AI Visibility Hypothesis: Semantic Proximity

Overview Symmetric Proximity Association Semantic proximity describes a failure mode distinct from semantic drift. A correctly recalled entity becomes incorrectly associated with third party activity through probabilistic co-occurrence inference from signals sharing the same namespace. This document records a hypothesis regarding semantic proximity as a distinct entity contamination mechanism operating at the probabilistic inference layer. Semantic proximity is distinguished from semantic drift as formalized in the AI Visibility Semantic Stability and Drift Theorem (Mas, 2026). The Mechanism Where semantic drift describes meaning degradation through inconsistent signal emission from within an entity corpus, semantic proximity describes a failure mode in which a correctly recalled entity absorbs incorrect attribution from third party signals occupying the same geographic, topical, or organizational namespace. The system calculates probable association from co-location rather than explicit attribution. Proximity Association Symmetry Proximity association may not be symmetric. When one entity in a shared namespace carries significantly greater signal weight than co-located third parties, observed association runs toward the higher weight entity. Documented Instances Three instances are recorded across two distinct proximity vectors: physical co-location involving Razor Rank and third party training companies operating in the same building, and organizational co-location involving remediation engagements within high-profile enterprise namespaces. Retrieval Layer Scope Observations occurred at the incident retrieval layer, not during a training cycle, and are treated as indicators of how probabilistic selection filters may operate across both layers. AI Visibility Framework Position This hypothesis extends the AI Visibility Semantic Stability and Drift Theorem and the Authorship and Provenance Determinism Theorem by identifying an external contamination vector not previously addressed in the framework.