---
title: "Witnessability Guides: Evidence, Verification and Claims"
description: "Read practical guides on verification, appraisal, signed logs, audit trail completeness and how witnessability relates to observability."
canonical: "https://witnessability.io/guides/"
representation: "Website reading format; consult linked publications for their exact status and terms."
---

Guides / Learn through questions

# What can this evidence establish?

Start with an ordinary record. Separate the claim, its grounds and the decision that follows.

Guide / Missing evidence

## [What does a missing audit record mean?](https://witnessability.io/guides/audit-trail-completeness/)

“No matching record was found” describes a search result. To use it as evidence that an event did not occur, you need additional grounds about what should have been recorded and what the search covered.

Comparison / Related concepts

## [Observability and witnessability ask different questions.](https://witnessability.io/comparisons/observability-vs-witnessability/)

Telemetry can help explain a system's behavior. Witnessability asks how available materials and arguments support a particular proposition under explicit conditions.

Examples / Three ordinary records

## [One record. Several possible claims.](https://witnessability.io/examples/)

These teaching examples separate what is observed, what the evidence can support and what needs an additional basis. They are illustrative cases, not reports of tests run on a real deployment.

Start here

## [What is Witnessability?](https://witnessability.io/what-is-witnessability/)

Witnessability concerns our ability to produce, preserve and make available the evidence and reasoning needed to assess a specific claim. It starts with a question: what can someone establish about something they did not observe themselves?

Guide / Three different questions

## [Verification, appraisal and reliance](https://witnessability.io/guides/verification-appraisal-and-reliance/)

A record can pass a technical check while leaving the claim you care about unresolved. A supported claim can still be insufficient for a decision. Keeping these questions separate makes the result more useful.

Guide / Execution evidence

## [A tool call is not yet an execution finding](https://witnessability.io/guides/tool-call-vs-execution/)

A request, an execution record and an observed effect answer different questions. To understand an AI agent or software workflow, identify the exact boundary behind each piece of evidence.

Reference / Plain-language terms

## [A vocabulary for precise conclusions](https://witnessability.io/glossary/)

These short explanations help you read the site. Definitions in a cited specification apply within that document's own scope and version. This glossary does not replace them or create a new standard.