Home How It Works Technology Pricing Agencies Blog About Contact (480) 588-2115 Explore Options
Evidence-Bound Architecture

Evidence-Bound.
Engineered for Review.

Six interconnected systems designed to observe evidence, prioritize work, execute authorized changes, and verify what actually happened. Here's how the architecture keeps a recommendation, an attempted change, a public readback, and an external outcome distinct.

System Architecture

Under the Hood

A source-backed architecture that keeps scoring, prepared work, effect authority, provider state, public readback, and external observations in distinct records.

27

Documented Product Areas

40

Schema Template Keys

57

Scheduled Job Definitions

5

Configured AI Integrations

5

Entity Platforms

6

Interconnected Systems

System 1

AI Readiness
Scoring Engine

For an admissible submitted site, the scanner reports a 0–100 AI Readiness estimate across five weighted evidence families when coverage clears the required threshold. It can inspect up to 100 observed pages; the score summarizes setup evidence and does not predict whether an AI engine will cite the business.

📊

Five Weighted Evidence Families

Off-site Authority (28%), On-page GEO (26%), Schema Engine (18%), Entity / E-E-A-T (18%), and Crawler Access (10%). These family weights are starting estimates, not weights fitted to customer outcomes. SEO Health and measured AI Presence are separate scores; insufficient evidence remains unavailable rather than becoming a zero.

🔍

Deep Page Analysis

Crawls up to 100 pages per scan. Evaluates heading structure, image alt text, internal linking, structured data, meta completeness, and more.

📈

Actionable Issue Reports

Eligible findings can carry a specific review recommendation, configured range, or prepared revision. A recommendation is not a deployed effect, and unavailable evidence remains unavailable.

Scan time depends on site and provider response. Completed measurements can be retained for dated trend review.
🎯
AI Readiness Score — illustrative sample
80
acmeplumbing.com

Scanned 47 pages
Last scan: 2 min ago

Off-site Authority 28%
88
On-page GEO 26%
72
Schema Engine 18%
85
Entity / E-E-A-T 18%
68
Crawler Access 10%
91
System 2

Auto-Fix Engine:
Closed-Loop Architecture

The auto-fix workflow separates generation, approval, provider execution, and public readback. A connector result is not represented as live until the configured verification path confirms the exact revision.

🧠

AI-Generated Fixes

The current individual paths prepare meta descriptions within configured length bounds, title tags, image alt text, and internal-link suggestions. Homepage structured-data generation is a separate, confirmed-fact path; heading restructuring is not part of this auto-fix contract.

🚀

3 Deployment Methods

WordPress REST API, JavaScript snippet on a compatible site, or Cloudflare Worker. Capability is proven per connection and effect. Snippet changes are client-side and not verifiable in the raw public-origin response. Reversal is available only where the connector declares and binds a supported revert path.

✅

Closed-Loop Verification

After an eligible effect is applied, ClickRadius records the provider receipt separately from an independent public-origin reread where that connector permits the check. An unavailable or unobservable result remains unknown and is surfaced for review.

Evidence-bound workflow — from detection through currently authorized execution over a connection proven capable of that effect and a separate applicable public-origin reread.
⚙️
Auto-Fix Pipeline
🔍
Scan & Identify

47 pages crawled — 12 issues found across meta, schema, and content categories

🧠
Generate Fixes

Current source paths prepare meta descriptions, title tags, and image alt text; internal-link analysis and homepage structured-data generation remain separate paths

🚀
Deploy to Site

Illustrative state: an eligible change may proceed only after current authorization and a successful capability probe

✅
Verify Results

Illustrative state only: a later scan can report changed readiness inputs; it does not prove all findings resolved or an external outcome

WordPress REST JS Snippet Cloudflare Worker
System 3

Strategy Engine:
Thompson Sampling

ClickRadius currently uses Thompson Sampling over configured Beta-Binomial arm state to order eligible auto-fix strategies. It records the selection, competing samples, and an available decision-time citation baseline. Customer-outcome training is not currently active.

🎰

Configured Strategy Sampling

Each eligible optimization strategy is represented as an arm. The engine samples current arm state to order candidates; selection alone is not evidence that one strategy caused a better outcome.

📐

Decision Evidence

Each selection records the chosen arm, competing samples, context, and any grounded citation baseline available at decision time so later review can distinguish missing evidence from a measured zero.

🔄

Outcome-Training Boundary

Observed outcomes without a baseline, sufficient coverage, or an unconfounded treatment are withheld from training. A shared causal and exactly-once eligibility contract must close before customer-outcome training can become an active capability.

