The Method, v1.1

Run every core buyer prompt at least 5 independent times per platform, on the models consumers actually use. Record the triple: visibility rate / median rank / cited domain. Never publish a single-run screenshot — AI answers are non-deterministic. EVIDENCE — SparkToro 2025: <1/100 repeat runs return the same brand list, <1/1,000 the same order. Cost via API ≈ $0.07 per prompt-run.

Each engine has its own source universe. In our category (beauty OEM, measured 2026-06-13, 120 runs): Perplexity feeds on alibaba.com & made-in-china.com directories; Claude on accio.com (104 citations of 120 runs) and Alibaba aggregate pages; ChatGPT reads owned websites deepest; Gemini crawls owned sites but hallucinates domains. Industry-wide, 36.7% of all beauty AI citations flow to just 10 domains — and no brand's own website is among them. EVIDENCE — our raw data (scoreboard); beauty citation concentration: 5.7M-answer study, Feb–Jun 2025.

One brand, one set of facts, everywhere. sameAs schema linking official site ↔ marketplace stores ↔ social profiles; full-site 301 from legacy domains; identical numbers (MOQ, capacity, certifications) across every property. AI engines cross-check — contradictory data gets you dropped. Defensive move: register the domains AI hallucinates for you and 301 them home. EVIDENCE — our Phase 1: Perplexity began citing the new domain within hours of sameAs + sitemap submission. Brand mentions correlate with AI visibility at r=0.664 vs 0.218 for backlinks (Ahrefs, 75,000 brands).

For each core buyer question, one CITABLE page: a 2–3 sentence direct answer up top, a 200–400 word extractable passage, FAQ/Article schema, verifiable numbers, synonym coverage. Add statistics, quotations and cited sources — the three tactics proven to lift generative-engine visibility 30–40%. EVIDENCE — Princeton/Georgia Tech, KDD 2024: cite-sources / quotation / statistics each +30–40% Position-Adjusted Word Count; keyword stuffing ≈ 0. Pages with 19+ data points average 5.4 ChatGPT citations vs 2.8 (SE Ranking, 129k domains).

Optimize where the engines eat, not just your own site: marketplace listings titled in the buyer's exact phrasing (Perplexity, Claude), Reddit answers written from real expertise (ChatGPT), a LinkedIn company page with citable posts (cross-platform #2 source), YouTube factory footage (Google AI), industry directories. Brands are 6.5× more likely to be cited through third-party sources than their own domain. EVIDENCE — Airops; AI search citation structure is ~75–92% earned/third-party media depending on category.

Attack thin-competition niches first (for Yanse: oil-control film, powder-coated blotting paper, one-stop oil-care lines) where owned content alone can reach near-100% visibility — then expand toward head terms with the entity authority you've built. GEO structurally favors challengers: the same optimization that lifts a #5-ranked site +115% drops the #1 site −30%. EVIDENCE — Princeton KDD 2024, per-rank visibility analysis.

Re-run step 01 every 2–4 weeks; after every major move (301, listing launch, content drop), verify causality within 1–2 weeks. Publish the scoreboard — wins and collapses alike. Citation pools rotate 40–60% monthly; occupation is an operating rhythm, not a one-time project. EVIDENCE — our Scoreboard #1: a 100% query collapsed to 0% in 48 hours while two others rose. Profound: 40–60% monthly citation-domain turnover.