Don't guess — you already built the tools to know
Every module above produces a number you can actually check, and you already have (or nearly have) the instruments for three of the four layers.
| Layer | How to measure it | Tool |
|---|---|---|
| Crawl | Indexed page count, crawl errors | Google Search Console (free) |
| Cite | Baseline query re-test delta; GA4's "AI Assistant" channel | GA4 + the Module 02 baseline spreadsheet |
| Discover / Social | Signups, funnel conversion, real revenue and cost | Your own Business Monitor dashboard |
| Uptime | Is the product even reachable while you're driving traffic to it | The JobPlnr Watchdog agent |
One honest caveat on GA4's AI-referral tracking: some AI assistants strip the referrer entirely, so a real share of AI-driven visits will land in "Direct" traffic and undercount the true total. Treat the number as a floor, not a ceiling.
The Business Monitor's funnel (signup → resume analysis → job search → DreamJob Prep → Apply Assist → paid) is your ground truth for whether any of Modules 01-04 are actually working. A spike in "AI Assistant" referral traffic that doesn't show up as new signups means the citation worked but the landing experience didn't convert — a different problem than not being cited at all.
Deep Dive GA4 channel grouping, attribution limits, and reading your own funnel correctly
Setting up a custom channel group for AI referrals
GA4's default "AI Assistant" channel picks up ChatGPT, Gemini, and Copilot by default, but as noted, Perplexity and Claude referrals sometimes arrive with different or stripped referrer patterns. In GA4 → Admin → Channel Groups, you can build a custom regex-based rule matching source/medium against known AI referral patterns (e.g. matching perplexity, claude, or ai-assistant as a medium) so these aren't silently absorbed into "Direct" or "Referral" traffic. Worth setting up once, then not touching again.
Why attribution here is inherently imperfect — and why that's fine
No analytics setup gives you a fully accurate picture of AI-driven discovery, because a meaningful share of the effect is zero-click: someone reads a ChatGPT answer that mentions JobPlnr, remembers the name, and later types jobplnr.com directly — which GA4 will record as "Direct" traffic with no way to trace it back to the citation that actually caused it. This is a known, structural limitation across the whole industry right now, not something your setup is doing wrong. The practical implication: treat the Module 02 baseline query test (are you actually being cited, checked by hand) as your primary signal for whether GEO is working, and GA4's AI referral numbers as a supporting, undercounted data point — not the other way around.
Reading the Business Monitor funnel like a diagnostic, not a scoreboard
The funnel's real value is in the ratios between stages, not the absolute numbers. If signups are rising but the signup → resume-analysis-completed ratio is flat or falling, that's a product/onboarding friction problem, not a marketing problem — no amount of additional Module 03/04 work fixes a leaky step later in the funnel. Conversely, if every ratio downstream of signup is healthy but signups themselves are flat, that confirms the problem is genuinely at the top of the funnel (Modules 01-03), and effort belongs there. Check the funnel ratios, not just the raw signup count, before deciding where to spend the next week's effort.
A simple weekly review habit
Once real traffic exists: a 15-minute weekly check of (1) Search Console's indexed-page trend, (2) the Business Monitor's funnel ratios week-over-week, and (3) whether the Watchdog has fired any alerts, catches the large majority of problems worth catching. Resist the urge to build more elaborate tracking before this simple habit is actually happening consistently — a dashboard nobody checks is worse than a simpler one that gets looked at every week.