This project began with a simple observation: UAP data has always existed in silos. Project Blue Book's 12,000+ Air Force case files sat in one archive. NUFORC's 134,000+ civilian reports sat in another. CIA FOIA documents were scattered across individual PDFs. The UK Ministry of Defence files were buried in the National Archives. Nobody had ever asked a single question across all of them at once.
The UAP Knowledge Graph — built under the pseudonym Redefine Zero — is the answer to that question. It aggregates twenty-seven independent government and civilian sources into a unified, queryable database, and links them through a geospatial algorithm that identifies when multiple sources independently document the same event.
The result: 17,126 verified cross-source corroborations computed for the first time — cases where two or more independent databases document the same incident at the same location within 30 days. Civilian witnesses and Air Force investigators, unknowingly recording the same event decades apart.
The Data Sources
Sixteen sources are currently active. Each is processed through a structured ingestion pipeline that normalises dates, geocodes locations, extracts structured data from PDFs using AI, and assigns a provenance-tracked record to every event.
Source
Tier
Events
Span
Notes
Blue Book
1
6,636
1947–1969
USAF official investigation program
CIA FOIA
1
525
1908–2001
Declassified intelligence reports
UK MoD
1
80
1909–2009
Ministry of Defence case files
AARO
1
35
1942–2024
Current DoD investigation office
DoW / PURSUE
1
162
1940s–2025
Pentagon PURSUE program — war.gov/UFO, May 2026 release
NUFORC
3
80,332
1906–2014
National civilian reporting center
27 active sources: NUFORC, Blue Book, CIA, UK MoD, AARO, GEIPAN (France), NARA-FAA, Brazil-Wiki, NARA-NSA, NARA-State, DIA, Army Intelligence, DoD, MUFON, NARCAP, CEFAE, NICAP, Magonia, Hatch-UDB, NUFORC-Recent, Lissoni Archives, Sign/Grudge, SETKA-MO, ODNI, RAAF-UAP, BlackVault-Australia, IUP-Press, and DoW/PURSUE (war.gov/UFO — May 2026, rolling releases). Ingestion pipeline active for future DoW tranches.
Credibility Scoring
Every record receives a credibility score from 1.0 to 9.0 based on five weighted dimensions. The model is deliberately source-aware: a declassified government intelligence cable is not evaluated on the same scale as an anonymous civilian phone-in report.
Longer observations correlate with stronger evidence
Geospatial Precision
+0.5
Confirmed coordinates vs. approximate location
Cross-Source Corroboration
Up to +4.0
Independent confirmation across databases — the most powerful factor
The cross-source corroboration bonus is the most significant factor. A civilian NUFORC record that matches a Blue Book Air Force investigation within 50km and 30 days receives a substantial score increase — because the independent agreement carries genuine evidential weight that neither source alone possesses.
NUFORC civilian records with Blue Book corroboration achieve High Credibility (score 7+) through independent confirmation. These are ordinary phone-in reports independently confirmed by official Air Force investigations at the same location and period. The knowledge graph makes that confirmation visible for the first time.
Key Discoveries
The cross-source linking algorithm has surfaced findings not visible in any individual source:
Forbes AFB, Kansas (1962) — A NUFORC civilian record matches a Blue Book investigation 0.4km away, 8 days apart. Confidence: 0.89. The highest cross-source match in the dataset. Neither source referenced the other.
Dalnegorsk, USSR (1989) — CIA documents describe a sphere crash at Hill 611 with recovered exotic alloys: gold, silver, nickel, alpha-titanium, molybdenum, beryllium. The strongest physical evidence case in the database.
The Wright-Patterson Effect — Dayton, Ohio has 114 Blue Book corroborations for nearby NUFORC records — far above baseline. The data reveals that Dayton's density reflects where Blue Book's own analysts were stationed, not just where unusual events occurred. A structural bias made visible by cross-referencing.
Sweden, 1943 — A CIA cable describes crashed vehicle recovery near Sweden, referring to physical evidence as "hardware" — wartime technical terminology suggesting material recovery was classified. Appears in no other source in the database.
