Citation Probability Score (CPS)
Citation Probability Score is a composite metric ranging from 0.0 to 1.0 that estimates the likelihood an entity will be cited by a large language model in response to a relevant query.
CPS was developed as part of Answer Authority Engineering to provide a measurable, repeatable assessment of AI visibility. It replaces subjective estimates of "AI readiness" with a structured probabilistic model grounded in mathematical principles.
Input Components
CPS integrates five measurable signals:
- Extractability — How cleanly a transformer-based system can parse and isolate the entity's content
- Entity Clarity — Definitional stability and semantic consistency of the entity across its content
- Structured Data Completeness — Machine-readable metadata coverage (Schema.org, JSON-LD, discovery manifests)
- Cross-Platform Retrieval Rate — Citation persistence across multiple LLM platforms (ChatGPT, Gemini, Claude, Perplexity, Grok, Copilot)
- Spectral Stability — Resilience of the entity's embedding position under model update cycles
How CPS Is Computed
The five input signals are combined using a proprietary multi-factor weighting model. The aggregation method draws on principles from:
- Composite index construction (quantitative finance)
- Multi-criteria decision analysis (operations research)
- Bayesian evidence accumulation (probability theory)
The specific weights, nonlinear transformations, and interaction terms are proprietary trade secrets. The mathematical principles are disclosed in our open methodology.
Score Interpretation
- 0.80 – 1.00 — Strong. High probability of consistent AI citation.
- 0.60 – 0.79 — Moderate. Citation occurs but may be inconsistent across platforms or vulnerable to drift.
- Below 0.60 — Vulnerable. Low citation likelihood. Structural remediation needed.
Most websites, prior to AEO intervention, score below 0.20. The structural gap between typical web content and AI-retrievable content is significant.
Update Cadence
CPS recalculates on a 6-hour monitoring cycle. Cross-platform citation stability is measured continuously. Drift and collapse probability are assessed on each cycle.