The AAO Score

01What the AAO Score is

The AAO Score is a rating, not a tool, not a scan, and not an audit grade. The word matters. A rating implies a transparent methodology, an independent assessor, a named scale with disclosed bands, and a confidence band on the result. The AAO Score is built on all four.

Three properties distinguish it from existing AI search visibility numbers:

1. Outcome-anchored. Gate 10, Weighted Citability, is measured from actual ICP queries run against ChatGPT, Perplexity, Microsoft Copilot, and other agents. Gates 1 through 9 are predictors that explain why Gate 10 lands where it does. Predictor gates and the measured outcome must agree, and divergence between them is itself a diagnostic signal.
2. Vertically calibrated. What counts as "strong reputation" or "current content" differs across Industrial, B2B SaaS, Professional Services, and Healthcare. The rating selects a Vertical Profile and scores against that vertical’s canonical channels, not a one-size-fits-all checklist.
3. Reliability disclosed. Every score reports how many of the 10 gates were assessable, what confidence the operator placed on each, and which reliability tier the snapshot ships at (Standard, Preliminary, or unable to ship). The number on the cover is never the only thing the report says.

02How to read your score

The numeric AAO Score is reported alongside a named band. The band is the one-glance verdict; the number is the resolution within the band.

An example score visualized as a gauge:

Example: an AAO Score of 67 sits in the AAO-C band. Foundation tier likely passing, advanced tier (Authority, Recency, Granularity, Differentiation, Weighted Citability) is where the gap lives.

03How the score is calculated

Three inputs combine to produce a single AAO Score: the 10 gate scores, the vertical profile that selects which canonical channels matter, and the weighting profile that determines how heavily each gate counts toward the composite.

The 10 gates

Each gate is scored 0 to 5 against documented tier criteria. A 5 means best-in-class against vertical-canonical signals; a 0 means catastrophically absent or actively blocked. Gates split into a Foundation tier and an Advanced tier:

1. 01
   Discovery Can AI agents find the content? Sitemaps, internal linking, llms.txt, and bot-accessible routing.
2. 02
   Structure Is content machine-parseable? Server-side rendering, JSON-LD, semantic HTML.
3. 03
   Content Is the content substantive, factual, and E-E-A-T compliant? Named authors, primary sources, operational specificity.
4. 04
   Reputation Are there third-party signals on the canonical channels for the brand’s vertical? Vertically calibrated, not a single checklist.
5. 05
   Infrastructure Core Web Vitals, HTTPS, mobile responsiveness, no JavaScript-only rendering. The technical foundation that retrieval depends on.
6. 06
   Authority Wikidata, Wikipedia, named principals with credentials on platform-canonical profiles, NAP consistency. The single highest-leverage advanced gate.
7. 07
   Recency Signs of currency, not frequency of body-text change. Schema dateModified, "Last reviewed" stamps, references to current standards.
8. 08
   Granularity Self-contained, quotable passages. FAQ schema, definitional sentences, structured chunks AI can cite verbatim.
9. 09
   Differentiation Original content AI cannot synthesize from other sources. Named accreditations, exclusive partnerships, primary research.
10. 10
    Weighted Citability The measured outcome. ICP queries run across ChatGPT, Perplexity, Copilot, sampled three times each. Layer 1 uses industry-benchmark queries; Layer 2 uses prospect-specific queries.

The AAO Framework page documents each gate in operational detail.

Vertical profiles

The same numeric criteria mean different things in different industries. A 4 on Reputation for an industrial manufacturer requires presence on ThomasNet, ANAB, and trade publications. A 4 on Reputation for a B2B SaaS company requires G2, Capterra, and Forrester or Gartner coverage. The rating selects one of five profiles at scoring time:

• Industrial / B2B Physical Goods. Manufacturers, distributors, fabricators.
• B2B SaaS / Digital Services. Software products, APIs, dev tools.
• Professional Services. Consulting, legal, accounting, RIA, marketing consultancies.
• Healthcare / Regulated. Clinical providers, pharma, medical devices.
• Mixed. Cross-category or no clear vertical fit. Uses the most universal channels.

Weighting profiles

The 10 gates do not contribute equally to the composite. The brand selects a weighting profile depending on its discovery model:

• Traditional Organic (SEO-Heavy). Google organic remains the dominant traffic source. Foundation tier weighted higher.
• AI-First (default). AI-driven discovery is the primary growth vector. Advanced tier weighted higher. Authority and Weighted Citability dominate.
• Balanced 2026. Mixed organic and AI-driven discovery.

Reliability tiers

Not every gate is assessable for every brand. The number of N/A gates determines what reliability tier the rating ships at:

• Standard. 0 to 2 gates N/A. Score ships without qualifier.
• Preliminary. 3 to 4 gates N/A. Score ships with a banner disclosing that the verdict is provisional.
• Cannot ship. 5 or more gates N/A. Rating returns to the operator for additional research, or the brand is documented as having insufficient public surface for the methodology.

04Why a rating, not a tool

Moody’s rates bonds. FICO rates borrowers. S&P rates corporate credit. The reason these institutions persist across multiple economic cycles is not that their numbers are uniquely accurate. It is that their methodologies are transparent, their processes are reproducible, and their disclosures are mandatory. Other parties can audit how the number was produced, challenge the criteria, and predict how the score will change as the underlying entity changes.

The AAO Score is built on the same posture, applied to a different surface: AI search citability instead of debt repayment. The numeric output sits on top of:

• A published 10-gate methodology with named criteria and observable signals.
• A versioned scoring rubric with documented changes between revisions.
• Vertical profiles that disclose which canonical channels apply to which industries.
• Reliability tiers that admit when the rating cannot be produced confidently.
• A validation methodology with agreement-threshold pass/fail criteria for any automation that scores on behalf of human operators.

The word "tool" implies a black box that produces a result. The word "rating" implies a chain of evidence from input to output that a skeptical reader can follow. AI search visibility is a new enough surface that the methodology needs the credibility the word "rating" implies.

05How the rating stays honest

The 2008 financial crisis showed what happens when rating agencies have their methodology gamed by the entities they rate. The AAO Score is built with three structural defenses against the same failure mode:

1. Outcome gate (Gate 10) is measured, not asserted. The brand cannot directly influence what ChatGPT, Perplexity, or Copilot return for a buyer-typed query. A brand that improves predictor gates 1 through 9 without moving Gate 10 produces a divergence the methodology flags as a finding rather than treating as an improvement.
2. Rating and remediation are separate engagements. The fee structure for producing a rating does not depend on the rating outcome. Remediation engagements that follow a rating are priced separately and do not modify the methodology used to produce the rating.
3. Methodology is published and versioned. Every rubric change is documented in the AAO Framework changelog. Issuers and third parties can audit the methodology against the score they received.

06Further reading

• AAO Framework: the 10-gate methodology in operational detail
• Bayesian MMM theory: how AAO Scores become priors in a Marketing Mix Model
• AAO Discovery Snapshot: the productized rating report
• Forthcoming whitepapers: methodology validation, band calibration, and cross-vertical comparability
• Original AAO framework article by Andres Plashal

07Frequently asked questions