Technical Stack
The platform is built entirely on open-source infrastructure at zero ongoing cost:
Database: PostgreSQL (master) + SQLite (deployment) · Ingestion: Python with AI-assisted PDF extraction · Geocoding: OpenStreetMap Nominatim · API: Datasette JSON at uap-knowledge-graph.onrender.com · AI Query Layer: Remote MCP server at uap-mcp-server.onrender.com/mcp · Hosting: Render.com + IONOS · Monitoring: UptimeRobot
The source code is available at github.com/uap-knowledge-graph. The credibility scoring methodology is designed for publication as a citable academic framework on arXiv.
Interested in Licensing or Collaboration?
We offer data licensing, API access, institutional subscriptions, and custom research partnerships. Whether you're a journalist, documentary producer, academic institution, or government agency — reach out to discuss what's possible.
The UAP Knowledge Graph aggregates publicly available data from government archives and civilian reporting databases. No warranty of accuracy, completeness, or fitness for any purpose is made. Data is provided on an as-available basis and reflects the source material as-is, including any errors, omissions, or inconsistencies present in the original records. Credibility scores reflect documentation quality as assessed by the scoring methodology — they do not constitute verification of the underlying events. All risk associated with use of this data rests solely with the user. Redefine Zero Research makes no representations and accepts no liability for decisions made based on information from this platform.
Redefine Zero Research
The UAP Knowledge Graph is an independent research initiative published under the name Redefine Zero.
For research partnerships, media licensing, institutional access, or source code: [email protected]
UAP Knowledge Graph
Frequently Asked Questions
For journalists, researchers, and curious minds.
The Basics
What is the UAP Knowledge Graph? ▼
It is the first public database to aggregate UAP records from multiple independent government and civilian sources, assign each record a credibility score, and compute cross-source links identifying when two separate databases document the same event. The result is 361,491 records spanning 1906 to 2025, across 80+ countries, from 38 sources including 9 government archives, with 17,126 verified cross-source corroborations. The most recent addition is the May 2026 Department of War PURSUE release (war.gov/UFO) — 162 newly declassified files from the Pentagon's Presidential Unsealing and Reporting System for UAP Encounters.
Is this data trustworthy? ▼
The raw data comes from official government archives — the US Air Force, CIA, DIA, NSA, US Army Intelligence, UK Ministry of Defence, the Department of Defense (AARO), FAA, US State Department diplomatic cables, GEIPAN (France's official UAP investigation office), RAAF (Australia), and the newly released Department of War PURSUE files (war.gov/UFO, May 2026). The credibility scoring system is designed to distinguish between well-documented cases and unsubstantiated reports. A score of 7+ indicates a case that is either officially documented by a government source, independently corroborated across databases, or both. The score reflects documentation strength — not a claim about what the phenomenon was.
Does this prove UAP are extraterrestrial? ▼
No. The database documents UAP reports — what was seen, by whom, and how thoroughly it was investigated. The official government position, reflected in Blue Book's final report and AARO's historical record, is that the vast majority of cases are misidentified conventional objects. The platform scores documentation quality, not anomaly significance. A highly credible case is one that was thoroughly investigated and well-documented — not necessarily one that defies explanation.
What does the credibility score actually measure? ▼
Five things: source authority (government vs. civilian), physical evidence (radar, photography, recovered material), observation duration, geospatial precision, and cross-source corroboration. The last factor is the most powerful — when two independent databases document the same event at the same location within 30 days, both records receive a significant score increase. A score of 9.0 indicates a government-documented case with confirmed coordinates, physical evidence, and independent corroboration from another database. A score of 1.0 indicates a single anonymous civilian report with no corroboration.
Using the Platform
How do I find cases in a specific area? ▼
On desktop, use the Map Area Filter — click Draw Area and drag a rectangle over any region to filter all results to that geographic area. On mobile, go to the Filters tab, tap Draw Area, then switch to the Map tab and drag your finger to draw the area. The list and map update automatically to show only events within that boundary.
