Wandered evidence protocol · public desk
Bring the claim. Keep the trace.
Six local instruments for interrogating AI traffic evidence. Each one exposes its input, names the test it applies, and leaves an inspectable result behind.
Local processing · documented rules · reproducible artifacts where supported
Protocol index / records 01—06
Choose the evidence you need to test
Every record below has its own input, test, and output. Portable artifacts move between tools only where a record explicitly says they do.
Acquisition evidence
GA4 AI Source Auditor
Interrogate source rows without silently assigning unknown traffic.
- InputGA4 source rows
- TestReferral registry
- OutputHash-bound summary
Identity evidence
AI Bot Verifier
Separate a self-declared token from a dated published-range match.
- InputUser-agent + optional IP
- TestIdentity / range snapshot
- OutputDated finding
Policy evidence
Robots AI Policy Simulator
Resolve the exact group, line, and rule governing a requested path.
- InputPolicy + token + path
- TestRFC 9309 precedence
- OutputLine-level verdict
Crawler evidence
AI Crawler Log Analyzer
Keep unknown requests visible while aggregating documented identities.
- InputServer access log
- TestCrawler / range registry
- OutputHash-bound summary
Evidence comparison
Crawl-to-Referral Matrix
Join only compatible summaries and keep causal claims outside the result.
- InputGA4 + crawler summaries
- TestSchema, hash, origin, coverage + period
- OutputExact-path matrix JSON / CSV
Planning evidence
AI Traffic Calculators
Make every price, denominator, formula, and uncertainty range explicit.
- InputManual values or crawler summary
- TestFormula / Wilson interval
- OutputScenario JSON
Methodology
One maintained evidence registry
Tool classifications come from a versioned registry with dated sources, explicit precedence, and stable rule identifiers. Results distinguish referral evidence, user-agent identity claims, published network-range matches, policy outcomes, and unknowns rather than collapsing them into one confidence label.
Limitations
Evidence is narrower than attribution
These tools do not prove that a crawler is authentic, that an AI system cited a page, or that a referral caused an outcome. Unknown and unsupported records stay visible so totals remain auditable.