Current boundary: strategy selection and evidence recording are implemented; causal customer-outcome training remains gated.
🎰
Thompson Sampling Engine
θ ~ Beta(effective α, effective β)

Each eligible arm is sampled from its stored state to order the current strategy candidates

Schema Fix
78%
Meta Rewrite
65%
Alt Text
42%
Internal Links
31%
Heading Fix
22%
System 4

Entity Orchestrator

The Entity Orchestrator records and compares supported business-identity evidence across five external platforms. A listing or profile remains publisher-controlled evidence; its presence does not prove accuracy, authority, retrieval, or use by an AI engine.

🌍

5-Platform Presence

Wikidata, Google Business Profile, Apple Maps, Data Axle and Yelp. ClickRadius records the available state and keeps platform, subject, authority and observation provenance separate.

🔗

Cross-Platform Coherence

Comparable NAP, service-description and entity attributes can be checked for differences. A discrepancy is flagged; any external correction requires the platform's eligibility, current authority and its own provider evidence.

✅

Scheduled Evidence Checks

Scheduled checks record what supported collectors can currently observe. Missing or unavailable evidence remains unknown, and no check guarantees that an external record stays consistent.

Entity coherence is a measured readiness input; no retained provider evidence ranks it as an AI engine's number-one recommendation signal.
🌍
Entity Orchestrator — illustrative evidence states
📖
Wikidata

Eligibility and destination-authority review

Not established
📍
Google Business Profile

Observable public business facts

Observed
🍎
Apple Maps

Destination-authority check

Unavailable
💾
Data Axle

Provider-evidence check

Not checked
⭐
Yelp

Provider-evidence check

Not checked
Illustrative coherence state
Unavailable
System 5

Citation Monitor:
5-Engine Intelligence

Record what sampled AI responses say about your business at a known time. Per-engine integrations query ChatGPT, Gemini, Perplexity, Claude and Grok monthly, then classify observed sentiment, extract grounded URL citations and retain confidence. The result is a bounded sample, not a real-time view of every response.

🤖

Per-Engine SDK Integrations

Provider-specific API paths record direct query-response observations on the governed monthly cadence. Provider acceptance, grounding and missing responses retain distinct states.

📊

Citation Frequency Tracking

Track the frequency observed in retained monthly samples from each configured engine. The sample does not represent every answer or claim greater accuracy than an unmeasured cadence. Trends appear beside delivered-work evidence as correlation, not revenue or citation causation.

🔎

Competitive Citation Analysis

Compare named businesses in the same sampled responses. Differences can guide evidence review, but the observation does not establish why an engine selected a competitor.

Monthly observations across 5 supported AI engines, with each query and provider result retained as evidence.
🔎
Citation Monitor — illustrative sample
ChatGPT
Cited
14
citations
Gemini
Cited
11
citations
Perplexity
Cited
18
citations
Claude
Tracking
7
citations
Grok
Tracking
3
citations
System 6

GEO Content Engine

Prepare blog drafts for Generative Engine Optimization (GEO). Where context supports them, drafts can use quotations, statistics and source citations — tactics associated with improved generative-engine visibility in the cited Princeton-led research, not guaranteed citations. Source and factual controls run before publication. Eligible WordPress publication requires current authorization and a connection proven capable of publication; otherwise the article waits for review. 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.

✍️

GEO-Optimized Articles

Each standard site includes 8 authority-grade blog posts and 10 standalone knowledge articles per monthly billing period, plus a one-time 50-article Knowledge Foundation during the first 60 days. Items are admitted only when they answer a distinct customer need and satisfy the applicable evidence, authority, and quality controls.

✅

Source and Claim Review

Governed drafts inventory visible factual units and require fetched-source, authority and admission evidence to clear. The GEO score is a separate model-based screen, not factual certification or citation proof.

📰

Auto-Publish + Schema

Once you have authorized automatic publishing over a capable WordPress connection, an admitted exact revision can publish with Article markup and any additional schema supported by its visible content and current policy; otherwise it waits for review. An IndexNow request is not proof of indexing.

The cited Princeton-led experiment reported visibility-metric gains for some tactics under its test design; it does not establish a 30–40% customer citation outcome.
✍️
GEO Content Engine — illustrative draft
How to Choose a Plumber in Austin: A Data-Driven Guide
GEO-Optimized Source Evidence Cleared Review Required

Illustrative draft layout only: any quotation, statistic, attribution, or source must clear the fetched-source and factual-unit controls for the exact article before publication.

GEO Score 87/100
✓ 4 quotations from authoritative sources
✓ 8 statistics with citations
✓ 6 source URLs embedded
✓ Article markup candidate prepared
✓ Publication and notification not yet claimed

See Your AI Readiness Score

Run a free scan to review the five AI Readiness evidence families. Completion time depends on the submitted site and provider availability, and the score does not report where a private engine can or cannot find the business.

View Pricing Scan My Site Free