Gate C — Measurement
Owner: Engineering + Marketing. Sign-off: written, jointly.
The 4 checks
Section titled “The 4 checks”| # | Check |
|---|---|
| C.1 | AI-referrer analytics instrumented and writing to durable store |
| C.2 | Citation tracker baselined — 50 prompts × 4 LLMs (ChatGPT, Claude, Perplexity, Gemini), zero-shot count recorded |
| C.3 | Framix scorecard dashboard accessible to client |
| C.4 | 90-day checkpoint scheduled; kill criteria documented and acknowledged |
Why C cannot be skipped
Section titled “Why C cannot be skipped”T=0 baseline is non-negotiable. Without it the 90-day post-launch checkpoint and kill criteria cannot evaluate change. See What We Don’t Do § 8.
v1.2 candidate — expand C.2 to 5 surfaces
Section titled “v1.2 candidate — expand C.2 to 5 surfaces”Decided 2026-06-03. C.2 currently lists four LLMs. Microsoft’s Bing AI Performance Report (Feb 2026, _sources/bing-position-on-seo-aeo-geo.md) gives per-URL citation tracking directly from the engine and has become a standard measurement surface in the wider industry. v1.2 will expand C.2 from 50 prompts × 4 LLMs to 50 prompts × 5 LLMs (ChatGPT, Claude, Perplexity, Gemini, Copilot/Bing), with Bing AI Performance Report instrumentation added to the gate. Engineering retains discretion to baseline Copilot/Bing through prompt sampling or direct Webmaster Tools export — both produce comparable T=0 counts.