Our public baseline
We publish completed observations and their limits. The figures below reproduce the historical Round 0 record from 12 November 2026; this website update did not rerun those measurements.
Scope reported in the historical summary
The previous website reported 44 samples across 4 engines on 12 Nov 2026, mostly one sample per question and engine, with a small number reaching three. We retain the original reporting scope below. The underlying records await verification; these figures are not currently independently verified evidence of impact.
| Measure and denominator | Round 0 record |
|---|---|
| Generic questions without the brand name (26 samples) | No recommendation mentions of ProofReach were recorded. |
| Citations of proofreach.cn (44 samples) | No sample cited this website. |
| Direct brand questions (18 samples) | Gemini: 5/5 identified correctly; ChatGPT: 5/5 identified incorrectly; Perplexity: mixed results; Tongyi: insufficient information. |
Coverage and gaps
| Platform | Completed samples | Round notes |
|---|---|---|
| Perplexity | 23 | Sampled after sign-in |
| ChatGPT | 12 | Historical record |
| Gemini | 6 | Historical record |
| Tongyi | 3 | Historical record |
| Claude / DeepSeek / Kimi | Not measured | Sign-in limits at the time |
| Doubao | Not measured | Regional limits at the time |
Initial observation: brand-name confusion
The previous summary reported that some ChatGPT and Perplexity answers identified ProofReach as another similarly named product, while relevant correct identification by Gemini used the DOO website. It reported no citations of proofreach.cn among those 44 samples. This is not a new test of proofreach.hk, nor proof that no engine has ever crawled or cited the brand.
The planned method is not the completed coverage
The original plan was 20 questions × 8 engines × 3 samples, or 480 sample slots. Round 0 did not achieve that coverage. Platforms and repetitions should reflect each project’s markets and goals. Planned coverage, valid answers, failures and unmeasured slots must be recorded separately.
How to make follow-up observations comparable
- Report direct brand recognition separately from discovery questions that do not name the brand.
- Keep consumer website or app observations separate from API model tests. API results do not fill missing website samples.
- Record question versions, platform entry points, languages, known regional conditions, browsing modes, time, original answers, citations and assessment reasons.
- Document changed conditions. Report non-comparable rounds separately and do not attribute random variation directly to content changes.
This page preserves a summary from the previous website. The complete questions and original answer records await verification, so the figures are not independently reproducible at present and are not proof of impact. Future traceable records will distinguish verified, unverified and non-comparable findings without reconstructing or altering historical data.
One issue to test in the next round
For identity confusion, align official facts and public pages, then repeat the same questions on the same platforms and modes. Report improvement, no improvement or non-comparability without assuming the outcome.
Start with a real question about your brand.
Tell us about your products, markets and the issue you want to resolve.
Select ProofReach on the contact form.