AigeoRadar Lab
AIPM Benchmark
The same AI model. The same page-derived questions. Three contexts: raw HTML, JSON-LD schema, and AIPM (.ai.json). We measure what actually burns AI budget: accuracy, latency, and tokens.
Accuracy
Gold answers come from AIPM fields. We score how often each context retrieves the fact.
Time
Wall-clock median and p95 latency per answer — not marketing fiction.
Tokens
Input + output tokens and estimated USD cost for the full question pack.
Latest public report
The Best GEO WordPress Plugin in 2026
AIPM won this run — accuracy +55.5 pts vs HTML, -51.6% fewer tokens, -30.7% faster median latency.
1/10
HTML matched
3/10
Schema matched
8/10
AIPM matched
Fair by design
- Questions are sampled from each page’s AIPM — a plumber site never gets SaaS quiz questions.
- HTML mode strips
.ai.jsonhints and JSON-LD so the model cannot fetch the sidecar. - Works for AigeoRadar blogs and customer WordPress installs alike.
Spec: AI Page Manifest (AIPM)