COSGEO.AI · PLAYBOOKSEVIDENCE: 120 RUNS · 2026-06-13 · BEAUTY OEM CATEGORY
ONE ENGINE, ONE PLAYBOOK

Platform Playbooks

Answer first: the four major AI engines source their manufacturer recommendations from almost entirely different universes — only ~11% of domains are cited by both ChatGPT and Perplexity (Search Atlas, 5.5M answers). A single strategy cannot cover them. Below: what each engine actually cited in our category across 120 measured runs, and the play for each.

PLAYBOOK 01 · PERPLEXITY

The directory eater

In our 30 runs, Perplexity's manufacturer answers were fed by B2B directories and competitor product pages — it is also the engine whose citations overlap most with Google's top 10 (28.6%, Ahrefs), so classic SEO still matters here. Its citation pool rotates violently: we watched a 100% query hit 0% in 48 hours.

alibaba.com 20 citationsmade-in-china.com 15accio.com 12youtube.com 11

The play: marketplace listings titled in the buyer's exact phrasing, structured spec fields complete, plus YouTube factory footage. Re-measure every 2 weeks — wins here decay fastest.

Full Perplexity breakdown →

PLAYBOOK 02 · CHATGPT

The deep reader with a long memory

ChatGPT reads owned websites deepest of the four — and remembers them: 100% of its Yanse citations in our test still pointed to the legacy domain weeks after our 301. It favors entities corroborated by Wikipedia-grade and editorial sources, and rewards citable on-page passages.

yansebeauty.com (legacy) 15 citationsmade-in-china.com 12beaza.com 6

The play: answer-first pages with statistics, quotations and named sources (the KDD 2024 trio, +30–40% visibility); aggressive entity unification (sameAs + 301) and patience — domain memory refreshes on its schedule, which our Scoreboard #2 will time.

Full ChatGPT breakdown →

PLAYBOOK 03 · GEMINI

The crawler that makes things up

Gemini gave us 100% visibility (30/30, median rank #1–3) on the strength of deep owned-site crawling — Google AI surfaces are also the one place where classic SEO carries through (99% of Google AI Mode citations come from top-20 organic results, Bounteous). The catch: 38% of its domain mentions for us were hallucinated variants.

Visibility 30/30Median rank #1–3Domain hallucination 38%

The play: keep the owned site server-rendered, schema-complete and date-stamped; defensively register the domains it invents and 301 them home; repeat the brand-name-plus-domain pairing across every public property.

Full Gemini breakdown →

PLAYBOOK 04 · CLAUDE

The aggregator loyalist

Claude's manufacturer answers ran almost entirely through B2B aggregators — accio.com appeared 104 times in 30 runs, Alibaba 73. Industry data shows Claude also over-weights medical/reference sources (10.85% vs 6.3% average) and reuses the same authority domains across languages more than any other engine.

accio.com 104 citationsalibaba.com 73tradekorea.com 5

The play: audit and claim your accio.com and Alibaba aggregate presence first — it's effectively Claude's index for this category. For skincare claims, dermatologist-grade sourcing pays double here.

Full Claude breakdown →

THE CROSS-PLATFORM CONSTANT

Across all engines, third-party corroboration beats self-description: brands are 6.5× more likely to be cited via third-party sources than their own domain, and brand mentions correlate with AI visibility at r=0.664 versus r=0.218 for backlinks. Own your entity; let others say your name.