UFOSEARCH

DECLASSIFIED ARCHIVE ← back to search

UFOSearch is a grounded, searchable index of the declassified government record on unidentified aerial phenomena — the actual files released by national governments, unified into one place where you can ask a question in plain English and read the real documents that answer it.

Every result is a verbatim page from a real government document, linked back to its official source — the exact scan where one’s available, or the government archive it came from. Nothing here is AI-generated, and nothing is embellished. When the record doesn’t support an answer, the honest response is “insufficient evidence” — not a guess.

6,980documents
436,169indexed passages
6national governments

Where the data comes from

Everything is sourced from official government archives and their public releases. Current holdings:

United States2,513 documents
Project Blue Book / Sign / Grudge
USAF sighting case files, 1947–1969
927
CIA UFO collection
FOIA case F-2020-02272, 1950s–1990s
713
NARA Record Group 615
2024-NDAA transfers — FAA ‘Skywatch’, NRC, ODNI, DoD
612
DoD PURSUE Release 01
Department of War UAP release (war.gov)
184
DIA AAWSAP/AATIP DIRDs
Pentagon advanced-aerospace technical reports
37
Modern US records
ODNI/AARO reports, NASA study, hearing transcripts
24
FBI Vault UFO files
FBI records, 1947–1954
16
United Kingdom1,358 documents
Ministry of Defence UFO files
National Archives DEFE/AIR series, incl. Rendlesham Forest
1,358
France2,768 documents
GEIPAN (CNES)
France’s official UAP office — cases classified A/B/C/D
2,768
Canada30 documents
Library and Archives Canada
DND / Transport / NRC / RCMP files, incl. CIRVIS
30
Brazil231 documents
FAB / Arquivo Nacional
incl. Operação Prato (Colares, 1977) and SIOANI
231
Australia80 documents
National Archives of Australia / RAAF
declassified Australian UFO files
80

How it was built

Each collection follows the same grounded pipeline:

  1. Acquire the files from the official archive (national archives, FOIA releases, agency portals).
  2. Extract the text — directly where documents are born-digital, or via OCR for scanned paper (many of these are decades-old typed and handwritten reports).
  3. Chunk each document into passages, keeping page anchors so results point back to a location in the source.
  4. Embed every passage with a multilingual model (bge-m3) into a vector index, so search works by meaning, not just keywords.
  5. Retrieve with a hybrid of keyword (BM25) and semantic (vector) search, fused for relevance.
  6. Link every document back to its original source so you can verify it yourself.

What you can search

What this is — and isn’t

These are investigation, intelligence, and sighting files. There is no “alien” document class, because no such thing exists in the government record — so this index never implies more than the documents say. The files record what people reported and what officials concluded; a sighting report is not proof of anything, and government conclusions (identified, balloon, aircraft, unexplained) are presented as what the agency determined, not as ground truth.

Text is shown verbatim, including redactions and OCR artifacts. That’s deliberate: the point is to show exactly what the record says, warts and all.

Honest caveats