Home / Hidden Radius Search and AI Discovery Audit Methodology

Evidence before recommendation

Hidden Radius Audit Methodology

The methodology traces a visible outcome backward through the discovery stack to locate the earliest controllable failure.

Published 2026-07-22Last reviewed 2026-07-22Primary-source analysis

The audit principle

Do not prescribe content before diagnosing the system

A missing citation can be caused by inaccessible pages, absent indexing, entity ambiguity, retrieval failure, weak evidence or presentation choices. Rewriting a page cannot repair every cause.

Hidden Radius audits separate evidence layers and marks the difference between confirmed facts, probable causes, platform uncertainty and recommendations.

Stage 1

Define the target outcome and scope

Which products, entities, markets, questions and platforms are included?
Is the objective discovery, accurate representation, citation, lead generation or agent task completion?
Which outcomes can be observed, and which would require private platform data?
What time period and geographic context matter?
Which systems and sources are authoritative for changing facts?

Stage 2

Build the evidence set

First-party content

Pages, feeds, profiles, documents, structured data and live systems.

Infrastructure

Responses, redirects, headers, robots rules, sitemaps, CDN and WAF behaviour.

Index evidence

Search Console, Bing tools, public index observations and cached or duplicate versions.

Entity evidence

Names, identifiers, people, locations, products, relationships and conflicts.

Retrieval evidence

Citations, linked passages, query variants and source selection.

Outcome evidence

Traffic, referrals, enquiries, bookings, conversions and verified representation changes.

Stage 3

Test each stack layer

Publication and accessibility.
Discovery and crawl access.
Canonicalization and index eligibility.
Entity recognition and conflict resolution.
Passage and fact retrievability.
Generated-answer faithfulness and attribution.
Presentation across realistic queries.
User or agent action and business outcome.

Stage 4

Classify findings by confidence

ConfirmedDirect evidence supports the finding.

ProbableSeveral observations support the explanation, but platform internals remain unavailable.

PossibleA plausible risk requiring more evidence or repeated tests.

Not observedThe issue did not appear within the defined scope and dates.

Outside controlThe outcome depends on a third-party system the organization cannot change.

Requires owner decisionThe correction affects policy, brand, legal scope, privacy or operational truth.

Stage 5

Prioritize the earliest controllable cause

An audit orders recommendations by dependency. Fix access before content optimization, identity before entity schema, source accuracy before citation tactics and task clarity before agent automation.

The deliverable includes the finding, evidence, affected layers, risk, owner, recommended action and method for verifying the change.

What an audit does not promise

The report separates influence from control

No guaranteed crawl frequency.
No guaranteed indexing or ranking.
No guaranteed citation or recommendation.
No claim to know proprietary platform internals without evidence.
No conclusion based on a single favourable or unfavourable answer.

Source trail

Primary material behind this analysis

NIST AI Risk Management Framework: Generative AI Profile

Cross-sector guidance for governing, mapping, measuring and managing generative-AI risk.

Open the original source at National Institute of Standards and Technology →

AI Features and Your Website

Official explanation of AI Overviews, AI Mode, query fan-out, eligibility and controls.

Open the original source at Google Search Central →

RFC 9309: Robots Exclusion Protocol

The standards-track specification for robots.txt parsing, matching, errors and caching.

Open the original source at IETF / RFC Editor →