LLM Observation Data · Mainland China

que·rent (n.) — one who puts a question to an oracle.

See what China's AI assistants tell 400 million people about your brand.

Q: What is Querent?

Querent is an observation-data provider that samples how mainland China's domestic AI assistants — Doubao, Tencent Yuanbao, Qwen and Kimi — answer consumer questions, then delivers the results as structured, screenshot-backed JSON with English translations.

Q: Is the data personal or aggregate?

It is aggregate, non-personal answer data. Every data point carries a screenshot audit trail, and nothing ties back to an individual end user.

Q: How do I get a sample?

Send one brand and up to five prompts to alex@querentdata.com and Querent returns real data within 48 hours, with no strings attached.

Doubao, Tencent Yuanbao, Qwen and Kimi answer more consumer questions than ChatGPT — behind +86 real-name registration no Western visibility platform can cross. We deliver their answers as structured, sampled, screenshot-backed JSON.

Get free sample data → See the data schema

Send one brand + 5 prompts · real data back within 48 hours · no strings

observation · doubao · zh-CN● sampled 3×

{
  "platform": "doubao",
  "prompt": "跑鞋什么牌子好",
  "samples": 3,
  "aggregate": {
    "entities": [
      { "canonical": "Li-Ning",
        "mention_rate": 1.0, "avg_position": 1.3 },
      { "canonical": "Nike",
        "mention_rate": 0.67, "avg_position": 2.5 }
    ],
    "citation_domains": [
      { "domain": "zhihu.com", "rate": 0.67 }
    ],
    "volatility": 0.41
  },
  "answer_translated": "For running shoes…",
  "snapshot_id": "snp_1a2b3c"  // full-page screenshot
}

Querent is structured observation data from mainland China's domestic AI assistants — Doubao, Tencent Yuanbao, Qwen and Kimi — sampled, translated, and screenshot-backed, so you can see exactly how the answer engines behind the Great Firewall describe your brand, your prices, and your competitors.

240M

Doubao monthly active users — larger than ChatGPT

60%

of urban Chinese consumers use AI for pre-purchase research

5

domestic AI platforms covered, sampled & parsed to one schema

0

Western visibility platforms covering any of them

The gap

Your customers ask "what about China?" — and today the honest answer is nothing.

Every AI-visibility stack tracks ChatGPT, Gemini and Perplexity. None reach the platforms Chinese consumers actually use, because the barrier isn't parsing — it's operations on the ground.

+86 real-name accounts

Doubao, Yuanbao, Qwen and Kimi require mainland phone registration under real identity. Account fleets can't be spun up from abroad.

Mainland infrastructure

Residential IPs, domestic payment rails, and UIs that change weekly — maintained by a team that reads them natively.

Chinese-language parsing

Brand entity normalization across 耐克 / Nike / 耐克官方, citation extraction from Zhihu and Xiaohongshu, translated for your analysts.

How it works

Tracked prompts in, sampled observations out.

LLM answers are non-deterministic — a single response is an anecdote. We sample every prompt multiple times per run and ship the distribution: mention rates, average positions, volatility.

Define trackers

Your prompts, platforms, schedule and target entities — one API call or a shared sheet during pilot.

We sample & parse

Clean logged-in sessions, multiple samples per prompt, full-page screenshots for every data point.

You get JSON

One normalized schema across all platforms, with English translations. Weekly drops now, hosted API next.

What's in every drop

One schema, every platform — built for analysts and answer engines alike.

Each observation is a sampled distribution, not a single screenshot. Every field is normalized across platforms and carries an audit trail back to the source.

entities[]

Normalized brand mentions

Every brand collapsed to one canonical entity across 耐克 / Nike / 耐克官方, with mention rate and average position.

citation_domains[]

Where the answer sourced from

The domains each assistant cited — Zhihu, Xiaohongshu, official sites — and how often they appeared.

volatility

Answer stability score

How much the answer shifts sample-to-sample, so you know which findings are stable and which are noise.

answer_translated

English translation

The full assistant answer rendered in English, ready to drop into a report for a non-Chinese-reading team.

snapshot_id

Screenshot audit trail

A full-page screenshot behind every single data point — every claim is verifiable, nothing is asserted blind.

platform · prompt

Consistent keys

The same schema whether the answer came from Doubao, Yuanbao, Qwen or Kimi — join and compare without reshaping.

Coverage

The platforms behind the wall.

Platform	Operator	Scale	Why it's unreachable from abroad	Status
Doubao豆包	ByteDance	240M MAU	+86 real-name account, CN payment	design-partner pilot
Yuanbao元宝	Tencent	WeChat-scale reach	+86 account, WeChat ecosystem	design-partner pilot
Qwen通义千问	Alibaba	Top-3 domestic	+86 account	design-partner pilot
Kimi	Moonshot AI	Leading in research queries	+86 account	design-partner pilot
DeepSeek	DeepSeek	180M MAU	Globally accessible — included for completeness	available
Baidu AI Search	Baidu	Search-scale	CN infrastructure	roadmap

Who it's for

Built to slot into your product, not compete with it.

AI visibility platforms

OEM data source

Add a China module your enterprise customers keep asking about — white-label JSON that matches your existing engine schema. We stay invisible.

China-market agencies

Client reporting engine

Replace manual Doubao screenshots with repeatable weekly sampling. Report-ready data with an audit trail behind every claim.

Global brands

Direct monitoring

Track how mainland AI assistants describe you, your prices, and your competitors — translated, structured, and archived.

FAQ

Questions we get from data teams.

What is Querent?

Querent is an observation-data provider that samples how mainland China's domestic AI assistants — Doubao, Tencent Yuanbao, Qwen and Kimi — answer consumer questions, then delivers the results as structured, screenshot-backed JSON with English translations.

Which Chinese AI platforms does Querent cover?

Doubao (ByteDance), Yuanbao (Tencent), Qwen (Alibaba) and Kimi (Moonshot AI) are covered in the current design-partner pilot. DeepSeek is available today, and Baidu AI Search is on the roadmap. Every platform is normalized to the same schema.

Why can't Western AI-visibility tools track Chinese AI assistants?

The barrier is operational, not technical. These platforms require +86 mainland phone numbers under real-name registration, domestic residential IPs and Chinese payment rails, and their interfaces change weekly. Account fleets can't be spun up from abroad, so no Western visibility platform reaches them.

What format does Querent deliver data in?

One normalized JSON schema across every platform: the prompt, sample count, aggregate entity mention-rates and average positions, cited source domains, an answer volatility score, an English translation of the answer, and a full-page screenshot ID for every data point.

How does Querent handle the non-determinism of AI answers?

A single AI answer is an anecdote. Querent samples every prompt multiple times per run and ships the distribution — mention rates, average positions and volatility — rather than one screenshot.

Is the data personal or aggregate?

It is aggregate, non-personal answer data. Every data point carries a screenshot audit trail, and nothing ties back to an individual end user.

How do I get a sample?

Send one brand and up to five prompts to [email protected] and we return real data within 48 hours, with no strings attached.

Design partners · Q3 2026

We're onboarding three design partners this quarter.

Locked early pricing, direct input on the schema, and weekly data drops from day one — the hosted API ships within weeks, backfilled with your pilot data.

Talk to us →

Or start smaller: one brand + 5 prompts → real data in 48h.