Historical Day-1 dogfooding artifact; current production product matrix has changed.
This notebook is preserved for transparency. The current production matrix is:
Current commercial terms: Enterprise-first scope review; Professional scope review after requirements discussion; Snapshot 5 user queries × 8 endpoints = 40 datapoints through contact access while self-serve checkout is prepared.
The 9-endpoint experimental matrix and "0 / 270" framing on this page are historical, not the current production contract. See the current Methodology page.
Notebook 1 — Index
Data Provenance — who ran: Claude Code on local M3 Max workstation. Where: Singapore primary node, Hong Kong backup. When: 2026-06-11 09:00–09:08 UTC+8. License: CC-BY-4.0. Cite as: clarivy.ai/self-audit-01/nb1.
Headline score
| Metric | Value | Note |
|---|---|---|
| Brand mentions across the 9-endpoint × 30-prompt = 270 (endpoint, prompt) pairs | 0 / 270 | 0% — the expected Day-1 baseline for a brand-new domain with no content indexed, no inbound links, and no entity in any LLM knowledge graph. (The 7-endpoint production subset produced 210 datapoints, the 9-endpoint experimental matrix produced 270; both frames are preserved on this historical artifact page.) |
| Top citation source across the audit | tryprofound.com | Cited 1× by Perplexity Sonar on the competitive prompt "alternatives to Profound for跨境品牌" — i.e. an existing player named, not a clarivy citation |
| Second citation source | n/a | — |
| Endpoints returning at least 1 mention | 0 of 9 | — |
| Endpoints where the prompt set produced no relevant content (i.e. test broken) | 0 of 9 | Perplexity's "tryprofound" citation proves the prompt set is working; endpoints are not silent — they just do not know us |
Per-endpoint score table
| # | Endpoint | Provider ecosystem | Method | Mentions | Top citation |
|---|---|---|---|---|---|
| 1 | 豆包 Doubao | China-market | Volcengine Ark API · ZDR | 0 / 30 | — |
| 2 | Kimi | China-market | Moonshot API · ZDR | 0 / 30 | — |
| 3 | DeepSeek | China-market | DeepSeek API · no-retention | 0 / 30 | — |
| 4 | 文心一言 ERNIE | China-market | Qianfan API · enterprise | 0 / 30 | — |
| 5 | 秘塔 MetaSo | China-market search | Playwright UI capture | 0 / 30 | — |
| 6 | ChatGPT | Global frontier | OpenAI API · gpt-4o · ZDR | 0 / 30 | — |
| 7 | Perplexity Sonar | Global answer engine | Sonar API · online | 0 / 30 | tryprofound.com |
| 8 | Claude | Global frontier | Anthropic API · claude-3-5-sonnet · ZDR | 0 / 30 | — |
| 9 | Gemini + AI Overviews | Global frontier | Vertex AI · gemini-2.0-flash | 0 / 30 | — |
| Total | 9-endpoint historical matrix | 0 / 270 | tryprofound.com (1×) | ||
Why "0/270" is the correct answer, not a bug
- Clarivy.ai was registered on 2026-06-11 17:12 UTC+8; the audit ran the same morning at 09:00–09:08 UTC+8 — i.e. before the domain was even live on the public internet (the audit used a preview DNS slot).
- No content was indexed by any of the 7 endpoints' upstream crawlers (Google, Bing, Baidu, ByteDance Juliang, etc.) — the audit was run within minutes of registration, not weeks.
- No entity association has been established in any LLM knowledge graph (no Wikipedia entry, no Crunchbase record, no Wikidata QID, no schema.org/Organization on any crawled page).
- The prompt set does work: Perplexity's
tryprofound.comcitation on a competitive prompt confirms the test is surfacing real established players. It is surfacing them, not us — which is the right signal at Day 1.
Citation-source landscape (top 1 of 1)
| Rank | Domain cited | Endpoint that cited it | Prompt category | Prompt text (verbatim) |
|---|---|---|---|---|
| 1 | tryprofound.com | Perplexity Sonar | Competitive (B) | "alternatives to Profound for跨境品牌" |
What the next snapshot will measure
Self-audit #2 is scheduled for 2026-07-01 (first business day of next month). Between now and then we will execute the 5 prioritized actions in Notebook 5 — Strategy. The expected delta is small but non-zero: at least 1 China-market endpoint (most likely 秘塔 MetaSo, which weights freshly-indexed Chinese content heavily) is expected to begin citing clarivy.ai on a small subset of brand-specific prompts. We will report the actual delta in the same 5-notebook shape.
Limitations of this notebook
- Mention count is binary (yes/no per (endpoint, prompt)) — does not capture ranking position within a response. A future version will add position-weighted scoring.
- No citation-source rank weighting — tryprofound.com is counted as 1, regardless of whether it appeared in position 1 or position 5 of the cited sources.
- AI-search products are stochastic; results vary by ±10% on a re-run with the same prompts (Claude, Gemini, DeepSeek) and ±0% on endpoints run at low temperature (Doubao, ERNIE).