GeoChina AI Visibility Glossary
The vocabulary of getting your brand cited by China's AI engines. Nineteen plain-English definitions covering GEO, AIVO, Share of Model, the Big Six and the source platforms that feed them.
The measurement discipline
GEO — Generative Engine Optimization
Generative Engine Optimization (GEO) is the practice of shaping a brand's content, entities and third-party sources so that generative AI engines retrieve, trust and cite that brand inside their answers. Unlike SEO, which competes for ranked links, GEO competes to be the recommendation the model states outright. In China it targets the Big Six engines rather than Google.
AIVO — AI Visibility Optimization
AI Visibility Optimization (AIVO) is a measurement-first discipline for growing how often and how favorably a brand appears inside AI-generated answers, treating AI visibility as a KPI you can track across engines over time. Where GEO names the tactics, AIVO names the outcome and the scoreboard. It reframes brand marketing around the AI answer layer instead of the results page.
Share of Model (SoM)
Share of Model is the percentage of relevant AI answers in which your brand is recommended versus competitors, measured per engine across hundreds of real, high-intent buyer questions. It is the successor to share of voice for the AI era. A brand can hold 40% Share of Model on Doubao yet 7% on Baidu, so GeoAIVO tracks it engine by engine and over time.
AIVO Score
The AIVO Score is a single 0-100 index of a brand's overall AI visibility health, combining visibility, infrastructure, competition and sentiment into one weighted number. It rolls up the results of GeoAIVO's eight scoring modules so teams can report progress with a headline figure. Because weights are fixed, scores stay comparable across audits and across competitors.
Entity graph
An entity graph is the structured web of facts an AI model holds about your brand as an entity — its category, attributes, relationships and the authority sources that confirm them. When the graph is consistent and well-sourced, engines describe you accurately and confidently. When it is thin or contradictory, they hedge, confuse you with rivals, or omit you entirely.
Citation sentiment
Citation sentiment is the tone an AI engine uses when it mentions your brand — describing you as a category leader, a reliable option, or merely a budget choice — independent of whether you are mentioned at all. Two brands can share equal Share of Model yet be framed very differently. Tracking sentiment reveals reputation risks that raw mention counts hide.
Zero-click
Zero-click describes a query in which the user receives a complete answer inside the AI response and never clicks through to any website. This is now the default behavior in generative search: the model summarizes and recommends, and the click never happens. Being ranked #1 on a results page is worthless if no results page is ever seen.
China's Big Six AI engines
The Big Six
The Big Six are the six generative AI engines that dominate Chinese consumer usage: Doubao, DeepSeek, Yuanbao, Qwen/Tongyi, Kimi and ERNIE/Baidu. Together they answer a growing share of buying questions for the 602M Chinese generative-AI users counted by CNNIC in 2025. Each sources answers from a different content layer, so visibility must be won engine by engine.
Doubao (豆包)
Doubao is ByteDance's consumer AI assistant and China's most-used AI app, with roughly 345M monthly active users. It leans heavily on Douyin and ByteDance's in-house content ecosystem when answering product and lifestyle questions. For consumer brands it is usually the single highest-traffic engine to win.
DeepSeek
DeepSeek is a Chinese frontier-model lab whose reasoning models are used directly by consumers and embedded across countless other Chinese products. Its answers tend to be structured and comparison-heavy, rewarding brands with clear, well-sourced facts. GeoAIVO often uses DeepSeek as its judge model when scoring audits.
Yuanbao (元宝)
Yuanbao is Tencent's AI assistant, notable for surfacing WeChat Official Account and Tencent-ecosystem content when it answers. This makes WeChat authority content a direct lever for Yuanbao visibility, unlike any other engine. Brands with strong WeChat publishing tend to over-index here.
Qwen / Tongyi (通义)
Qwen, or Tongyi Qianwen, is Alibaba's model family and assistant, strongest in commerce-adjacent contexts and connected to Alibaba's ecosystem. Because of its shopping lineage, it is especially relevant for brands selling on Tmall and Taobao. Its answers frequently weigh practical purchase criteria.
Kimi
Kimi is Moonshot AI's assistant, known for long-context handling and thorough, research-style answers that cite web sources. It tends to reward depth and reputable references, so authoritative long-form content performs well. Its answers are typically slower to generate — a property of its thoroughness, not a fault.
ERNIE / Baidu (文心)
ERNIE is Baidu's model line, and its AI answers are tightly linked to Baidu Search and Baidu Baike as trusted knowledge sources. A well-maintained Baike entry and strong Baidu presence therefore feed directly into ERNIE visibility. It remains a default reference point for many Chinese users seeking factual grounding.
The content layer AI cites
Xiaohongshu (RED / 小红书)
Xiaohongshu, known internationally as RED, is China's dominant lifestyle and product-discovery community and a primary source layer AI engines draw on for consumer recommendations. Its authentic user posts form much of the consensus that models reflect. For consumer categories, a credible Xiaohongshu footprint is often prerequisite to being recommended at all.
Zhihu (知乎)
Zhihu is China's leading long-form Q&A knowledge community, frequently cited by AI engines for reasoned, comparison-style answers. Its detailed expert threads map neatly onto the way models justify a recommendation. Winning the relevant Zhihu questions is a high-leverage move for considered-purchase categories.
Knowledge atom
A knowledge atom is a single, self-contained, quotable fact about a brand, structured so an AI engine can extract and reuse it verbatim in an answer. Examples include a precise claim, a specification, or a defining differentiator. Publishing your brand as clean knowledge atoms makes it far easier for models to cite you accurately.
Answer-first content
Answer-first content states its conclusion in the opening sentence before elaborating, maximizing the chance an AI engine lifts it directly into a generated answer. It mirrors how models themselves respond, so it is easy for them to extract. This glossary is written answer-first for exactly that reason.
种草 (seeding)
种草 (zhòng cǎo), or "seeding," is the Chinese marketing practice of planting authentic product recommendations across social communities to build the third-party consensus that AI engines later cite. The phrase literally means "planting grass" — creating the desire to buy. Because roughly 85% of AI brand mentions come from third-party content, seeding is a foundational input to AI visibility.
See these terms measured for your brand.
Run a free sample audit and get your Share of Model across the Big Six in about 60 seconds.
Start free audit →