Spectral Margin Index is the differential between an entity's Citation Probability Score and the mean CPS of its competitive set. It quantifies relative AI visibility advantage.
A positive SMI indicates that the entity holds a measurable citation advantage over competitors. A negative SMI indicates vulnerability — the entity is at risk of competitive displacement in AI retrieval systems.
AI retrieval is competitive. When an LLM responds to a query, it selects from a pool of candidate sources. The entities with stronger embedding alignment and higher structured authority are more likely to be cited.
CPS measures absolute citation likelihood. SMI measures relative position. An entity with a CPS of 0.75 may still be vulnerable if competitors average 0.78.
SMI captures this competitive dynamic. It answers: "Am I ahead, behind, or at parity with my competitive field?"
Competitor CPS is estimated using observable structured authority signals: structured data presence, glossary and definition density, schema coverage, machine-readable discovery files, and public citation frequency sampling.
These estimates are clearly labeled as modeled values based on publicly observable signals. 411bz does not fabricate competitor data.
The estimation algorithms are proprietary. The methodology principles are disclosed in our open methodology.
SMI is one of the inputs to collapse probability modeling. A narrowing spectral margin increases the probability that embedding drift or model updates will result in citation loss. Monitoring SMI over time provides early warning of competitive erosion.