The canonical specification

The Standard.

The published specification of what AI discoverability is — its definitions, signal taxonomies, structured-data subsets, authority taxonomy, and entity-resolution rules. The reference document of the field.

VERSION
1.0 In force
FIRST PROMULGATED
2026 — by the Standard Committee
NEXT REVIEW
Q4 2027
FORMAT
PDF · 32 pp · Free in perpetuity
Read the document

The Standard, in its canonical edition.

CANONICAL EDITION · AI DISCOVERY

The AI Discovery Standard™

v1.0
2026
PAGES
32
WORDS
~9,500
LICENSE
Editorial · Free
LAST UPDATED
2026-05-09
The signal taxonomy

Eight signals. One composite.

01

Mention Frequency

The share of relevant AI responses, across measured systems and the prompt corpus, in which the constituent is named.

16%WEIGHT
02

Mention Authority

The qualitative weight an AI system places on the constituent when surfacing it — Primary, Secondary, Neutral, or Cautionary.

18%WEIGHT
03

Citation Source Quality

The independent quality of the public sources AI systems cite when discussing the constituent, weighted by prominence.

14%WEIGHT
04

Entity Resolution Strength

Whether AI systems consistently resolve the constituent to a single, correct identity — measured across systems and prompts.

12%WEIGHT
05

Schema and Structured Data

The completeness and validity of structured data describing the constituent, against the Standard's structured-data subset.

8%WEIGHT
06

Authority Graph Distance

The graph distance between the constituent and the small set of nodes treated as canonical for its field.

12%WEIGHT
07

Behavioural Reinforcement

Publicly observable proxies for the downstream user behaviour reinforcing the constituent's presence in AI responses.

10%WEIGHT
08

Recency Stability

Whether the constituent's discovery state is stable across recent observation windows, as opposed to situationally visible.

10%WEIGHT
The composition

How the score composes.

The composite formula

Each signal is normalised to a 0–100 scale within its vertical and observation window. The headline AI Discovery Score is the weighted sum, scaled to 0–1000:

Score = 10 · ( 0.16·S₁ + 0.18·S₂ + 0.14·S₃ + 0.12·S₄ + 0.08·S₅ + 0.12·S₆ + 0.10·S₇ + 0.10·S₈ )

The 0–1000 scale was chosen for two reasons. First, it provides sufficient resolution to distinguish constituents in dense bands without resorting to decimals. Second, it is far enough from the 0–100 percentile mental model that buyers do not confuse a score with a probability.

Why these weights

  1. INo signal carries more than 20%. No signal can dominate; the composite is robust to any one signal's manipulation.
  2. IIThe two trust-anchored signals (2 and 3) together carry 32%. Trust outweighs volume, by design.
  3. IIIThe four operational signals (1, 5, 7, 8) together carry 44%. Half the score is movable through documented practice; half is not. The latter half is what makes the Index expensive to game.
The bands

Six bands. Population shares within each vertical.

A · Discovery-leading
The top of the table. Constituents whose authority and frequency place them at the top of the answers in their vertical.
Score 850–1000
≤ 5% of population
B · Strongly discoverable
Routinely surfaced in AI responses on relevant prompts; with strong trust framing in most appearances.
Score 700–849
≈ 15%
C · Adequately discoverable
Surfaced in roughly the share an unweighted observer would expect for their vertical and locality.
Score 550–699
≈ 25%
D · Discovery-marginal
Surfaced inconsistently; trust framing is mixed; the engagement opportunity is largest at this band.
Score 400–549
≈ 25%
E · Discovery-weak
Surfaced rarely; structural barriers to discoverability that respond well to twelve-month engagement.
Score 250–399
≈ 20%
F · Discovery-invisible
Effectively absent from AI responses on relevant prompts. Foundational structural and authority work required.
Score 0–249
≈ 10%
Governance

Maintained by the Standard Committee.

The Standard Committee

The Standard is maintained by a Committee of seven, appointed for staggered three-year terms. The Committee comprises a Chair, three Methodology Members, two Practice Members, and one Public Member. Members' names, terms, and disclosed conflicts are published.

The Committee operates constitutionally independently of the commercial leadership of AI Discovery the firm under Article IV of the Institutional Charter. Definitional disagreements between the Standard and any other institutional artifact are resolved by reference to the Standard, with the Committee as final adjudicator.

View the Committee →

Revision procedure

  1. IProposal — submitted in writing by a Committee member or qualifying observer, with rationale and continuity analysis.
  2. IIPublic consultation — sixty-day public posting; written responses accepted from any party; all responses published.
  3. IIICommittee vote — two-thirds majority required for revisions; voting record published in the Reproducibility Pack.
  4. IVObservation period — six months parallel publication of prior and revised series.
  5. VAdoption — revised methodology becomes canonical at the next quarterly cycle, with continuity guarantees.
Change log

Every revision, archived in perpetuity.

v1.0
PromulgationThe Standard's first canonical edition. Eight-signal taxonomy, structured-data subset, authority taxonomy, entity-resolution rules. Promulgated by the Founding Convocation.
2026 · IN FORCE
v1.0.1
Refinement (M-2026-01)Structured-data subset clarified to include three additional schema.org properties relevant to professional-services entities. No effect on continuity.
2026-03 · APPLIED
v1.0.2
Refinement (M-2026-02)Authority-signal taxonomy expanded from three bands to four (Primary, Secondary, Neutral, Cautionary). The Cautionary band introduced to capture qualified or hedged framings.
2026-08 · APPLIED
v1.1 (draft)
Methodology refinement on schema step-functionCurrently in 60-day public consultation. Addresses the discontinuous effect observed in Founding Quarter data at structured-data completeness ≈ 0.7. Adoption Q1 2027.
CONSULTATION OPEN
For practitioners and researchers

How to cite the Standard.

Editorial citation

AI Discovery (2026). The AI Discovery Standard, v1.0. AI Discovery Standard Committee. Available at: aidiscovery.org/standard.

Editorial use of the Standard is free, in perpetuity, with attribution. Quotation, reproduction, and reference in editorial, academic, and professional contexts are permitted without further licence.

Academic citation (APA)

AI Discovery Standard Committee. (2026). The AI Discovery Standard (Version 1.0) [Methodology specification]. AI Discovery. https://aidiscovery.org/standard

The Standard's archival URL is permanent; prior versions remain accessible. Academic citation may quote up to 1,000 contiguous words without further licence; longer quotation requires editorial-class licence (free, on application).

Read the canon

The Standard
is the field.

Free in perpetuity. Reviewed annually. Maintained under public governance. Begin with the document; if you want to apply it to your own business, begin with the Audit.

Download the Standard