Knowledge base

GEO and related terms

Our working definitions of core terms — every fact can be verified via Facts and Evidence.

GEO (Generative Engine Optimization)
Generative Engine Optimization
Short for Generative Engine Optimization — optimizing a brand's visibility and factual credibility for generative AI answer engines, so the brand becomes the primary source cited, summarized and recommended by AI.
Structured data / Schema / JSON-LD
Schema.orgJSON-LDstructured markup
A standard vocabulary embedded in page source to mark up entities (Organization, Product, Service, FAQPage, Article, etc.) in a machine-readable way — the doorway through which AI engines read the facts of a page. Schema.org supplies the vocabulary; JSON-LD is the dominant injection format.
Content twin
content mirrorpublic twin
A stable, public, crawlable and citable twin of content from walled-garden platforms (WeChat OA, Channels, Xiaohongshu). Solves "content is invisible to AI because it lives in walled gardens".
Citability
AI citabilityCitable Score
A brand's ability to be directly cited, summarized and recommended by AI engines in answers. Measured by 8 sub-dimensions — Robots / llms.txt / Schema / Meta / Content / Entity / Signals / AI-discoverability — quantified on a 0–100 scale.
Evidence system
Evidence Graphfactual evidence
A traceable, verifiable evidence chain for a brand's key facts — each fact tied to one or more primary sources (filings / announcements / press / official docs), so AI engines can weigh credibility and choose what to cite.
llms.txt
llms-full.txt
A Markdown site-manifest file for LLMs, typically served at the site root, listing the site structure, key entities and citable excerpts. Often the first stop for ChatGPT / Perplexity and similar engines when they ingest a brand.
Dual-track audit
Dual-Track Audit
Track A compliance in 8 dimensions (crawl / Schema / evidence / channels / performance / accessibility / security / compliance placements — P0/P1/P2 grading) + Track B citability 0–100 (8-dimension scoring). The two merge into an overall score plus a blocker list.