denominator and coverage.

  • Visibility = eligible category answers mentioning the brand ÷ all eligible category answers.
  • Share of Voice = brand-mentioning category answers ÷ the sum of mentioning-answer counts for the brand and tracked competitors.
  • Average position = the mean first-mention rank of the brand among tracked entities in category answers that mention it; lower is better.

Mentions and citations are different signals

A mention is detected only in captured answer text using confirmed brand and competitor aliases. An answer counts once for an entity even if the name appears several times. The first matching character position is used for position ranking.

A citation is a source URL exposed with an answer. Citation metrics group sources by registered domain and preserve full URLs for page-level evidence. A brand can be mentioned without its website being cited, and a cited page does not by itself prove a favorable mention.

Category and direct brand questions stay separate

Category prompts ask broad buyer questions without relying on the monitored brand name. Only qualifying category answers enter Visibility, Share of Voice, and Average Position, which prevents direct brand questions from inflating earned visibility or position.

Direct brand prompts ask about the brand by name and measure brand recognition. Brand-mentioning answers from either category or direct brand prompts can enter sentiment review, but direct brand answers never enter Visibility, Share of Voice, or Average Position.

Missing data stays missing

When a metric has no valid denominator, simplyBrand reports n/a rather than zero. Zero means the denominator exists and the measured event did not occur; n/a means the metric cannot be calculated from the available evidence.

An engine below 80% completeness for a period is excluded from cross-engine rollups and listed as a caveat. Its per-engine result may still appear with that caveat. Refused answers and paused prompts are excluded according to the metric contract.

Sampling and prompt-set consistency

Monitoring is configured by market, language, topic, engine, and prompt mode. Results describe that configured sample; they are not a census of every question a person could ask or every answer an engine could produce.

Each run captures the metric-significant prompt fields used at run creation. Later edits cannot move historical denominators. Material prompt-set changes create a new version, and prompt churn above 30% resets the trend baseline.

Engine coverage and citation limits

Results are reported for the engines included in the configured scope. Models, transports, retrieval behavior, and answer availability differ by engine, so cross-engine comparisons must retain their engine and basis context.

Some engine interfaces do not expose citations. Citation rollups include only citation-capable evidence and disclose that limitation; unavailable citation evidence is never converted into a zero citation rate.

How sentiment is evaluated

Sentiment is judged toward the monitored brand as positive, neutral, or negative for answers that mention it. It is evaluated from answer meaning rather than keyword matching.

Only judged answers enter sentiment shares. Unjudged answers are excluded, judge coverage is reported, and the judge model and version are stored so methodology changes can be identified.

Updates and basis changes

Metrics can be recomputed as new scheduled runs arrive. Trend values compare the current and prior periods within the same scope, without smoothing; a delta remains n/a if either period has no calculable value.

Changes to engine transport, pinned model, prompt-set version, or sentiment judge alter the measurement basis. simplyBrand marks these changes rather than silently joining unlike time-series segments.

Known limitations

AI answers are variable observations produced under a defined configuration. The methodology is designed for repeatable monitoring and evidence review, not for claiming complete knowledge of an engine or causal attribution for ranking changes.

  • Results depend on the selected prompts, markets, languages, engines, topics, and collection period.
  • Engine and model updates can change answers even when the monitored configuration is unchanged.
  • Citation availability depends on what each engine interface exposes.
  • Metric movement shows an observed change; it does not by itself prove which external action caused that change.

Continue exploring simplyBrand

See how the product connects measurements to captured answers and sources, review common questions, or read the privacy policy.