Diagnostics and Action Center
GEO diagnostics
Agentic Readiness evaluates whether a target is understandable, attributable and citable by AI systems. It covers technical discoverability, canonical identity, structured data, content clarity, evidence quality, entity consistency, machine-readable artifacts and answer readiness.
It does not score cart, checkout, payment, Masterpass or Mastercard readiness. Commerce endpoints discovered during crawling may remain in raw technical evidence, but they must not influence GEO scoring or recommendations.
Diagnostic lifecycle
- Select target, scope and provider.
- Discover reachable pages and machine-readable assets.
- Evaluate technical, content, entity, evidence and citation signals.
- Store an immutable report with score breakdown and evidence.
- Convert actionable gaps into findings and recommendations.
History preserves old reports; improvements to the engine do not rewrite previous snapshots.
Diagnostic score composition
| Area | Representative questions | Excluded concepts |
|---|---|---|
| Discoverability | Can approved crawlers reach and identify the relevant assets? | Cart/checkout availability |
| Entity clarity | Is the target canonical, unambiguous and related correctly? | Payment provider support |
| Citability | Are claims specific, sourced, current and answer-ready? | Transaction completion |
| Machine readability | Are appropriate metadata, structured data and artifacts valid? | Commerce protocol readiness |
| Evidence quality | Can important statements be traced to authoritative sources? | Masterpass/Mastercard presence |
| Answer readiness | Can an AI system describe the target accurately for relevant prompts? | “Can the AI buy here?” scoring |
Recommendation roadmap
Recommendations include rationale, proposed change, impact, confidence, effort, priority score, expected lift, owner, due date, dependencies and verification plan. Prioritization must remain explainable.
Governed actions
Actions carry risk class, scope, idempotency key, proposed diff, policy result, execution result, verification result and rollback plan. Row actions are managed through modal workflows; opening a separate edit page is not required for routine table operations.
Controlled autopilot
Autonomy begins at suggest. Higher automation levels may prepare approval-bound steps, but the planner returns executable=false unless all external execution requirements are satisfied. Budget, dependency order, approval checkpoints, safe stop and emergency stop remain mandatory.
Action responsibility matrix
| Step | System responsibility | Human/policy responsibility |
|---|---|---|
| Finding | Preserve evidence and explain the gap | Confirm relevance and business context |
| Recommendation | Estimate impact, confidence and effort | Accept, dismiss or refine |
| Proposed action | Produce diff, dependencies and verification plan | Assign owner and risk class |
| Approval | Enforce required state and record actor | Review scope, claims, destination and rollback |
| Execution | Apply idempotently and record external reference | Intervene on unsafe/unknown state |
| Verification | Check expected destination and later measurement | Accept outcome or request rollback |