Presence, SoV & citations

Every component score is built from three complementary metrics. Presence asks were you there at all, share-of-voice asks how much of the answer was you, and citations record exactly which URLs the engine used. Together they turn a fuzzy “are we visible?” into three measurable dimensions.

Presence

Presence is the probability that an engine surfaces your brand for a prompt — by citing, linking, recommending or naming it. Because answers are non-deterministic, presence is never a single yes/no: each prompt is run K times across temperature, persona and locale, and the outcomes become a Monte-Carlo presence probability with a 95% confidence interval. Detection combines URL matching, brand named-entity recognition and semantic matching so a mention counts even when your exact URL is not linked.

Share of voice

Share of voice measures how much of the answer space you occupy relative to competitors. Being mentioned once in a crowded answer full of rivals is very different from being the primary recommendation; SoV captures that difference and is the metric to watch when you are present but being out-shouted.

Citations

Citations are the concrete URLs an engine linked or cited in its answer — the raw evidence behind presence and SoV. They tell you which of your pages (or which third-party pages) are doing the work, and they surface the earned-media sources engines lean on, such as Reddit, Wikipedia and G2. Citation data is what drives the optimization levers.

Provenance and honesty

Each metric value is labelled live, cached or mock (sample), and mock data is never shown as live. A wide confidence interval on presence is a signal to sample more, not a defect. See sampling & confidence intervals and statistical honesty for how the numbers are produced. These metrics roll up into the per-engine breakdown and the overall Lokrix AI Score.