From Audit to Verified Work and AI Presence
A defensible workflow does not draw a straight arrow from audit to citation. It keeps setup evidence, proposed work, authorization, provider delivery, independent public observation, sampled engine answers, and business outcomes as separate records. That separation makes the system useful when an expected effect does not occur: the evidence shows where the path stopped instead of inventing success.
Why the stages must stay separate
The same word “done” can hide several different states. A recommendation may be prepared but not authorized. A provider may accept a request without changing the public site. A public page may reflect an approved revision without any external engine having crawled it. A later mention may occur without evidence that the revision caused it. Each state needs its own receipt and its own unknown state.
Sequence is not causation: approved, delivered, publicly observed, crawled, cited, and converted are different events controlled by different parties.
— ClickRadius Institute
The seven-stage evidence workflow
- Define scope — identify the organization, submitted URL, evidence sources, requested effect, and observation period.
- Measure AI Readiness — report the five setup-evidence families with version and coverage.
- Prepare exact work — retain the proposed revision, source evidence, intended effect, and limitations.
- Authorize and deliver — require current authority and a connection proven capable of that effect, then retain the provider receipt.
- Verify the public origin — independently reread the public surface where supported and retain correction or withdrawal state.
- Observe AI Presence — keep monthly sampled engine answers separate from readiness and delivery.
- Compare and decide — review like-for-like evidence without turning correlation into attribution.
Stage 1: define the evidence boundary
Record what is being measured before interpreting a number: the canonical organization, submitted URL, crawl or observation time, evidence age, supported connectors, and the effect under review. If a source is unavailable or a relationship is ambiguous, preserve that state. Unknown is evidence about the measurement, not permission to infer a favorable answer.
Stage 2: measure the five AI Readiness families
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 AI Presence are separate and never blended into the score.
Missing evidence does not become zero or pass. If supported evidence does not clear the versioned minimum coverage threshold, the composite remains unavailable. A later result is comparable only when the target, score version, and evidence basis are sufficiently alike.
Stage 3: prepare an exact revision
A measured finding can support a proposed change to content, schema, metadata, crawler policy, or another eligible surface. The proposal should retain the exact before-and-after revision, the evidence behind it, its intended effect, and any claims that require owner verification. Research can inform what to inspect without promising the same effect on a customer site.
Preparation is not production. Off-site listings, reviews, earned coverage, and other third-party work also remain governed by the authority and rules of those surfaces.
Stage 4: authorize, prove capability, and deliver
Applying a site effect requires current authorization for the exact revision and a connection proven capable of that effect. If either gate is absent, the work remains for review or customer implementation. A provider response should be retained as accepted, rejected, or unknown rather than converted into a generic success message.
Stage 5: verify the independent public origin
Where the effect supports it, reread the public origin independently of the write path and compare it with the approved revision. This can establish what the public surface returned at that time. It still does not prove that an external crawler saw, rendered, indexed, or used it. A correction or withdrawal path must remain available if the public result is wrong or authority changes.
Stage 6: sample AI Presence monthly
ClickRadius separately records monthly sampled answers from ChatGPT, Gemini, Perplexity, Claude, and Grok. Microsoft Copilot is not currently supported. Retain the tested prompt set, engine, successful-response denominator, mentions, citations, and unknown or failed observations. The result describes that configured sample, not every answer users saw and not continuous monitoring.
Stage 7: compare without inventing causation
Compare equivalent score versions, public observations, and prompt samples. A changed readiness family says that supported setup evidence changed. A changed AI Presence sample says that the retained answers changed. Neither result, alone or in sequence, proves why an engine made a selection.
Keep referrals, campaign identifiers, optional intake answers, qualified leads, booked revenue, cost, and margin in a separate business-outcome layer. Attribute only what the retained provenance supports; leave the rest unknown.
How ClickRadius supports the workflow
ClickRadius measures AI Readiness and separate SEO Health, prepares governed work, records authorization and provider receipts for supported effects, supports independent public-origin rereads where available, and reports monthly AI Presence across the five supported engines. External engines retain control over crawling, indexing, retrieval, ranking, recommendation, and citation.
The direct plan is $499 per month with a 14-day retail trial and payment required; it is month-to-month. The agency plan is $200 per activated client site per month with no setup fee or trial; each site has its own four-month initial term from activation and then continues month-to-month. Commercial terms do not change any evidence or effect boundary.
Frequently asked questions
How long does the workflow take to produce an external-engine result?
There is no supported universal timeline. ClickRadius can timestamp measured evidence, authorization, provider receipt, public-origin reread, and monthly sampled engine observations. External engines control their own crawl, index, retrieval, ranking, recommendation, and citation timing.
Can ClickRadius apply every prepared correction?
No. Applying an exact supported revision requires current authorization and a connection proven capable of that effect. Otherwise the work remains for review or customer implementation. Provider acceptance and public effect are retained as separate states.
Does AI Presence prove that prior work caused an answer to change?
No. AI Presence is a monthly sample across ChatGPT, Gemini, Perplexity, Claude, and Grok. It reports what the configured sample observed, with denominators and unknown states; it does not establish why an answer changed. Microsoft Copilot is not currently supported.
Start with supported evidence. Submit a URL for a free AI Readiness assessment, then see ClickRadius plans for governed work and separate monthly AI Presence observations. The assessment does not contact the five engines or predict an external result.