How AIOTruth decides

Evaluation methodology

Each evaluation type publishes its inputs, retrieval method, observable signals, weighting, not applicable logic, failure states, the evidence it shows, its scoring or status method, its version, its change history, its known limitations, what it does not measure, its data retention, and its security controls.

• Trust Visibility (Live). Evaluates the observable signals a business publishes across clarity, entity consistency, authority, AI discoverability, trust, and local presence.
• Content Credibility (Beta). Evaluates whether a page is readable, attributable, sourced, specific, original, and honestly maintained. It does not claim to prove whether a human or an AI wrote the page.
• WebMCP Readiness (Beta, Early Preview). Inspects the current evidence that a website exposes structured tools for browser agents. Absence is not a current failure, and presence is not a security certification.

The exact dimension weighting for a live type is published on the AIOInsights methodology page, applied identically to every domain, and stamped with the engine version that produced each result.

Research method

Separately from evaluating a business, AIOTruth keeps a research record of the developments shaping the field. That record is scored by the deterministic Research Value Score, which is a different measurement from any evaluation score. The full research method, including discovery, source verification, the evidence labels, and the score, is published on the Editorial Method page.

The rule that does not move

A threshold is never lowered to produce more output, and the engine changes only to be more accurate. When it changes, the change is recorded on the changelog.