The AIVO Standard Methodology, Version 4.1: From Visibility to the Decision

What changed and why

Version 3.5 of the AIVO Standard asked how a brand becomes discoverable by AI assistants. Measurement since then has established that discoverability is not the binding constraint. Recognition at the opening question runs high enough that possession of knowledge poorly explains who is eventually chosen. A large majority of the facts an assistant demonstrably holds go unused at the point of recommendation. A subject named at the start of a conversation is frequently displaced before that conversation reaches a decision.

Those findings carry their sample sizes and intervals in the sources cited under Supporting Evidence in the full document. They are properties of the corpora measured, not constants restated here.

Version 4.1 is a restructure rather than an addendum. The nine foundational stages from version 3.5 are retained in full and moved to an Annex, reclassified as prerequisite hygiene beneath a diagnostic layer that now governs. Each stage is tested against one question: what does it move at the decision.

Structure

Part I, Diagnosis, specifies probe conditions (open and closed book, generic and anchored, API and consumer application, machine and human execution, research simulation and agentic), the assistant set and its weighting, what counts as a multi-turn decision conversation, the annotation layer and its validation requirement, the buying journey probe, the reasoning chain score, displacement initiation, AI Win-Rate, the AI Recommendation Quotient, the Linkage Gap in both its measurements, the AI Reputation Quotient, the Role of Brand Probe, the Ownership Verdict Probe, the Suppression Gap, run integrity, and reporting requirements.

Part II, Substrate, covers legibility rather than volume, the context atom and its required components, knowledge architecture, brand.context, and the attribution convention.

Part III, Remediation, specifies the loop from diagnosis through authoring, authorisation, publication and re-probe, the two gap modes, the authorisation conditions governing publication on a subject's behalf, the permanent exclusion of political subjects from remediation, and the conditions a remediation claim must satisfy.

Part VI, Attribution, sets out the boundaries an attribution must cross and what may be claimed where one is open.

Part VII, Governance, covers limitations, reproduction, contestability, versioning and change control.

Scope

The document specifies method only. It carries 53 numbered rules and a conformance clause stating the conditions under which a measurement, and the report carrying it, are conformant. It describes no particular implementation, platform or vendor, and does not restate the measurement results of any individual practitioner.

Changes from version 4.0

Parts I, II, III, VI and VII were rewritten to specify method only. Probe conditions, the assistant set and weighting, the definition of a multi-turn decision conversation, the annotation layer and its validation requirement, run integrity and conformance are newly specified. Descriptions of any particular implementation, and the measurement results of any individual practitioner, have been removed. Two corrections are made: the platform weighting example published in version 4.0 named the wrong assistant, and a category figure quoted in version 4.0 has been superseded by a verified recount.

Version 4.1 supersedes the diagnostic framing of version 3.5.

DOI-anchored, not peer-reviewed.

DOI: 10.5281/zenodo.22708359 Published: September 11, 2026 Authors: Paul Sheals, Tim de Rosen