Skip to main content

What the PURSUE files conclude

How many of the declassified files explain what was seen, how many leave it unresolved, and what the unresolved ones rest on. Every number counts only what the record itself states — the agency, the investigators, a witness or a document it cites — not our opinion. How this was counted. For the overall statistics, see PURSUE by the numbers.

Of 351 incident reports in the PURSUE files, 24 (7%) are explained in the record, 187 (53%) remain unresolved, 24 are under investigation and 116 state no conclusion. 127 files are not single incident reports and are left out.

  • Explained in the record
  • Unresolved
  • Under investigation
  • No conclusion stated
  1. All incident reports351 files

FBI has the largest share of files left unresolved or under investigation: 72% of its 69 files.

  • Explained in the record
  • Unresolved
  • Under investigation
  • No conclusion stated
  1. Department of War225 files
  2. FBI69 files
  3. NASA22 files
  4. CIA15 files
  5. Other agencies20 files

Bars show each group's share; the number of files is on the right.

Files about incidents in the 1970s are the most often explained in the record: 58% of 12 files.

  • Explained in the record
  • Unresolved
  • Under investigation
  • No conclusion stated
  1. 1940s8 files
  2. 1950s30 files
  3. 1960s27 files
  4. 1970s12 files
  5. 1990s5 files
  6. 2010s18 files
  7. 2020s195 files
  8. Other decades4 files
  9. Date unknown52 files

Bars show each group's share; the number of files is on the right. Decades with fewer than 5 reports are grouped; the year is read from war.gov's free-text incident date.

91 files name at least one mundane explanation; the most common is other (28 files).

  • Named in files explained in the record
  • Named in other files
  1. Other28 files
  2. Aircraft or drone22 files
  3. Sensor or optical artifact18 files
  4. Star, planet or meteor15 files
  5. Balloon14 files
  6. Satellite or space object8 files
  7. Weather or atmospheric8 files
  8. Bird5 files
  9. Hoax2 files

An explanation counts when the record mentions it, even if the record then rejects it — so a file explained in the record may also name alternatives it ruled out.

49 of 211 unresolved files (23%) rest on witness testimony alone; 13 rest on more than one sensor.

  • Several sensors
  • One sensor or recording
  • Witness testimony only
  • No observation
  1. Explained in the record24 files
  2. Unresolved or under investigation211 files

Bars show each group's share; the number of files is on the right.

Cases

Each case page brings together the files about one incident. Its label sums up what those files conclude.

Unresolved 55

All files by conclusion

Every classified file, with one sentence on what the record states.

Explained in the record 24
Unresolved 187
Under investigation 24
No conclusion stated 116
Not a single incident report 127

How this was counted

Each file was read by Claude Haiku, an AI model, from our analysis of that file: its summary, key findings, stated conclusion and credibility notes. It recorded what the record itself concludes, which mundane explanations the record raises, and what the account rests on. Explanations that only our own analysis suggests are not counted.

“Explained” means the record presents a mundane cause as its assessment, even when hedged as probable. “Assessed as benign” without an identification counts as unresolved: it is a threat call, not an explanation. A record AARO filed on its “Unresolved UAP Report” form, or that says the operator was unable to positively identify the object, counts as unresolved even when it also disclaims any analytical judgment. Files that report no single incident — papers, letters, artwork and compilations of many reports — are left out of the shares.

The classification is AI-assisted and can be wrong. If a file is misread, please tell us. The verdicts are also columns in the open PURSUE dataset.