Information Density Score

The fundamental calculation dividing unique semantic units by total token count. A high ratio indicates content where every word carries informational weight.

• Formula: (Unique Entities + Novel Claims) / Total Tokens
• Penalizes: Filler phrases, redundant explanations, marketing fluff
• Rewards: Concise definitions, data-rich statements, precise terminology

Example: A 500-token article introducing 12 new entity relationships scores higher than a 2,000-token article restating common knowledge.

The systematic down-weighting of low-information tokens that add length without substance. This mechanism directly reduces scores for content that wastes the AI's context window.

• High-penalty patterns: "In today's fast-paced world...", "It is important to note that..."
• Medium-penalty patterns: Excessive adverbs, redundant adjectives
• Zero-penalty content: Named entities, numerical data, causal statements

Content with >30% filler tokens typically falls below the retrieval threshold for generative engines.

A sub-metric measuring how efficiently complex ideas are encoded. High compression means conveying sophisticated concepts in minimal tokens without information loss.

• Measures: Information retained per token after aggressive summarization
• Ideal pattern: Topic sentences followed by structured data
• Anti-pattern: Circular explanations that restate the same concept

Technical documentation with structured schemas consistently achieves the highest semantic compression scores.

The proportion of content that introduces previously unindexed information relative to the AI model's training corpus. This directly feeds into Information Gain Scoring.

• High-novelty signals: Proprietary data, original research, post-cutoff events
• Low-novelty signals: Wikipedia summaries, common definitions
• Threshold effect: Content crossing 40% novelty density triggers preferential retrieval

Combining high novelty density with high token efficiency produces the optimal Information Density Score.

A specific calculation tracking the frequency of named entities, attributes, and relationships per hundred tokens. This serves as a proxy for factual richness.

• Target range: 8-15 distinct entity references per 100 tokens
• Below target: Indicates vague, abstract writing lacking concrete referents
• Above target: May indicate entity stuffing without coherent relationships

Optimal content balances entity density with clear predicate connections between those entities.

A modifier that adjusts the base density score based on vocabulary range and precision. Repetitive language signals low information value, while precise domain terminology signals expertise.

• Positive signals: Domain-specific terms used correctly, varied sentence structures
• Negative signals: Repeated boilerplate, synonym stuffing without added meaning
• Balance point: Technical precision without unnecessary jargon inflation

This prevents gaming through simple keyword repetition while rewarding genuine subject-matter depth.

A metric quantifying the unique, novel value a document provides beyond an AI model's existing training data. It predicts content visibility in generative search results by measuring the delta between pre-existing knowledge and new information.

• Directly correlates with citation probability in AI overviews
• Rewards original research, proprietary data, and novel synthesis
• Penalizes commoditized, widely-replicated content

The temporal and factual void between an AI model's last knowledge update and real-world events. This gap represents a critical opportunity for content to provide post-training knowledge.

• Content covering events after the cutoff date has maximum information gain
• Includes new product launches, regulatory changes, and scientific discoveries
• Requires explicit freshness signals like publication dates and versioning

The strategic introduction of new named entities, relationships, or attributes into content to expand a knowledge graph's coverage. This establishes the source as a primary origin for that entity.

• Introduce new products, features, or methodologies with clear definitions
• Define entity relationships (predicates) not previously documented
• Use Schema.org markup to formally declare new entities to crawlers

A trust metric evaluating the verifiable origin, chain of custody, and authority of data used in content. It directly influences an AI model's citation confidence and willingness to surface information.

• Primary sources (original research, raw data) score highest
• Transparent methodology sections boost provenance
• Multi-source corroboration creates triangulated trust

A content strategy focused on systematically addressing documented blind spots, unanswered questions, and zero-volume queries within an AI model's knowledge base.

• Mine 'People Also Ask' boxes and AI chat logs for unanswered queries
• Target long-tail entity coverage where training data is sparse
• Document edge cases, failure modes, and deprecated knowledge markers

The unique informational advantage conveyed by publishing non-public, first-party data such as internal benchmarks, telemetry, or user research. This data cannot be replicated by competitors.

• Creates an unassailable moat of unique information
• Includes statistical significance markers for credibility
• Transforms tacit organizational knowledge into explicit, indexable assets