The Information Verification Field Manual
Evaluate claims, websites, screenshots, charts, experts, search results, and AI answers with a repeatable process that separates presentation quality from evidence quality.
1 of 4
Turn verification from a feeling into a repeatable process.
Break information into exact claims, identify what evidence would support them, investigate sources from outside their own presentation, and match confidence to the quality and independence of the evidence.
Investigate from the outside.
Use lateral reading to understand who is behind a source, what incentives may exist, and what independent sources say before accepting self-description as proof.
Trace claims to original material.
Follow consequential assertions to studies, datasets, documents, transcripts, filings, or other primary evidence when available.
Act proportionately to uncertainty.
Replace binary reactions with explicit confidence levels tied to evidence quality, source independence, missing context, and the cost of being wrong.
A field system for modern information environments.
Claim and source analysis
Decompose arguments into factual, causal, interpretive, predictive, and recommendation claims, then identify the evidence each type actually requires.
Lateral reading and context reconstruction
Investigate unfamiliar sources, restore context around screenshots and quotes, inspect charts and numbers, and trace citation chains backward.
AI verification protocol
Check AI-generated citations, calculations, summaries, and consequential claims outside the model before treating fluent output as resolved fact.
Search results and screenshots are evidence maps, not verdicts.
A persuasive crop, polished snippet, credential, or AI answer can point toward useful evidence without being the evidence itself. Trace the claim until the support is visible.
Define. Investigate. Trace. Verify. Calibrate.
State the exact claim, determine what evidence would support it, investigate the source laterally, trace important assertions to primary material, restore missing context, and assign a confidence level before acting or sharing.
