Vestige Verum

Most AI detection tools use another AI model to guess whether text was written by AI. Vestige Verum utilises a different approach. Instead of relying on a black-box classifier, Vestige uses Critical Discourse Analysis (CDA), stylometry, and deterministic mathematical scoring to examine how a document is constructed. It measures structural and linguistic patterns associated with modern language model outputs, including assertion density, hedging behaviour, passive framing, burden shifting, structural uniformity, and language patterns commonly associated with RLHF-trained systems. Every metric is calculated locally on-device and every result can be inspected, exported, and reviewed. No hidden model scores. No unexplained percentages. No competing AI making an unchallengeable claim. Vestige Verum evaluates writing structure rather than topic or opinion. The engine examines multiple independent signals, including: • Structural repetition and consistency across a document • Passive versus assertive language balance • Hedging and qualification patterns • Attribution and agency assignment • RLHF-aligned discourse markers commonly observed in commercial language models • Lexical diversity and stylometric variation These signals are combined into a transparent assessment framework designed to identify writing that resembles patterns commonly observed in AI-generated text. Every result includes a complete audit trail showing how the assessment was reached. Thresholds, weightings, sub-scores, and contributing factors are visible and configurable. Users can inspect the evidence behind a result rather than relying on a proprietary confidence score. A score without reasoning is difficult to evaluate. A conclusion supported by documented metrics can be independently reviewed and challenged. Vestige does not force every document into a binary human-or-AI outcome. Results are presented in three categories: • Human-Like Patterns • Structural Anomaly • AI-Like Patterns When evidence is mixed or inconclusive, a Structural Anomaly is declared instead of overstating confidence. This provides a clearer distinction between strong indicators and cases that warrant further review. Portfolio authorship tracking Single-document analysis has limitations. Vestige can build a baseline profile across multiple writing samples and measure how new submissions differ from a writer's historical patterns. Significant shifts in structure, consistency, assertion behaviour, or discourse style can be identified and flagged for review. This approach can provide context that is often unavailable when evaluating a document in isolation. Privacy first All analysis runs locally on-device. No cloud processing. No external model calls. No account creation. No subscriptions. No network dependency. The documents you analyse remain on your device. Vestige Verum is designed for educators, academics, legal professionals, researchers, compliance teams, and anyone who needs an explainable assessment of AI-like writing patterns and authorship consistency. Vestige is an investigative tool, not a decision-maker. Results should be considered alongside context, supporting evidence, and professional judgment.