How we reach a verdict
Every Odd Evidence investigation ends in one of four verdicts. This page defines what each one requires. It was published before our first investigation, deliberately — a standard written after the fact is not a standard.
The threshold for publishing at all
No investigation runs without at least one primary or peer-reviewed source. Not a blog post, not an aggregator, not a Wikipedia article — Wikipedia's citation, followed to its origin. If we cannot reach a primary source, the story does not run, however good it is.
This rule is enforced by our publishing system, not by memory: an investigation without a qualifying source will not build.
The four verdicts
Confirmed
Reliable evidence supports the central claim.
Mostly True
The story is real, but the popular version exaggerates or omits important context.
Unresolved
The event is documented, but the evidence does not establish a decisive explanation.
Myth
The popular claim is unsupported or contradicted by reliable evidence.
Written thresholds
Confirmed — multiple independent reliable sources support the central claim, and no substantial body of credible evidence contradicts it. Where a verdict applies to an event but not to its cause, we say so explicitly in the verdict note.
Mostly True — the underlying claim is supported, but the version in general circulation adds, omits or distorts something that changes how a reader would understand it. We name the specific distortion rather than gesturing at inaccuracy.
Unresolved — the event or observation is documented, but the available evidence does not establish a decisive explanation. This verdict is not a shrug. It requires that the phenomenon itself be well attested and that we set out what is actually known.
Myth — the popular claim is unsupported by reliable evidence, or contradicted by it. We apply this verdict to claims, never to the people who repeated them.
What we do not do
- We do not publish AI-written article text. Research assistance, source-finding and draft critique, yes. The prose and the judgement are human, because the entire premise of this site is that a person checked.
- We do not name or characterise living individuals when issuing a Myth verdict. The claim is the subject, not its carriers.
- We do not present a single confident cause where the evidence supports several.
- We do not quietly edit. See the corrections log.
Review dates
Every investigation carries a published date and a last-reviewed date. When new evidence changes an assessment, we update the investigation, move the review date, and log the change.