AI & Methodology Disclosure

Factrail is an evidence-first, AI-assisted analytical platform. Every output begins with retrievable, citable source material and ends with structured analysis that you can independently verify.

The core principle is simple: no claim should exist without a traceable source, and no conclusion should be presented without visible reasoning. AI is a tool in this process — a powerful one — but the evidence is the foundation, not the model.

Our analytical framework, CAUSE 13.0 (Causal Assessment of Unified Societal Effectiveness), evaluates public figures, organizations, and events across 31 independent scoring dimensions. Every dimension score starts at zero and must be earned through cited evidence. This zero-based approach ensures that the absence of evidence is never mistaken for positive performance.

Each Factrail investigation follows an 8-step pipeline, executed in sequence:

1. Intent Classification — the system interprets the query and determines the investigative scope.
2. Source Retrieval — public sources are gathered from news archives, government databases, academic repositories, fact-checkers, and web search providers.
3. Claim Extraction — discrete factual claims are identified and isolated from source material.
4. Evidence Evaluation — each claim is assessed against multiple independent sources and assigned a confidence score and evidence tier.
5. Entity Identification — people, organizations, and events relevant to the investigation are identified and linked.
6. Network Graph Construction — relationships between entities are mapped, including causal chains, affiliations, and influence paths.
7. Confidence Scoring — per-claim and aggregate confidence values are computed using Bayesian analysis.
8. Synthesis & Scoring — the CAUSE 13.0 framework produces a structured dossier with dimension scores, responsibility maps, and domino effect analysis.

For full technical detail on the CAUSE 13.0 framework, including all 31 dimensions, the scoring formula, global coefficients, and the Dynamic Harm Exponent, see our Methodology page.

Factrail outputs contain three distinct layers of information. Understanding the difference between them is essential to interpreting any dossier correctly.

Layer 1: Factual Evidence. Direct quotes, data points, dates, and statements drawn from retrieved sources. These are the raw materials of the analysis. Every factual claim includes a source citation and an evidence tier classification. This layer represents what the sources say — not what Factrail concludes.

Layer 2: Analytical Inference. Structured reasoning applied to factual evidence using the CAUSE 13.0 framework. This includes causal attribution (separating a subject's direct impact from external factors), responsibility mapping, domino effect chains, and dimension scoring. Analytical inferences are computed deterministically from the evidence and the framework's published parameters. They represent Factrail's structured interpretation of the facts.

Layer 3: Model-Assisted Synthesis. Natural-language summaries, narrative explanations, and contextual observations generated with the assistance of large language models. This layer helps make the analysis readable and coherent, but it is the output most subject to the limitations inherent in AI language generation. Synthesis text is always grounded in the underlying evidence and analytical layers, but users should treat it as an aid to understanding — not as a primary source.

Throughout every dossier, these layers are presented together but remain distinguishable. Source citations, confidence indicators, and evidence tier markers allow you to assess the provenance of any specific statement.

Factrail uses third-party large language model (LLM) providers for analytical processing, evidence synthesis, and natural-language generation. The system routes queries to providers based on task requirements, using a primary provider with automatic fallback to secondary providers if the primary is unavailable.

Queries sent to LLM providers contain investigation-related content only: search topics, entity names, extracted claims, and evidence text for analysis. Factrail does not include user personal data (such as your email address, account credentials, or payment information) in any prompt sent to an AI provider.

Source retrieval is handled by dedicated web search and research providers that return publicly available information. These providers receive search queries derived from the investigation topic — not user account data.

For a full list of the categories of third-party processors we use and the types of data shared with each, refer to our Privacy Policy.

Every claim in a Factrail dossier carries a confidence score (0 to 1) reflecting the strength and consistency of its supporting evidence. Claims are also assigned an evidence tier:

• Tier A — strongest evidence: multiple independent, high-quality sources in agreement.
• Tier B — solid evidence: credible sources with minor gaps or limited independent corroboration.
• Tier C — moderate evidence: fewer sources, partial corroboration, or sources of mixed reliability.
• Tier D — weak evidence: single-source, unverified, or disputed claims.
• Tier S — self-reported: claims originating from the subject themselves, treated with appropriate caution.

CAUSE scores (0–100) aggregate across all 31 dimensions and are subject to guardrails, multipliers, and penalty mechanisms documented in our methodology. A CAUSE score is a structured analytical output — not a verdict or definitive judgment.

