AI Visibility Audit Methodology V5

What the current audit measures

The AI Visibility Audit evaluates observable website evidence that supports discovery, understanding, first-party credibility and decision-making. Product version 5.0 uses model identifier v5.0-website-ai-readiness-1.0. It is a point-in-time website-readiness assessment, not a measurement of an AI platform's private ranking or recommendation algorithm.

The free preview and paid detailed report use the same readiness model. The detailed report explains findings, evidence, limitations and prioritized next steps. Payment does not change the scoring methodology.

Four scored Website AI Readiness layers

AI Presence

Reviews whether the website can be reached, crawled and interpreted, using response status, crawl directives, sitemap evidence, canonical information, rendered content and navigation where available.

AI Understanding

Reviews whether the website identifies the business, offerings and relationships clearly. Evidence includes business descriptions, entity and website schema, naming consistency, headings, contact-page presence and offering coverage.

AI Authority

Reviews first-party credibility and verification evidence, such as visible founder or leadership identity, meaningful About content, contact details, applicable location and customer policies, reviews, and original supporting content. Website claims are not automatically independent external endorsement.

Recommendation Readiness

Reviews decision-useful website evidence: audiences and use cases, service or product specifications, relevant prices, benefits, comparisons, FAQs, answer-ready content, service-area clarity and supporting proof. This layer describes evidence available on the website; it does not establish that ChatGPT or another system has recommended the business.

Citation Authority is a separate assessment

Citation Authority is not included in Website AI Readiness. There is no combined five-layer overall score.

Independently validated external evidence is required for an external Citation Authority score. Ordinary website audit creation does not independently validate that evidence, so Citation Authority may remain unassessed. An unassessed external dimension is not a measured zero and should not be presented as proof that external authority is absent.

Links, references and review profiles observed on a website can inform future verification, but they are not a substitute for an independent external assessment.

Evidence collection and requirements

The audit inspects accessible website HTML, rendered content, metadata, structured data, headings, page sections and links. Relevant linked pages may supply additional evidence about identity, contact information, location, offerings, FAQs and policies. This is a bounded crawl, not an exhaustive review of every page or external source.

Findings retain captured evidence and its source where available. Structured information and verified page content are distinguished from weak keyword inference. A loose phrase is not equivalent to a clearly supported section, offering, identity or policy.

The model maps source signals into the current layers. Some source signals are consolidated to avoid counting duplicate evidence; others are informational, withheld or not applicable. The current product should not be described as 53 independently scored checks. The historical 53-signal framework is documented separately.

Statuses, applicability and capture completeness

• Detected: The available evidence supports the requirement.
• Partial: Some relevant evidence exists but does not fully satisfy it.
• Missing: A requirement was evaluated from sufficient relevant evidence and was not established.
• Unknown: Capture or evidence is insufficient to verify the requirement. Applicable points remain in the denominator and no credit is awarded; Unknown does not prove the information is absent from the website.
• Not Applicable: The requirement genuinely does not apply to the supported business model. Its points are excluded from the score denominator and it does not reduce evidence confidence.

Business-model applicability is evidence-based. Physical businesses, service-area businesses and online-only businesses have different location requirements. A private mailing address is not automatically a customer-facing premises. Physical-product shipping and returns are separate from consultation cancellation, rescheduling and treatment-package terms. Booking or checkout links alone do not establish physical-product sales.

Incomplete rendering or a failed linked-page capture can leave affected findings Unknown. Verified positive evidence can still receive credit, and independently complete sources can still support Missing findings. One incomplete section does not automatically invalidate the entire audit.

Scoring and evidence confidence

Website AI Readiness summarizes the four scored website layers on a 0–100 scale. Applicable supported evidence contributes to layer scores; legitimate Not Applicable requirements are excluded. The public documentation does not disclose the proprietary formula, weights or thresholds.

Evidence confidence describes the strength and completeness of captured evidence, not the probability of an AI recommendation. Unknown evidence reduces certainty; legitimate Not Applicable requirements do not. Confidence and readiness are distinct concepts, and neither is an accuracy guarantee or a live visibility measurement.

What is not included

No live ChatGPT, Gemini, Perplexity or Google AI prompt testing is included in this audit. It does not measure live mentions, citations, shortlist placement or recommendation frequency. Such testing is a separate exercise requiring its own prompts, dates, platform settings and evidence.

The audit does not guarantee rankings, citations, recommendations, traffic or revenue. A strong readiness score does not prove an AI system will recommend a business. Rendering limitations, inaccessible pages, changing website content and incomplete external evidence can affect findings. Recommendations concern verified website gaps or further verification, not guaranteed platform outcomes.

Version history and research boundaries

The historical AI Visibility Audit Methodology V4.2 remains available at its original URL. Its source document was published on 3 August 2026; repository history records presentation changes on 13 August 2026. It documents the earlier five-dimension, 53-signal framework and is not the current commercial V5 methodology.

The V5 Stage 1 code records a freeze date of 19 August 2026. That is a configuration milestone, not this page's publication date or a proven production-launch date. This public V5 documentation is published separately on 6 October 2026.

The AI Visibility Benchmark 2026 retains its own published methodology, dimensions, dates and results. This page does not retrospectively claim that the July research used the August V4.2 release, or substitute V5 layers for the Benchmark dimensions.

Related documentation and tools

• Run the current AI Visibility Audit
• Read the historical V4.2 methodology
• Browse original research
• Read AI visibility resources