get-cited-by-ailisted

# Get Cited by AI: citation optimization that starts from measured data ## Overview Every recommendation this skill produces traces to an evidence-tagged rule in RULEBOOK.md: a rulebook derived from 26,767 measured AI citations, a 751-page same-template controlled regression, and a scored-predictions ledger. The core finding, converged on by two independently-run analyses (including an adversarial verification pass) of the same portfolio: **Demand selects citations. Craft is a floor, not a lever.** (RULEBOOK A1: demand odds ratio 4.42, p=0.003; word count 1.09, not significant. Uncited pages score HIGHER on FAQs, tables, and schema than cited ones, B3.) **Violating the letter of this workflow is violating its spirit.** The workflow order exists because untrained agents reliably skip step 1 and prescribe craft. ## The Iron Rule **NO recommendation before a demand probe.** Not for "quick checklists," not under time pressure, not when the user asks specifically about schema. If the user's niche has no grounding demand (or engines answer it from memory without searching), citation is structurally impossible and every craft-hour is wasted (A5, F3). Run `modules/demand-probe.md` first, or state exactly what inputs you still need to run it. ## Workflow 1. **Demand probe**: `modules/demand-probe.md`. Establish: does grounding demand exist, what shape is it (aggregate vs entity), and do engines actually search this niche or answer from memory? Output gates everything dow