Known limitations include:

• Source availability: analysis quality depends on the breadth and depth of publicly available information. Subjects with limited public records will produce less comprehensive dossiers.
• Temporal gaps: events that are very recent may not yet be reflected in indexed sources.
• Language bias: source retrieval currently emphasises English-language material, which may introduce geographic or cultural bias.
• Model limitations: LLMs can produce errors in reasoning, miss nuance, or reflect biases present in their training data. Adversarial validation passes mitigate but do not eliminate this risk.
• Causal attribution uncertainty: separating an individual's direct impact from external factors involves informed estimation, not certainty.

We display these limitations openly because understanding them is essential to responsible use of the platform.

Every factual claim in a Factrail dossier links to its source material. Source citations include the origin (publication, database, or record), the evidence tier classification, and the confidence assessment.

This design serves two purposes. First, it allows you to verify any claim by examining the original source directly. Second, it makes the analytical process auditable: you can trace any CAUSE dimension score back through the claims that support it to the sources those claims were drawn from.

Factrail uses only publicly available sources. We do not access private databases, confidential records, or non-public personal data as part of our investigative process.

Factrail is designed for the analysis of public figures, organizations operating in the public sphere, and events of public significance. This includes elected officials, corporate leaders, public institutions, and widely reported incidents.

The platform is not intended for the investigation of private individuals who have no meaningful public role or profile. Non-public figures receive stricter analytical protections by default, and hypothesised relationships involving such individuals are visually distinguished from confirmed ones.

This public-interest orientation reflects a core belief: people and institutions that wield significant influence over others' lives should be subject to structured, evidence-based scrutiny. That scrutiny must itself be conducted responsibly, which is why the platform applies the same rigorous standards to its own outputs.

Factrail applies the following content standards to all analytical outputs:

• No assertions of guilt or criminality without evidence. The platform presents evidence-grounded analysis with explicit confidence levels. It does not declare individuals guilty, corrupt, or criminal as statements of fact.
• Analytical predictions are framed as analysis, not fact. Forward-looking assessments (such as legacy projections and intergenerational impact estimates) are clearly identified as analytical projections derived from current evidence, not factual predictions.
• Uncertainty is visible. Where evidence is contested, incomplete, or inconclusive, the dossier states this explicitly. Alternative explanations and data gaps are presented alongside findings.
• Adversarial validation. A dedicated red-team pass challenges findings for bias, narrative manipulation, cherry-picking, and underweighted harm. This does not guarantee perfection, but it provides a systematic check against one-sided analysis.

Auto-generated long-form articles (Revelations) are produced from completed dossier data and undergo structured editorial processing. They are analytical commentary grounded in the dossier evidence, not independent reporting.

To use Factrail responsibly, it is important to understand what the platform does not do and is not designed to be:

• Not legal advice. Nothing on this platform constitutes legal counsel. If you need legal guidance, consult a qualified professional in the relevant jurisdiction.
• Not a court or tribunal. Factrail does not adjudicate disputes, determine liability, or render binding judgments. A CAUSE score is an analytical output, not a legal finding.
• Not a credit or background-check agency. The platform is not a consumer reporting agency and its outputs should not be used for employment screening, credit decisions, or any purpose governed by consumer reporting legislation.
• Not a surveillance tool. Factrail analyses publicly available information about public figures and events. It does not monitor individuals, access private communications, or conduct real-time tracking.
• Not infallible. AI-assisted analysis has inherent limitations. Sources can be incomplete, models can err, and causal attribution involves estimation. We are transparent about these constraints because we believe honesty about limitations builds more trust than claims of perfection.

We built Factrail to make structured, evidence-grounded analysis accessible. But accessibility is not a substitute for critical thinking.

We encourage every user to:

• Review the cited sources directly, especially for claims that inform important decisions.
• Consider the evidence tier and confidence score of each claim before relying on it.
• Read the stated limitations and alternative explanations in each dossier.
• Form your own conclusions based on the evidence presented, not solely on scores or summaries.
• Consult appropriate professionals (legal, financial, journalistic) when the subject matter warrants it.

Factrail is a tool for informed inquiry. The conclusions are yours to draw.

If you have questions about our methodology, AI usage, or analytical approach, you can reach us at [LEGAL_CONTACT_EMAIL].