How does the dataset size selector work? ▼
Choose your dataset size before clicking Load Data. TOP (~500 records, score 8+) loads instantly and shows government-verified cases. HIGH (~5,000 records, score 5+) gives a good balance for slider filtering. ALL GEO loads up to 5,000 geolocated records across all credibility levels. FULL loads the same but resets all filters. The platform currently streams up to 5,000 records per session — more than enough for interactive research.
Can I query the database directly? ▼
Yes. The full database is available as a JSON API at uap-knowledge-graph.onrender.com. It accepts standard SQL SELECT queries via the ?sql= parameter. If you use Claude or any Anthropic API-compatible AI, you can query it in natural language via the remote MCP server at uap-mcp-server.onrender.com/mcp using the mcp-client-2025-04-04 beta.
Is the data free to use? ▼
The underlying source data is public domain — declassified government records and public civilian reports. The credibility scores, cross-source link computations, and unified schema are the intellectual property of Redefine Zero Research. For academic citation, media use, institutional licensing, or source code access, contact [email protected].
For Journalists and Researchers
What is the Department of War PURSUE release? ▼
On May 9, 2026, the Pentagon launched war.gov/UFO and released 162 newly declassified UAP files as the first tranche of PURSUE — the Presidential Unsealing and Reporting System for UAP Encounters. The interagency effort includes the White House, ODNI, Department of Energy, AARO, NASA, and the FBI. Files include Air Force mission reports with UAP sightings from Syria (2022), UAE (2023), and other theatres; Apollo-era astronaut observations; Cold War intelligence cables; and civilian reports submitted to government agencies since the 1940s. Additional files will be released on a rolling basis. The UAP Knowledge Graph has ingested this initial release and will automatically incorporate future tranches as they are published.
What is the most significant finding in the database? ▼
The 17,126 verified cross-source corroborations are the primary finding — never previously computed at this scale across government and civilian sources. Among individual cases, the Dalnegorsk (USSR, 1989) CIA record is the strongest physical evidence case: a crashed sphere, multiple witnesses, recovered material with an exotic alloy composition (gold, silver, nickel, alpha-titanium, molybdenum, beryllium) that Soviet scientists could not attribute to known aerospace materials. The Forbes AFB (Kansas, 1962) case has the highest cross-source confidence score at 0.89 — a civilian NUFORC record and an Air Force Blue Book investigation 0.4km apart, 8 days apart, neither referencing the other.
How were CIA PDF documents processed? ▼
Over 800 individual government FOIA documents were processed using pdfplumber to extract raw text page by page, then sent to an AI (Claude Haiku via the Anthropic API) with a structured extraction prompt. The AI returned JSON with datetime, location, shape, evidence types, and resolution status for any specific incidents. This extracted 525 structured records from documents that were previously unsearchable. Every extraction is logged with the source document hash and extraction script version for full auditability.
How does the cross-source linking algorithm work? ▼
For each government record with confirmed coordinates, the algorithm runs a Haversine geospatial query against all other records within 50km and 30 days. Each matching pair receives a confidence score: 60% weight on geographic proximity, 40% on temporal proximity. Pairs above 0.7 confidence are classified as same_event links. The algorithm processed over 800,000 record pairs with strict date and proximity requirements, producing 1,552 verified same-event links — dominated by NARA-FAA aviation records matching NUFORC civilian reports at the same location. It runs against anchor events from Blue Book and AARO — sources with the most reliable coordinates.
What is the data accuracy disclaimer? ▼
This platform aggregates publicly available data from government archives and civilian reporting databases. No warranty of accuracy, completeness, or fitness for any purpose is made. Data is provided on an as-available basis and reflects the source material as-is, including any errors, omissions, or inconsistencies in the original records. Credibility scores reflect documentation quality under our scoring methodology — they do not constitute independent verification of the underlying events. All risk associated with use of this data rests solely with the user. Redefine Zero Research makes no representations and accepts no liability for decisions made based on information from this platform.
UAP Knowledge Graph
UAP News
Latest UAP news. Click Search in Knowledge Graph on any article to find related historical events.
Select a topic above or type a search to load news.