eastbound
Book a Consultation Run AI visibility audit

China AI visibility · Onboarding

Getting started with China AI visibility

Eastbound publishes a lot of material on Mainland Chinese AI visibility — five long-form research studies, a measurement methodology, eleven platform / engine / industry pages, and free tools. New visitors usually have one of five jobs in mind. This page routes you to the right place based on what you're trying to do, plus three caveats everyone should know before going deeper.

Last reviewed 2026-05-10. Updated as new studies and pages ship.

30-second orientation

Mainland Chinese consumers ask AI assistants — DeepSeek, Qwen (Alibaba's Tongyi) and Doubao (ByteDance) — for product recommendations, comparisons and how-to advice in roughly the same way Western consumers ask ChatGPT, Claude or Perplexity. The recommendations differ. In our 540-call May 2026 panel, top-15 cited-source overlap (Jaccard) between any two Chinese engines was 0.20–0.30 — and overlap with Western engines was lower still. A brand strong on ChatGPT and Gemini can be effectively invisible on DeepSeek and Qwen, and vice versa.

China AI visibility is the discipline of measuring and improving how these Mainland engines surface, cite and recommend a brand. It overlaps with GEO at the framework level but uses different platforms (百度百科 not Wikipedia, 知乎 not Reddit, 小红书 not YouTube), different language, and different methodology. The pillar reference is the longest single page on this — 2,500 words on what it is, why it matters and how it differs.

Are you here because…

Path A — You're a brand marketer who hasn't thought about Chinese AI engines yet

Recommended order:

  1. China AI visibility for global brands — the pillar. What this discipline is, why it matters, how it differs from generic AI visibility.
  2. DeepSeek vs Qwen vs Doubao: Why Brands Look Different — the headline 540-call study showing engines share only 20–30% of their cited sources.
  3. Luxury AI Visibility in China — The 5 Cited Sources — concrete category example showing which third-party sources matter at which price tier.
  4. Run the free audit on your domain. The audit returns per-engine selection, absorption and brand-mention scores against a stratified zh-CN consumer prompt panel.

Path B — You're an SEO or GEO practitioner

Recommended order:

  1. Generative engine optimization for China — how the GEO framework adapts when the engines are DeepSeek, Qwen and Doubao instead of ChatGPT and Claude.
  2. Per-engine playbooks — DeepSeek SEO, Qwen optimization, Doubao optimization. What predicts citation rate on each, where they differ.
  3. Eastbound's measurement methodology — two-stage citation framework, stratified prompt panels, per-engine reliability stats (top-5 + top-15 κ).
  4. AI crawler readiness + llms.txt vs robots.txt — infrastructure work that applies to Chinese engines too.
  5. Off-site substrate study — why traditional SEO won't get you into Chinese AI answers and what does.

Path C — You're a decision-maker evaluating whether this matters for your business

Recommended order:

  1. The business case for China AI visibility — what is and isn't measurable today, three concrete buyer scenarios, what spending typically looks like.
  2. China GEO agency services — what an Eastbound engagement covers and how we work.
  3. Methodology — the measurement discipline. Decision-makers should read this to understand what gets reported back.
  4. Book a 30-minute fit check — we'll tell you whether we're the right fit and if not, we'll say so.

Path D — You're a researcher or journalist looking for citation-worthy material

Recommended order:

  1. Eastbound's measurement methodology — read this first. The disclosure section flags which claims are measured, hypothesised, or planned-test.
  2. The five published studies: three Chinese AIs source-attribution, luxury AI sources, off-site substrate, risk-management un-pivotable brands, awareness-vs-decision intent shift.
  3. Insights drilldowns: Wikipedia at 21% on DeepSeek, YouTube at 20% despite the geo-block, SMZDM tier collapse, Zhihu, Xiaohongshu.
  4. For data requests, custom cuts on a vertical, or quotes attributed to "Eastbound Research" — email info@eastbound.ai.

Path E — You're a competitor or fellow agency benchmarking us

Recommended order:

  1. Competitor AI visibility audit guide — our public methodology for benchmarking competitors on AI engines.
  2. AI visibility tools comparison — honest comparison including our position vs Profound / AthenaHQ / Otterly / Goodie.
  3. Profound alternatives — where we fit (China specialty + free audit) vs where Profound wins (US-engine dashboard at scale).
  4. Run the free audit on your own domain. We don't restrict it.

Three caveats every reader should know

1. China is a separate generative-search surface — not a translation of the Western one

The Western GEO frameworks (Aggarwal KDD 2024, Profound, AthenaHQ) all default to ChatGPT / Claude / Perplexity / Gemini. The source ecosystems Mainland Chinese engines draw from differ by 70–80% from those Western engines — different platforms, different language, different consumer prompt patterns. Don't expect a Western GEO playbook to port. See AI search optimization for China for the framework adaptations.

2. Selection ≠ absorption ≠ user-visible mention

Per the two-stage measurement framework (Aggarwal KDD 2024 + Yao Jingang 2026), a page can be selected — pulled into the engine's source pool — without being absorbed (its language actually shaping the answer the user reads), and absorbed without being mentioned (your brand name surfacing in the visible output). Tw93's 2026 instrumentation showed ChatGPT retrieves ~100 pages per query but only ~15% surface in the answer. Three different metrics. Different optimisation tactics for each. Our methodology page details how we measure all three.

3. Off-site source-graph work is the slowest but most-compounding layer

Aggarwal et al.'s KDD 2024 benchmark found brands cited by third parties were referenced roughly 6.5× more often than brands cited only on their own domain. Technical infrastructure is a 1-hour layer. Content design is a multi-week layer. Off-site source-graph work — Wikipedia, Reddit, ArXiv for Western; 百度百科, 知乎, 小红书, SMZDM for China — is a multi-quarter layer that compounds. Plan accordingly.

The fastest first step is the free audit

If you only do one thing, run the audit on your domain. It returns per-engine selection, absorption and brand-mention scores against a stratified zh-CN consumer prompt panel. From there the path forward is concrete.

Run AI visibility audit  or  book a 30-minute fit check.