Methodology & Disclaimer

1 · The six rules we hold ourselves to

1. Claims, not facts. Every stored and displayed item is attributed to a person and a source. We never present a claim as verified.
2. Corroboration means independent people, not mentions. One person repeated across thirty interviews is a single source. When one speaker is plainly restating another, we flag it as derivative and it does not count as independent corroboration.
3. Everything is grounded. A claim exists only if it has a verbatim quote present in the source text and a resolvable link to that exact moment. No quote and locator, no claim.
4. The brief is a SIMULATION. The disclosure brief is a synthesis of attributed claims, labeled as a simulation everywhere it appears. It is not a real government document.
5. Attributed testimony, never accusation. Many claims allege wrongdoing by named, living people and agencies. We render these strictly as "X said…," linked to the source — never as this site's own assertion.
6. The method is transparent. Every scoring input — independent-speaker count, source credibility, specificity, contradictions — is shown on the claim, so you can judge for yourself.

2 · Where the material comes from

The corpus is built from public interviews, books, academic papers, and government documents in which vetted "insiders" — former intelligence and military officials, program veterans, scientists, and firsthand witnesses — discuss UAP, non-human intelligence, and government secrecy. Video interviews are transcribed from captions; books and documents are ingested as page- or section-anchored text.

A roster of scored insiders is maintained on the Sources & Roster page, each assigned a credibility tier reflecting how directly they would plausibly have knowledge. Every source is recorded with its type, publisher, and a reliability grade.

3 · How a claim is made

Each source is read in overlapping windows. For every disclosure-relevant assertion we record: a neutral paraphrase; the verbatim quote it rests on; the topic, stance (assertion / denial / hedge), and specificity (vague / moderate / specific); the speaker; and the basis — primary (the speaker's own firsthand claim) or secondary (the speaker relaying someone else, whose name we capture). The quote is matched back to the source to attach a timestamp or page and a deep link. Anything that cannot be grounded is dropped.

A limitation we do not hide: transcripts have no speaker labels. When a source has exactly one roster name, its claims are attributed to that person. When several people appear (a panel or debate), the claim is marked speaker-uncertain and attributed at the source level. We do not fabricate certainty about who said what.

4 · Clustering & the independence rule

Equivalent raw claims are grouped into a single canonical claim with a neutral statement. Its independent-speaker count is the number of distinct people asserting it, excluding anyone marked derivative. This is the number that matters: a claim asserted by eight people who each independently say they know it is very different from a claim eight people heard from the same one person.

5 · Scoring, in the open

Each canonical claim carries a credibility score (a tier-weighted sum over its distinct independent speakers), combined with specificity and any contradiction into a corroboration score. Claims are then placed in a bucket:

• strongly corroborated — many independent, higher-tier speakers; specific; uncontradicted.
• multiply asserted — several independent speakers, but thinner or less specific.
• single source — asserted independently by only one person.
• contested — has a direct contradiction elsewhere in the testimony.

The buckets, counts, and every input are shown on the Data page. We publish the formula rather than hide it.

6 · Contradictions

Within a topic we look for canonical claims that oppose each other on the same proposition (for example, "biological remains were recovered" versus "no biologics exist"). These are linked as contradictions and flagged, while taking care not to mistake differences of scope or wording for genuine disagreement.

7 · The simulated brief

The disclosure brief synthesizes the best-corroborated testimony into a single narrative — "if formal disclosure occurred, the collected testimony points to…" — with every line footnoted back to canonical claims, their sources, and deep links. It is produced from the corpus on each publish and tied to a snapshot version. It is a simulation. It represents what this body of testimony collectively asserts, not a factual account and not a leaked document.

8 · Known coverage gaps

The corpus is not exhaustive, and we say so plainly:

• Some interviews are members-only or paywalled and have no transcript available.
• Several roster names have little or no presence in the sources we could gather, because they appear mainly in venues outside this collection.
• Most video transcripts are auto-generated captions: no punctuation guarantees, no speaker labels, and occasional transcription errors. Extraction tolerates messy text, but errors are possible.

As new material is ingested, the archive is regenerated and re-scored, and a new snapshot is published — so the "best guess" can change over time as the evidence base grows.

9 · Disclaimer

IfTheyToldUs is an independent research and media project of the Emergent Frontiers Institute. It reports what others have said; it does not endorse, verify, or vouch for the truth of any claim. References to specific people, agencies, programs, and companies reflect attributed statements made by the cited speakers and are presented for documentation and public interest. They are not assertions of fact by this site. If you are named in a source and wish to respond or correct the record, please contact us.

Content is provided "as is," without warranty. Deep links point to third-party platforms whose content we do not control and which may change or be removed.