MEASUREMENT STANDARD / V1.0
AEO Metric Dictionary
The CBI AEO Metric Dictionary defines 24 measures for AI visibility research. Every metric includes a calculation and a reporting guardrail so presence, recommendations and citations are not collapsed into a misleading vanity score.
MEASUREMENT STANDARD / V1.0
The Measurement System
24defined metrics
5core dimensions
1unknown state preserved
0synthetic results
MEASUREMENT STANDARD / V1.0
Core Measures
| Metric | What it measures | Critical guardrail |
|---|---|---|
| Presence rate | Eligible observations in which the entity appears | Publish entity rules and sample size |
| Recommendation rate | Explicit recommendations for the tested use case | Separate from incidental mention |
| Citation diversity | Distinct canonical domains or URLs in the sample | Report domain and URL views separately |
| Answer stability | Agreement across controlled repetitions | Not a guarantee of future behavior |
| Unknown rate | Fields that cannot be classified | Unknown is preferable to forced certainty |
MEASUREMENT STANDARD / V1.0
Composite-score Policy
Each dimension should be reported separately before any composite index is calculated. If a composite is used, its weights must be frozen before analysis and published with the result. A composite describes the defined sample; it does not establish universal brand quality or model preference.
MEASUREMENT STANDARD / V1.0
Complete AEO Metric Dictionary
All 24 AEO metric names, definitions, calculations and reporting guardrails are published below in the page HTML so people, search engines and AI systems can inspect the measurement standard without downloading a file.
24 records published online. Browse The Complete Resource
presence_rate
Presence rate
Definition: Share of eligible answer observations in which the defined entity appears.
Calculation: entity-present observations / eligible observations
Reporting Guardrail: Report the prompt corpus, entity rules and sample size.
Evidence Class: Methodological
Version: v1.0
shortlist_inclusion_rate
Shortlist inclusion rate
Definition: Share of eligible observations where the entity is presented as an option for consideration.
Calculation: shortlist observations / eligible observations
Reporting Guardrail: Separate from incidental mentions.
Evidence Class: Methodological
Version: v1.0
recommendation_rate
Recommendation rate
Definition: Share of eligible observations containing explicit recommendation language for the tested use case.
Calculation: recommended observations / eligible observations
Reporting Guardrail: Code the recommendation rule before collection.
Evidence Class: Methodological
Version: v1.0
conditional_recommendation_rate
Conditional recommendation rate
Definition: Share of recommendations tied to a stated condition, audience or use case.
Calculation: conditional recommendations / recommendations
Reporting Guardrail: Conditions must be captured verbatim or coded transparently.
Evidence Class: Methodological
Version: v1.0
caveat_rate
Caveat rate
Definition: Share of entity mentions accompanied by a limitation, risk or qualification.
Calculation: caveated entity mentions / entity mentions
Reporting Guardrail: A caveat is not automatically negative or inaccurate.
Evidence Class: Methodological
Version: v1.0
citation_rate
Citation rate
Definition: Share of observations containing at least one visible source reference.
Calculation: observations with citations / eligible observations
Reporting Guardrail: Define what counts as a visible citation on each surface.
Evidence Class: Methodological
Version: v1.0
citation_diversity
Citation diversity
Definition: Number of distinct canonical source domains or URLs in the sample.
Calculation: count of distinct canonical sources
Reporting Guardrail: Report domain- and URL-level views separately.
Evidence Class: Methodological
Version: v1.0
citation_concentration
Citation concentration
Definition: Degree to which citations cluster among a small number of sources.
Calculation: declared concentration index
Reporting Guardrail: Publish the formula and canonicalization rules.
Evidence Class: Methodological
Version: v1.0
first_party_citation_share
First-party citation share
Definition: Share of captured citations controlled by the entity being evaluated.
Calculation: first-party citations / all citations
Reporting Guardrail: Ownership classification must be dated and documented.
Evidence Class: Methodological
Version: v1.0
independent_citation_share
Independent citation share
Definition: Share of captured citations from sources classified as editorially independent.
