Wang Bo is the founder of GeoCheckTool, the China AI search measurement firm. From Beijing, on real consumer clients and a pinned API layer, his team measures what China's most-used AI assistants actually answer when shoppers ask about brands — with frozen question banks, per-citation hand classification, published methods, and open data (CC BY 4.0).
Why hosts and editors book this topic: Western tooling structurally cannot see China's answer market (language, real-name accounts, network barriers). Every number below comes from published research with raw tables attached — nothing anecdotal.
Contact: [email protected] ｜ Research: geochecktool.com/research
Speaking topics
For brand-recommendation questions on China's most-used assistant, 73.0–75.5% of citations were confirmed marketing-ecosystem sources (163 citations); across all intents in the same collection, 51.3–54.5% (641 citations). Two denominators, never mixed.
In dietary supplements, 61.4% of answers named no identifiable candidate at all — the lowest occupancy of six industries measured. One of the names that keeps appearing is the regulator.
Across preregistered paired questions, candidate sets changed in 31 of 31 pairs, mean Jaccard overlap 0.026, and 0 of 31 pairs shared enough names to compare an ordering.
On matched pairs inside a 30-minute window, the web surface displayed more search terms in 15 of 16 pairs and more sources in 14 of 16 — one assistant, two answer paths.
In luxury, the most-named entity was a resale platform (9/70), ahead of Hermès and Chanel (6/70 each). In pet food, the leaderboard top is split between a domestic brand and SGS, the certifier, at 5/70 each.
Evidence
Recording
Beijing time zone; remote recording any tool; English; slides or raw tables available on request. Methods and limits stated on every study — API and consumer-app surfaces are never pooled.