Two independent reputable sources, minimum — or it does not get published. Where the record is contested, we print DISPUTED and name the contested part. That rule is the whole business.
Because in most of the strange stories you have been told, the strangest part is the invented part. Not the whole story — the detail. The number. The dying quote. The motive. Strip that one element away and what is left is usually still true, still documented, and still worth telling. The reason so much real history feels fake is that the version that reached you was optimised for retelling, not for accuracy, and every retelling sanded one more caveat off it.
There are three different things happening under that one feeling, and it is worth being able to tell them apart:
The tell for all three is the same, and you can apply it in about ninety seconds: the single most quotable detail is the one to check first. If the story survives losing it, the story is probably real. If the story only exists because of it, be careful.
Nothing is published under this name unless it traces to at least two independent reputable sources. Reputable means one of four things:
Popular-history sites, encyclopaedia entries and magazine features are used only to corroborate a claim that already rests on one of those four. They never carry a claim on their own. Independent means the two sources do not both descend from the same original — two articles quoting the same 1970s paperback are one source wearing two coats.
And where the record is genuinely contested, the story says Disputed and names which part is contested. Not a disclaimer at the bottom. A flag on the sentence.
What survives checking: people danced compulsively in the street for weeks; the authorities concluded the cure was more dancing and cleared guildhalls and hired musicians. This sits in municipal records, physicians' notes and cathedral sermons of the period.
Disputed The cause and the death toll. Ergot poisoning is the popular explanation and is medically weak — ergotism causes convulsions and gangrene, not sustained coordinated dancing. The "fifteen people a day died" line traces to a single much later chronicle and is not supported by the 1518 sources. Which is exactly the pattern: the most repeatable number is the least supported one.
What survives checking: Australian soldiers were deployed to Western Australia with two Lewis guns and ten thousand rounds against emus damaging wheat crops; the birds scattered, a gun jammed early, and the operation was withdrawn without meeting its objective. That is in the operation's own records and in the contemporary press.
Disputed The kill count. The much-quoted "986 killed" is the commanding officer's own end-of-operation claim, never independently verified, and contemporary reports conflict. It is a claimed figure, not a counted one — and every viral version prints it as a counted one.
What survives checking: the broadcast happened, and some listeners were genuinely frightened — the surviving letters prove it.
Disputed The nationwide mass panic. Audience measurement from the night shows a comparatively small live audience, and the scholarship argues the panic was substantially a newspaper construction — the press had a commercial interest in portraying radio as dangerous. This is the rare case where the famous version is the invented one and the boring version is the record.
That is the entire method. It is not sophisticated. It is just rarely done, which is why so much confidently delivered history is wrong in one specific place.
The Rabbit Hole List — twelve true histories strange enough to sound invented, each with what is genuinely disputed about it and the specific first source to open. Free.
Get the Rabbit Hole List — free The collections
Every product carries its own sources appendix and its own disputed ledger — the list of every flagged claim in the volume, in one place, so you can see what was checked and what could not be settled.