Data Accuracy & Evidence Standards
Not all sources carry equal weight. Disclosure Navigator uses a three-tier evidence system to make the quality and origin of every claim transparent — so you can make your own judgment about what to trust and to what degree.
The Source Hierarchy
T1 — Government Documents are the evidentiary foundation. These are official institutional records: declassified files, agency reports, congressional testimony transcripts, and formal government communications. T1 is the highest confidence tier because these documents represent what institutions put on the record under their own authority. They are not proof of the claims they contain — governments can be wrong, incomplete, or deliberately misleading — but they are the primary record.
T2 — Research Sources cover researcher analysis, witness testimony, documentary interviews, and investigative journalism. T2 sources are valuable for context, corroboration, and access to witnesses who aren't represented in official records. They are treated as secondary to T1 — informative and often credible, but not primary institutional evidence.
T3 — AI-Enriched Data fills biographical and contextual gaps that neither government nor research sources address. T3 content is always clearly labeled with an AI BG badge and is never cited as primary evidence. It exists to help orient you when background context is missing from the record, not to make claims about what happened.
How Relationship Confidence Works
Relationships between entities are extracted from source text and assigned a confidence score based on how explicitly the connection is stated. A document that says "General X directed Program Y" yields a high-confidence DIRECTED relationship. A document that says "Program Y may have involved General X" yields a lower-confidence relationship. Low-confidence relationships are visible in entity profiles and the graph but are marked so you can weigh them appropriately. The source passage is always accessible so you can read the original language yourself.
AI-Generated Content
Entity profile summaries and entity fact tables are AI-synthesized from source documents. They are not copied verbatim — the AI reads across multiple documents about an entity and writes a coherent summary. This means occasional oversimplification is possible. The underlying source documents are always linked so you can verify specific claims.
Briefing answers are drawn only from the library. The AI does not pull information from the open web when answering a briefing question. Every claim in a briefing is accompanied by a citation to a specific library document. If the library doesn't contain evidence for something, the briefing will say so rather than speculating.
Limitations to Be Aware Of
- Redactions and deliberate obscurement — Early UAP documents, particularly pre-1970s files, often contain heavy redactions. The library reflects what was released, not the complete underlying record. Gaps in the data are real gaps in what has been disclosed, not platform errors.
- Incomplete entity profiles — An entity that appears in only one or two documents will have a thinner profile than one with extensive documentation. Thin profiles are a signal about the state of the record, not about the entity's actual significance.
- AI extraction errors — The AI ingestion pipeline is thorough but not perfect. Occasional misattribution of a relationship, an incorrectly parsed date, or a misidentified entity is possible. If you find an error, contact us — corrections improve the platform for everyone.
How to Evaluate Claims Yourself
Every briefing citation links to the source document in the library. Every entity profile lists the documents that mention that entity. Every relationship edge shows the source passage. The tier badge on any piece of data tells you immediately whether you are reading from a government record, a researcher's account, or AI background context. Use those signals to apply your own standard of evidence. The platform is designed to make that evaluation easy, not to make it for you.