Every one of the 216 analyzed PURSUE documents was put through the same filter: witness credibility, sensor corroboration, specificity of the anomaly, and institutional routing. 174 did not survive it. These are the 28 that did, and this is what they agree on.
A note on yardsticks: these rate the strength of documents, not incidents. The Evidence Database rates each incident as a standalone case, so a document can stand out here for its specificity or provenance while the matching incident there still rates only circumstantial. The two pages weigh different things by design. And the filter cuts both ways: the 21 documents rated contested are examined, reason by reason, on The Contested Cases.
Noise to signal: roughly 7 to 1. About one document in eight rates as strong. This page is only that residue.
Where the strongest documents come from, in six clusters.
Six agencies, 1949–2026. Eighteen of the twenty-five are from 2020 or later; the rest anchor the same signatures back to the founding of the nuclear complex.
Cross-cutting patterns that appear independently across agencies, eras, and sensor types.
The skeptic's strongest objection is reporting bias: one culture seeing what it expects to see. The rebuttal is the grid below. The same behaviors recur across separate agencies, decades, and sensor types, with no document referencing another. That independence is the engine of the case.
| Recurring behavior | Dept of War | ODNI | FBI | State | Energy | Era span |
|---|---|---|---|---|---|---|
| Parent-child ejection | ● 2023, 2026 | ● 2025 | ● 2026 | 2023–2026 | ||
| Nuclear-adjacent geography | ● 1949, 2023 | ● 2025 | ● 2022 | ● modern | 1949–2025 | |
| 90-degree turns | ● 2023 | ● 1994 | 1994–2023 | |||
| Multi-sensor corroboration | ● 2023 | ● 2025 | 2023–2025 | |||
| Sphere with rigid appendage | ● 2024 | 2024 | ||||
| Official advanced-capability flag | ● 2023 | 2023 |
Two behaviors, parent-child ejection and nuclear-adjacent geography, each surface across three or four independent agencies. The CIA's four strong documents sit outside this behavioral grid by design: they are the documentary foundation (the 1952 Scientific Advisory Panel, Blue Book Special Report No. 14), not field observations.
The single most-documented case in the corpus. A restricted national-security site in the western United States, October 2023. Six trained observers across multiple vantage points, an AARO field investigation, five first-person written narratives, and a formal AARO unresolved-case analysis concluding that roughly 40 percent of the phenomena has no plausible explanation. The parent-child ejection signature is documented by five independent witnesses independently describing the same behavior.
Active-duty ISR crews and AFSOC platforms across six theatres: Greece, Syria, Iraq, UAE, East China Sea, Persian Gulf, and Kazakhstan. Trained military observers with calibrated sensors. Each report was formally filed through AARO and declassified under the PURSUE program. The kinematic signatures, 90-degree turns, co-altitude high-speed approaches, and disappearance from targeting pods, are documented independently in separate theatres with no shared observation chain. A Navy Range Fouler report (Tranche 4) adds the U.S. East Coast: a roughly 28-year veteran aviator's high-speed rectangular contact that outran the sensor field of view, captured on video.
The same event, documented twice: the USPER Statement (Tranche 1, SECRET//NOFORN, the handwritten account) and ODNI-UAP-D001 (Tranche 2, the prose narrative filed through ODNI channels). A senior US intelligence community official on a dedicated UAP investigation mission, over a western U.S. weapons test range, late 2025. FLIR, NVG, radar, and naked eye across multiple observer positions. An orb that split in two, approached the helicopter to within ten feet, and sped away.
Three modern FBI field reports rated Strong: the Cheyenne Mountain sighting (February 2022, multiple Army soldiers, daylight, motionless bean-shaped object) and two corroborating FD-302s from the Northeastern U.S. (February 2026, paired red orbs). Both the Cheyenne Mountain location, adjacent to the primary nuclear command center, and the paired-orb morphology in the FBI 2026 reports echo patterns found in the DoD and ODNI record.
A 116-page SECRET-classified Sandia Base correspondence file from 1949, documenting a sustained multi-agency investigation of unexplained aerial phenomena over the New Mexico nuclear complex. Tranche 4 supplies its founding primary source: the verbatim transcript of the 16 February 1949 Los Alamos conference, with Teller, Bradbury, Reines, and LaPaz in the room, concluding the green fireballs were most likely not conventional meteorites. And two Pantex Plant records from Texas, the highest-value nuclear weapons assembly facility in the country: the existing radar-tower image and, now, the full incident report behind it, a silent object with no identifiable propulsion that evaded the Protective Force. The nuclear-proximity association spans 77 years and two separate agencies.
Four CIA documents that define the institutional context rather than the phenomena themselves. The 1953 Robertson Panel recommended public debunking and surveillance of civilian UFO groups, and its recommendations are still traceable in the institutional behavior documented across the rest of the corpus. The U-2/OXCART history explains a documented share of Cold War-era sightings. Project Blue Book Special Report No. 14 found statistically that the better the report quality, the more likely it remained an "unknown." The Australian Defence assessment argues that the problem has genuine intelligence dimensions no allied government can assess alone.
The 28 Strong-rated documents establish, across six agencies and 77 years, that something is happening that trained observers cannot explain, that the institutional response has been consistent and documented, and that the kinematic signatures recur across independent observation chains with no shared personnel or equipment.
They do not prove what the phenomena are. They cannot, because no classified sensor data has been released, no physical sample has been publicly confirmed, and the redaction wall around the most specific reports remains intact. The Standout Cases are the strongest cards in a hand that does not yet include the trump card.
What they do prove is that the question is real, the witnesses are credible, the sensors are calibrated, and the institutional seriousness is documented. That is a different and more defensible claim than the one this field usually makes.