How to Audit Your Own Site for AI Readiness
This is a reproducible inspection method, not a claimed ClickRadius case study. It shows how to retain evidence across the same five families used by AI Readiness while keeping manual observations distinct from the served score, governed site effects, and external-engine outcomes.
Write the audit identity before collecting evidence
Record the organization, submitted URL, canonical destination, inspection time, pages included, source for every off-site observation, and the model or checklist version. Without those fields, a later comparison may silently mix different sites, scopes, or rules.
Use explicit states: measured value, measured zero, unavailable, not applicable, stale, and needs measurement. Do not turn a blocked request or missing source into a zero or a pass.
The five-family model
ClickRadius AI Readiness estimates five weighted setup-evidence families: Off-site Authority (28%), On-page GEO (26%), Schema Engine (18%), Entity / E-E-A-T (18%), and Crawler Access (10%). These are current starting estimates, not weights fitted to customer outcomes. SEO Health and monthly sampled AI Presence are separate measurements.
A manual audit may inspect the same kinds of evidence, but it is not automatically a ClickRadius score. Producing that score requires the versioned calculation, supported evidence rules, and minimum coverage threshold. When coverage is insufficient, the served result remains unavailable.
1. Inspect Off-site Authority evidence
Record the external directory, review, syndication, social, link, and entity-record surfaces you can actually observe. Retain the source URL and time. Separate a missing listing from an unavailable request, and do not infer third-party facts from the organization's own site.
2. Inspect On-page GEO evidence
Review the selected pages for clear answer passages, specific supported claims, useful headings, attributed quotations, statistics with scope, and citations that genuinely support the sentence. Also identify purely promotional or duplicated sections. Record the page and passage rather than reducing the finding to “content weak.”
The Princeton-led GEO study reported visibility changes for several interventions in its own benchmark. That result can inform what to inspect; it does not calibrate the ClickRadius weights or predict a result for the audited site. The benchmark began with the top five Google sources, randomly selected one source for optimization, and measured that source's relative visibility/impression in generated responses; it did not measure retrieval or citation-inclusion probability, and reported effects varied by domain.
3. Inspect Schema Engine evidence
Extract supported structured data and check syntax, stable identifiers, duplicates, relationships, and agreement with visible content. A validator result proves only what that validator tested. It does not verify every declared fact or establish that an external engine used the markup.
4. Inspect Entity / E-E-A-T evidence
Check whether the available evidence supports one identifiable organization, real authors, claimed credentials, and consistency between owned and independent sources. Preserve contradictions rather than choosing the most favorable version. Claims that require owner confirmation remain unverified until that confirmation exists.
5. Inspect Crawler Access evidence
Read the declared robots.txt policy and record whether an llms.txt is published. This family describes site policy and raw-page evidence; it does not prove that an external crawler arrived, rendered JavaScript, indexed content, or cited a page.
Keep adjacent measurements separate
SEO Health covers conventional search hygiene and is not part of AI Readiness. AI Presence records monthly sampled answers from ChatGPT, Gemini, Perplexity, Claude, and Grok; Microsoft Copilot is not currently supported. Retain the prompt set, successful-response denominator, mentions, citations, and unknown or failed observations. A sample is not continuous monitoring or a census of every user answer.
Turn a finding into governed work
- Retain the finding and its evidence.
- Prepare the exact revision and identify its intended effect.
- Confirm current authorization for that revision.
- Prove the connection can perform that effect.
- Retain the provider response as accepted, rejected, or unknown.
- Reread the independent public origin where supported.
- Keep correction or withdrawal available.
A provider receipt is not a public effect, and a public effect is not proof of a later crawl, citation, lead, or sale. External engines control crawling, indexing, retrieval, ranking, recommendation, and citation.
Frequently asked questions
What does the ClickRadius AI Readiness assessment check?
It estimates five weighted setup-evidence families: Off-site Authority, On-page GEO, Schema Engine, Entity / E-E-A-T, and Crawler Access. SEO Health and monthly sampled AI Presence are separate. The assessment does not grade current citability or predict an external-engine outcome.
Can I audit my own site without ClickRadius?
You can inspect public pages, structured data, declared crawler policy, identity evidence, and selected off-site surfaces manually if you retain source, time, scope, and unknown states. A manual result should not be presented as the ClickRadius score unless it uses the same versioned model and evidence rules.
What should I change first after an audit?
There is no universal first change. Resolve target and coverage uncertainty, inspect the supported family findings and dependencies, then prepare the exact revision. A site effect requires current authorization and a connection proven capable of that effect; external engines still control their own outcomes.
Run a free AI Readiness assessment for a submitted URL, or see ClickRadius plans for governed correction preparation and separate monthly AI Presence samples. Insufficient supported evidence remains unavailable rather than becoming a confident score.