Score anatomy
The headline Lokrix AI Score is not a black box. It is a weighted roll-up of per-engine component scores, each of which is itself built from presence, share-of-voice and citation signals for that engine. This page walks the stack from the raw metric to the single number.
From metric to headline number
Reading bottom-up: each prompt run yields a presence outcome for a given engine; those outcomes aggregate into a Monte-Carlo presence probability with a confidence interval; per engine, presence, share-of-voice and citations combine into a component score; the component scores are then weighted into the overall 0–100 Lokrix AI Score.
Per-engine component scores
Every engine in scope produces its own 0–100 component score with its own confidence interval. Because the engines ground on different indexes, the same property can score highly on one and poorly on another — a Bing-grounded engine like ChatGPT or Copilot can diverge sharply from a Google-grounded engine like Gemini. Keeping the scores separate is what makes the breakdown actionable.
How components are weighted
The overall score weights each engine’s component score and each prompt’s contribution, so that heavily-sampled, high-confidence results carry more of the headline than sparse or cached ones. This is why running more live samples both tightens the interval and can shift the number: you are giving well-evidenced engines more weight.
Provenance carries through
Every component score inherits the provenance of its inputs — live, cached, or mock (sample). A component built partly from predicted rather than live data is flagged as such, so you always know how much of the headline is measured versus modelled. See statistical honesty for the full policy.
Related
Explore the engine-by-engine view in the per-engine breakdown and the metric definitions in presence, SoV & citations.