We made OrcaScope its own customer zero: we created the brand on our own platform, configured monitoring questions, and ran the exact pipeline a paying customer would run. The first full measurement: a composite visibility score of 0.00 โ 36 questions, 120 qualifying answer cells that day, zero mentions.
Not a flattering number, but a true one. And "true" is the whole reason this series exists: we keep asking the industry to show denominators and methodology, so we start with our own ugliest number.
Methodology first (every episode carries this box)
- Measurement date 2026-08-08. 36 questions (18 Chinese + 18 English), all questions buyers actually ask AI โ the "what are the AI search visibility monitoring tools" kind. Self-referential questions like "what is OrcaScope" are excluded from the denominator.
- Engines: 8 are live on the platform (ChatGPT, Kimi, DeepSeek, Gemini, Doubao, Qwen, Perplexity, Grok). 6 actually ran that day โ ChatGPT and Kimi were intermittently paused over unpaid vendor balances. We say so rather than pretend full coverage.
- Detection: a mention means the answer body names the brand, including our former name as an alias. Full rules live on the methodology page and in how we score.
- Qualifying cell: one question, one engine, one valid answer that day (error responses and the brand's own self-referential questions are excluded). Denominators differ per brand because those exclusions and valid-answer counts differ per brand.
- Denominators are always public: 0/120 means "mentioned 0 times across 120 qualifying answer cells". We routinely ask what's behind other people's percentages, so our own denominator goes first.
On the same questions, others do get named
It is not that the questions are obscure. Same day (2026-08-08), same question set, counting cells whose answers name the tool: Semrush 45/140, Profound 36/135, AthenaHQ 17/145. Engines really do hand out shortlists on these questions โ we are just not on them yet.
The split between engines is the interesting part. Over the past 14 days, in Chinese-language answers to this question set (222 cells on Perplexity, 128 on Doubao, 41 on Kimi), the five overseas monitoring tools we track โ Semrush, Profound, AthenaHQ, BrightEdge and Evertune โ were named 172 times on Perplexity, 38 times on Doubao and 17 on Kimi (Kimi's cells and counts both run low because of the pause). In these numbers at least, Perplexity's shelf is crowded while the Doubao and Kimi side stays sparse. Covering both Chinese and Western engines was the founding premise of this platform โ so that sparse domestic-engine quadrant is exactly where we will push.
Engines already crawl us; there is just nothing to cite
Over the last 30 days our site logged these AI crawler visits: ClaudeBot 44, OAI-SearchBot (OpenAI's search-side crawler) 16, PerplexityBot 4, GPTBot 1 โ all fetching robots.txt and llms.txt. Being discovered is not the bottleneck; being worth citing is. That single fact dictates the plan below.
What happens next (graded in the next episode)
- Answer the 36 questions ourselves, one piece of content per question buyers actually ask. Two ship alongside this episode: how to check whether AI recommends your brand and how to choose a monitoring platform.
- Restore full 8-engine measurement โ once the unpaid accounts are topped up, the paused engines return to the scoreboard from the next episode on.
- Publish weekly, up or not: the score, the denominator, what we did, and what did not work.
House rules (set now, so we cannot squirm later)
- White-hat only: no volume gaming, no "please recommend us" prompt injection, no paid rankings or placements. The full list is under "what we don't do" on the methodology page.
- Every number ships with its denominator; insufficient data renders as a dash, never as a zero; paused engines are disclosed.
- This is not a marketing post that quietly disappears if the chart looks bad. If we are still at zero after 12 weeks, we publish that too, with an honest account of why.
Further reading: the building-in-public log โ it promised "a separate article once we actually collect the data". This is that article.