// methodology · v2 · 5 layers
Every report runs the same 5-layer check. Scores are weighted, human-read, and traceable to a line of code or a live AI answer. No simulation.
M1 · 25%
12 AI bots vs robots.txt + live fetch as GPTBot + WAF untrusted flag. Door first, then house.
M2 · 30%
12 buyer questions × 9 intents fan-out × 3 engines, N=3 majority vote. Honest search-enabled endpoints only.
M3 · 20%
Top-10 domains + type (listicle/review/directory/media/community) + who cites competitor not you.
M4 · 15%
28-point GEO audit (15 structure + 13 AI-readiness). Same checklist live at /tools/geo-score.
M5 · 10% ★
≤7 ranked verdicts, each “do this → expect that” + copy-paste artifact. First one <30min.
api.openai.com)
·
Traceable to a line or a live answer
Dashboards give you another tab. We give you a verdict + to-do, traceable to robots.txt:14 or a live citation.
Search-enabled endpoints only. Not simulated memory — what buyers actually see.
Ranked by impact × effort. First fix is always a 30-min settings change, not a backlog.
robots.txt:14 line number + curl -A GPTBot wc -c byte ratio + WAF statuscitations[3].url + N=3 vote logllms.txt / schema / FAQ copy-pasteOne page · no fluff · raw AI answers · human-reviewed in 72h