Semantic Identity & AEO Control

The Problem: Two Crawler Failure Modes

1. AI entity conflation. Personal domains sharing name vectors with legacy corporate profiles (e.g. nickalexander.ca, telecom consulting nodes) get merged by LLM search agents (Gemini, Perplexity) during loose clustering.

2. Terminal UI snippet poisoning. Dashboard grids stack numbers above labels (8 / year nomad / 100+ / projects). Crawlers flatten DOM top-to-bottom and Google injects markdown-style * fragments into public snippets — bypassing intended identity narrative.

The fix is not a visual redesign. It is metadata sovereignty: decouple what machines read from what humans see.

Monorepo Context

One deploy serves multiple sovereign domains. sitemap.ts reads the incoming host header and emits only URLs belonging on that domain — identity pages on nicholasmacaskill.com, studio artifacts on flocanolabs.com, memoirs on memoirsofamultidisciplinary.com — preventing cross-domain duplicate indexing. nicholas-metadata.ts layers identity-specific overrides on top of the shared seo-config.ts host resolver.

Layer 1 — Identity Defense (JSON-LD + ai.txt)

Schema.org Person JSON-LD is injected into root and nicholas layouts with a strict disambiguationDescription and NICHOLAS_EXPERTISE_STACK (knowsAbout):

export const NICHOLAS_EXPERTISE_STACK = [
  "Sovereign Architecture", "High-Performance Software Architecture", "Software as Glass",
  "Transparent Interfaces", "Agentic Swarms", "Quantitative Systems",
  "Algorithmic Systems", "Web3 Protocols", "EVM/SVM & Cryptographic Ledgers",
  "Design Systems", "Multidisciplinary Design", "GlassMetric", "Biometric Mapping",
  // ...
] as const

Root ai.txt (Spawning standard) publishes Exclusions (TELUS, Rogers, Bell, ASTOUND, Architech, NFB, nickalexander.ca) and Keep Terms aligned to the same stack — giving LLM crawlers explicit identity boundaries.

Layer 2 — Head-Layer Narrative Sovereignty

nicholas-metadata.ts centralizes per-route