GeoCheckTool
China AI measurement hub

China AI Visibility

China AI visibility is the measurement of how China's AI assistants — including Doubao, Qwen (Tongyi), Tencent Yuanbao, and Baidu's Ernie — describe, compare, and recommend brands when Chinese consumers ask. Western AI-visibility suites generally do not provide verified access to these mainland consumer clients. GeoCheckTool measures the supported routes directly and publishes its methods, evidence boundaries, and open research data.

Coverage is engine-specific

Doubao is verified on real app and web clients; Qwen and Yuanbao are available as verified second-engine scopes; Ernie telemetry is directional; DeepSeek mainland work is custom research; Kimi is not currently measured.

Product identity

GeoCheckTool is the China AI search measurement product of Super Language (Beijing) Technology Co., Ltd., measuring what China's AI assistants actually say about brands — on real consumer surfaces, with published methods and open data.

Published evidence · bounded scope

Doubao ranked first among AI-native apps in QuestMobile's June 2026 China mobile-internet report with 382 million monthly active users; review the source report. This audience figure describes Doubao's reported June 2026 app reach, not the reach or performance of GeoCheckTool.

The mainland consumer surfaces Western dashboards do not verify

Mainstream international visibility platforms focus on ChatGPT, Gemini, Perplexity, and other internationally accessible routes. Even when a platform names a Chinese-origin model, buyers should ask whether the measurement came from the mainland consumer client, an international app, or an API proxy.

GeoCheckTool does not merge those routes. Every result retains the engine, surface, login state, collection date, and evidence fidelity needed to understand what was actually observed.

A China AI visibility measurement

On a fixed date, GeoCheckTool runs frozen questions on the named consumer client or directional route and records the full answer, exposed sources, and evidence state. Verified work uses dedicated research accounts and screenshot evidence; directional telemetry is labeled separately.

In the published Doubao surface study, identical questions produced different source sets in 14 of 16 valid pairs and different displayed search terms in 15 of 16. Those counts are a bounded snapshot, not proof that every question or future run will differ.

What the published research found

In GeoCheckTool's flagship recommendation-question sample, 119 confirmed marketing-ecosystem citations appeared among 163 total citations, a 73.0%–75.5% sensitivity range. Across all citation types, the corresponding range was 51.3%–54.5%. The denominators, coding rules, and limitations are published with the report.

The companion beauty, luxury, and automotive note shows why entity resolution matters: assistants can name products, marketplaces, authentication services, and brands within the same buyer answer. A brand mention cannot be interpreted correctly without checking which entity the answer actually recommends.

Start with research, a snapshot, or a verified audit

The research and open data are free. A $390 China Evidence Snapshot measures five frozen questions and credits that fee toward an audit booked within 30 days. A verified audit expands the scope to twenty frozen questions and per-question evidence packets.

  • Read the research for methods, open data, and limitations.
  • Use the Snapshot for five named real-client questions about one brand.
  • Use the audit for a broader question bank, citation analysis, and record-layer gap assessment.

Independent research boundary

We publish independent research on Chinese AI search. Paid clients are permanently excluded from our public research samples. Measurement cannot be purchased; our reports state what the engine said, favorable or not.