About this archive
A plain account of what we did, what we didn't do, and where to verify anything you read here against the original record.
Source
Every file in this archive originated from the U.S. Department of War's public release portal at war.gov/UFO, an interagency effort officially titled the Presidential Unsealing and Reporting System for UAP Encounters (PURSUE). We are not affiliated with the U.S. government. The original files are their work; we have only restructured them.
Every incident page links to the original document on war.gov so you can verify any claim against the source.
Preventing hallucination
The structured fields you see in case files — date, location, branch, sensor type, summary — are produced by a language model reading the source text. To prevent fabricated incidents from appearing here, every extracted incident must include a verbatim excerpt that we substring-validate against the original document. If the quote isn't actually in the source, the incident is dropped. The validator was tested against real source text plus deliberately fabricated rows; it preserved every real quote and dropped every fabricated one.
When in doubt, the verbatim excerpt is ground truth — not the structured fields. And when in further doubt, click through to the original PDF.
What we do not claim
This site does not claim that UAPs are extraterrestrial, that the U.S. government has recovered alien technology, or that any single incident has a paranormal explanation. The Pentagon's All-domain Anomaly Resolution Office (AARO), which reviewed many of these incidents officially, has reached no such conclusion either. The dataset is observational — what was seen, by whom, with what sensors, and what remained unresolved.
Limits and known gaps
The model occasionally misclassifies sensor types or branches when the source is ambiguous. We mark such fields as unknown rather than guess.
Heavily redacted documents may produce incidents with very sparse fields. They are still listed, because their existence is itself signal.
Roughly 60% of incidents could be placed on the map; the rest lack precise coordinates in the source text and appear in lists but not on the map. Older FBI case files often describe sightings without specific geographic detail.
Videos are linked, not transcribed. Their official war.gov captions appear in their place.