Calculation: independent citations / all citations
Reporting Guardrail: Independence can be unknown; do not infer it.
Evidence Class: Methodological
Version: v1.0
source_freshness
Source freshness
Definition: Distribution of source publication or update age at collection time.
Calculation: age in days by source
Reporting Guardrail: Missing dates must remain missing.
Evidence Class: Methodological
Version: v1.0
answer_stability
Answer stability
Definition: Consistency of coded outcomes across controlled repetitions.
Calculation: agreement rate across repeated observations
Reporting Guardrail: Not a guarantee of future behavior.
Evidence Class: Methodological
Version: v1.0
entity_consistency
Entity consistency
Definition: Rate at which core facts remain consistent across observations.
Calculation: observations without fact conflicts / entity observations
Reporting Guardrail: Requires a verified fact reference set.
Evidence Class: Methodological
Version: v1.0
accuracy_issue_rate
Accuracy issue rate
Definition: Share of entity observations containing a fact that conflicts with the verified reference set.
Calculation: observations with verified errors / entity observations
Reporting Guardrail: Unknown or disputed facts should not be coded as errors.
Evidence Class: Methodological
Version: v1.0
source_supported_claim_rate
Source-supported claim rate
Definition: Share of coded claims linked to a source that visibly supports that claim.
Calculation: supported cited claims / cited claims reviewed
Reporting Guardrail: Citation presence alone is insufficient.
Evidence Class: Methodological
Version: v1.0
competitor_displacement_rate
Competitor displacement rate
Definition: Share of eligible observations where a competitor is recommended and the focal entity is absent.
Calculation: competitor-recommended focal-absent observations / eligible observations
Reporting Guardrail: Does not establish causation.
Evidence Class: Methodological
Version: v1.0
use_case_fit_coverage
Use-case fit coverage
Definition: Share of priority use cases with at least one eligible observed entity answer.
Calculation: covered priority use cases / priority use cases
Reporting Guardrail: Priority use cases must be set before analysis.
Evidence Class: Methodological
Version: v1.0
journey_stage_coverage
Journey-stage coverage
Definition: Share of defined buyer-journey stages represented in the prompt corpus.
Calculation: represented stages / defined stages
Reporting Guardrail: Coverage quality matters more than raw prompt count.
Evidence Class: Methodological
Version: v1.0
prompt_family_balance
Prompt-family balance
Definition: Distribution of observations across discovery, comparison, risk, proof, implementation and procurement.
Calculation: observations by family / all observations
Reporting Guardrail: Report imbalance rather than hiding it in a composite.
Evidence Class: Methodological
Version: v1.0
model_surface_coverage
Model-surface coverage
Definition: Number and identity of AI products or surfaces included in the study.
Calculation: declared surface count and labels
Reporting Guardrail: Never extrapolate beyond the tested surfaces.
Evidence Class: Methodological
Version: v1.0
collection_window
Collection window
Definition: Elapsed period during which observations were captured.
Calculation: first timestamp to last timestamp
Reporting Guardrail: A longer window may introduce product or index changes.
Evidence Class: Methodological
Version: v1.0
unknown_rate
Unknown coding rate
Definition: Share of fields that could not be classified with the available evidence.
Calculation: unknown fields / coded fields
Reporting Guardrail: Unknown is preferable to forced certainty.
Evidence Class: Methodological
Version: v1.0
correction_rate
Correction rate
Definition: Share of published records materially changed after documented review.
Calculation: corrected records / published records
Reporting Guardrail: Keep a public correction log.
Evidence Class: Methodological
Version: v1.0
evidence_completeness
Evidence completeness
Definition: Share of required provenance fields populated for a research record.
Calculation: complete required fields / required fields
Reporting Guardrail: Do not treat completeness as evidence quality.
Evidence Class: Methodological
Version: v1.0
Next Useful Step
Request An AI Visibility Diagnostic
Start with the commercial decision, evidence and measurement conditions that matter most. The first recommendation will define the most useful scope before any larger commitment.