What's in it
Started in 2026. The first wave covers 8 engines; Avowd now measures 11, and new waves are added here as they are published.
Client measurements are never sold or shown. The panel's questions never contain a client's name.
Open data — free
Aggregate tables, CC BY 4.0. Use them, quote them, credit Avowd.
| Chinese-brand share | EN | 中文 |
|---|---|---|
| ChatGPT | 2.2% | 18% |
| Perplexity | 2.5% | 31.7% |
| Claude | 0% | 31.3% |
| Grok | 0% | 22.9% |
| Doubao 豆包 | 2% | 40.5% |
| Qwen 通义 | 1% | 23.8% |
| DeepSeek | 2% | 31.3% |
| Kimi | 0% | 26.7% |
| Most-cited sites · Western engines | Answers |
|---|---|
| cnblogs.com | 27 |
| blog.csdn.net | 27 |
| learn.g2.com | 21 |
| zoho.com.cn | 18 |
| youtube.com | 17 |
| zhuanlan.zhihu.com | 16 |
| cloud.tencent.com | 14 |
| reddit.com | 14 |
Wave 2026-07. Chinese engines returned no citation URLs, so they have no domain column.
Three ways in
Tier 1Open data
Aggregate tables for every wave, CSV and JSON.
Details
- Licence: CC BY 4.0 — credit Avowd.
- Aggregates only — no single answer text.
Free · CC BY 4.0
Download ↑
Tier 2Licensed dataset
Every row of the panel — full answer text included — for internal use.
Details
- One row per engine × question × language × repeat: full answer text, brands named, cited links, answer length, errors.
- Internal use; no resale or republication of rows.
- One wave so far (2026-07). New waves added monthly from November 2026.
Building — reserve now
Reserve · info@avowd.ai →
Tier 3Custom panel
Your categories, cities and questions — measured for you only.
Details
- Up to all 11 engines, English and Chinese.
- Exclusive: never added to the open data.
- Open panel: one wave so far (2026-07), monthly from November 2026.
Building — reserve now
Reserve · info@avowd.ai →How to cite
Avowd (2026). AI-answer research panel, wave 2026-07. avowd.ai/data. CC BY 4.0.
How questions are chosen and answers scored: avowd.ai/method. Questions: info@avowd.ai.